update to resolve jmorganca comments

Signed-off-by: Matt Williams <m@technovangelist.com>

.dockerignore

@@ -2,7 +2,7 @@
 ollama
 app
 dist
-scripts
-llm/llama.cpp/ggml
 llm/llama.cpp/gguf
 .env
+.cache
+test_data

.gitignore

@@ -6,3 +6,7 @@
 dist
 ollama
 ggml-metal.metal
+.cache
+*.exe
+.idea
+test_data

.gitmodules

@@ -1,8 +1,3 @@
-[submodule "llm/llama.cpp/ggml"]
-    path = llm/llama.cpp/ggml
-    url = https://github.com/ggerganov/llama.cpp.git
-    ignore = dirty
-    shallow = true
 [submodule "llm/llama.cpp/gguf"]
     path = llm/llama.cpp/gguf
     url = https://github.com/ggerganov/llama.cpp.git

Dockerfile

@@ -5,8 +5,8 @@ ARG GOFLAGS="'-ldflags=-w -s'"
 
 WORKDIR /go/src/github.com/jmorganca/ollama
 RUN apt-get update && apt-get install -y git build-essential cmake
-ADD https://dl.google.com/go/go1.21.1.linux-$TARGETARCH.tar.gz /tmp/go1.21.1.tar.gz
-RUN mkdir -p /usr/local && tar xz -C /usr/local </tmp/go1.21.1.tar.gz
+ADD https://dl.google.com/go/go1.21.3.linux-$TARGETARCH.tar.gz /tmp/go1.21.3.tar.gz
+RUN mkdir -p /usr/local && tar xz -C /usr/local </tmp/go1.21.3.tar.gz
 
 COPY . .
 ENV GOARCH=$TARGETARCH
@@ -19,5 +19,11 @@ RUN apt-get update && apt-get install -y ca-certificates
 COPY --from=0 /go/src/github.com/jmorganca/ollama/ollama /bin/ollama
 EXPOSE 11434
 ENV OLLAMA_HOST 0.0.0.0
+
+# set some environment variable for better NVIDIA compatibility
+ENV PATH=/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
+ENV LD_LIBRARY_PATH=/usr/local/nvidia/lib:/usr/local/nvidia/lib64
+ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility
+
 ENTRYPOINT ["/bin/ollama"]
 CMD ["serve"]

Dockerfile.build

@@ -1,24 +1,65 @@
+# Ubuntu 20.04 amd64 dependencies
+FROM --platform=linux/amd64 ubuntu:20.04 AS base-amd64
+ARG CUDA_VERSION=11.3.1-1
+ARG CMAKE_VERSION=3.22.1
+# ROCm only supports amd64
+ARG ROCM_VERSION=6.0
+ARG CLBLAST_VER=1.6.1
 
-# centos7 amd64 dependencies
-FROM --platform=linux/amd64 nvidia/cuda:11.8.0-devel-centos7 AS base-amd64
-RUN yum install -y https://repo.ius.io/ius-release-el7.rpm centos-release-scl && \
-    yum update -y && \
-    yum install -y devtoolset-10-gcc devtoolset-10-gcc-c++ git236 wget
-RUN wget "https://github.com/Kitware/CMake/releases/download/v3.27.6/cmake-3.27.6-linux-x86_64.sh" -O cmake-installer.sh && chmod +x cmake-installer.sh && ./cmake-installer.sh --skip-license --prefix=/usr/local
-ENV PATH /opt/rh/devtoolset-10/root/usr/bin:$PATH
+# Note: https://rocm.docs.amd.com/en/latest/release/user_kernel_space_compat_matrix.html
+RUN apt-get update && \
+    apt-get install -y wget gnupg && \
+    wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin && \
+    mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600 && \
+    apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/3bf863cc.pub && \
+    echo "deb [by-hash=no] https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/ /" > /etc/apt/sources.list.d/cuda.list && \
+    wget "https://github.com/Kitware/CMake/releases/download/v${CMAKE_VERSION}/cmake-${CMAKE_VERSION}-linux-x86_64.sh" -O /tmp/cmake-installer.sh && \
+    chmod +x /tmp/cmake-installer.sh && /tmp/cmake-installer.sh --skip-license --prefix=/usr && \
+    mkdir --parents --mode=0755 /etc/apt/keyrings && \
+    wget https://repo.radeon.com/rocm/rocm.gpg.key -O - | gpg --dearmor > /etc/apt/keyrings/rocm.gpg && \
+    echo "deb [arch=amd64 signed-by=/etc/apt/keyrings/rocm.gpg] https://repo.radeon.com/rocm/apt/${ROCM_VERSION} focal main" > /etc/apt/sources.list.d/rocm.list && \
+    echo "Package: *" > /etc/apt/preferences.d/rocm-pin-600 && \
+    echo "Pin: release o=repo.radeon.com" >> /etc/apt/preferences.d/rocm-pin-600 && \
+    echo "Pin-Priority: 600" >> /etc/apt/preferences.d/rocm-pin-600 && \
+    apt-get update && \
+    DEBIAN_FRONTEND=noninteractive apt-get -y install cuda=${CUDA_VERSION} rocm-hip-libraries rocm-device-libs rocm-libs rocm-ocl-icd rocm-hip-sdk rocm-hip-libraries rocm-cmake rocm-clang-ocl rocm-dev
 
-# centos8 arm64 dependencies
-FROM --platform=linux/arm64 nvidia/cuda:11.4.3-devel-centos8 AS base-arm64
-RUN sed -i -e 's/mirrorlist/#mirrorlist/g' -e 's|#baseurl=http://mirror.centos.org|baseurl=http://vault.centos.org|g' /etc/yum.repos.d/CentOS-*
-RUN yum install -y git cmake
+# CLBlast
+RUN wget -qO- https://github.com/CNugteren/CLBlast/archive/refs/tags/${CLBLAST_VER}.tar.gz | tar zxv -C /tmp/ && \
+    cd /tmp/CLBlast-${CLBLAST_VER} && mkdir build && cd build && cmake .. && make && make install
+
+ENV ROCM_PATH=/opt/rocm
+
+# Ubuntu 22.04 arm64 dependencies
+FROM --platform=linux/arm64 ubuntu:20.04 AS base-arm64
+ARG CUDA_VERSION=11.3.1-1
+ARG CMAKE_VERSION=3.27.6
+RUN apt-get update && \
+    apt-get install -y wget gnupg && \
+    wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/sbsa/cuda-ubuntu2004.pin && \
+    mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600 && \
+    apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/sbsa//3bf863cc.pub && \
+    echo "deb [by-hash=no] https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/sbsa/ /" > /etc/apt/sources.list.d/cuda.list && \
+    wget "https://github.com/Kitware/CMake/releases/download/v${CMAKE_VERSION}/cmake-${CMAKE_VERSION}-linux-aarch64.sh" -O /tmp/cmake-installer.sh && \
+    chmod +x /tmp/cmake-installer.sh && /tmp/cmake-installer.sh --skip-license --prefix=/usr && \
+    apt-get update && \
+    apt-cache madison cuda && \
+    DEBIAN_FRONTEND=noninteractive apt-get -y install cuda=${CUDA_VERSION} 
 
 FROM base-${TARGETARCH}
 ARG TARGETARCH
 ARG GOFLAGS="'-ldflags -w -s'"
+ARG CGO_CFLAGS
+ARG GOLANG_VERSION=1.21.3
+
+# Common toolchain
+RUN apt-get update && \
+    DEBIAN_FRONTEND=noninteractive apt-get install -y gcc-10 g++-10 cpp-10 git ocl-icd-opencl-dev && \
+    update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-10 100 --slave /usr/bin/g++ g++ /usr/bin/g++-10 --slave /usr/bin/gcov gcov /usr/bin/gcov-10
 
 # install go
-ADD https://dl.google.com/go/go1.21.1.linux-$TARGETARCH.tar.gz /tmp/go1.21.1.tar.gz
-RUN mkdir -p /usr/local && tar xz -C /usr/local </tmp/go1.21.1.tar.gz
+ADD https://dl.google.com/go/go${GOLANG_VERSION}.linux-$TARGETARCH.tar.gz /tmp/go${GOLANG_VERSION}.tar.gz
+RUN mkdir -p /usr/local && tar xz -C /usr/local </tmp/go${GOLANG_VERSION}.tar.gz
 
 # build the final binary
 WORKDIR /go/src/github.com/jmorganca/ollama
@@ -27,6 +68,7 @@ COPY . .
 ENV GOOS=linux
 ENV GOARCH=$TARGETARCH
 ENV GOFLAGS=$GOFLAGS
+ENV CGO_CFLAGS=${CGO_CFLAGS}
 
 RUN /usr/local/go/bin/go generate ./... && \
     /usr/local/go/bin/go build .

README.md

@@ -15,6 +15,10 @@ Get up and running with large language models locally.
 
 [Download](https://ollama.ai/download/Ollama-darwin.zip)
 
+### Windows
+
+Coming soon! For now, you can install Ollama on Windows via WSL2.
+
 ### Linux & WSL2
 
 ```
@@ -23,9 +27,9 @@ curl https://ollama.ai/install.sh | sh
 
 [Manual install instructions](https://github.com/jmorganca/ollama/blob/main/docs/linux.md)
 
-### Windows
+### Docker
 
-coming soon
+The official [Ollama Docker image](https://hub.docker.com/r/ollama/ollama) `ollama/ollama` is available on Docker Hub.
 
 ## Quickstart
 
@@ -43,24 +47,29 @@ Here are some example open-source models that can be downloaded:
 
 | Model              | Parameters | Size  | Download                       |
 | ------------------ | ---------- | ----- | ------------------------------ |
-| Mistral            | 7B         | 4.1GB | `ollama run mistral`           |
 | Llama 2            | 7B         | 3.8GB | `ollama run llama2`            |
+| Mistral            | 7B         | 4.1GB | `ollama run mistral`           |
+| Dolphin Phi        | 2.7B       | 1.6GB | `ollama run dolphin-phi`       |
+| Phi-2              | 2.7B       | 1.7GB | `ollama run phi`               |
+| Neural Chat        | 7B         | 4.1GB | `ollama run neural-chat`       |
+| Starling           | 7B         | 4.1GB | `ollama run starling-lm`       |
 | Code Llama         | 7B         | 3.8GB | `ollama run codellama`         |
 | Llama 2 Uncensored | 7B         | 3.8GB | `ollama run llama2-uncensored` |
 | Llama 2 13B        | 13B        | 7.3GB | `ollama run llama2:13b`        |
 | Llama 2 70B        | 70B        | 39GB  | `ollama run llama2:70b`        |
 | Orca Mini          | 3B         | 1.9GB | `ollama run orca-mini`         |
 | Vicuna             | 7B         | 3.8GB | `ollama run vicuna`            |
+| LLaVA              | 7B         | 4.5GB | `ollama run llava`             |
 
-> Note: You should have at least 8 GB of RAM to run the 3B models, 16 GB to run the 7B models, and 32 GB to run the 13B models.
+> Note: You should have at least 8 GB of RAM available to run the 7B models, 16 GB to run the 13B models, and 32 GB to run the 33B models.
 
-## Customize your own model
+## Customize a model
 
-### Import from GGUF or GGML
+### Import from GGUF
 
-Ollama supports importing GGUF and GGML file formats in the Modelfile. This means if you have a model that is not in the Ollama library, you can create it, iterate on it, and upload it to the Ollama library to share with others when you are ready.
+Ollama supports importing GGUF models in the Modelfile:
 
-1. Create a file named Modelfile, and add a `FROM` instruction with the local filepath to the model you want to import.
+1. Create a file named `Modelfile`, with a `FROM` instruction with the local filepath to the model you want to import.
 
    ```
    FROM ./vicuna-33b.Q4_0.gguf
@@ -69,18 +78,22 @@ Ollama supports importing GGUF and GGML file formats in the Modelfile. This mean
 2. Create the model in Ollama
 
    ```
-   ollama create name -f path_to_modelfile
+   ollama create example -f Modelfile
    ```
 
 3. Run the model
 
    ```
-   ollama run name
+   ollama run example
    ```
 
+### Import from PyTorch or Safetensors
+
+See the [guide](docs/import.md) on importing models for more information.
+
 ### Customize a prompt
 
-Models from the Ollama library can be customized with a prompt. The example
+Models from the Ollama library can be customized with a prompt. For example, to customize the `llama2` model:
 
 ```
 ollama pull llama2
@@ -94,7 +107,7 @@ FROM llama2
 # set the temperature to 1 [higher is more creative, lower is more coherent]
 PARAMETER temperature 1
 
-# set the system prompt
+# set the system message
 SYSTEM """
 You are Mario from Super Mario Bros. Answer as Mario, the assistant, only.
 """
@@ -117,6 +130,10 @@ For more examples, see the [examples](examples) directory. For more information
 
 `ollama create` is used to create a model from a Modelfile.
 
+```
+ollama create mymodel -f ./Modelfile
+```
+
 ### Pull a model
 
 ```
@@ -148,10 +165,17 @@ For multiline input, you can wrap text with `"""`:
 I'm a basic program that prints the famous "Hello, world!" message to the console.
 ```
 
+### Multimodal models
+
+```
+>>> What's in this image? /Users/jmorgan/Desktop/smile.png
+The image features a yellow smiley face, which is likely the central focus of the picture.
+```
+
 ### Pass in prompt as arguments
 
 ```
-$ ollama run llama2 "summarize this file:" "$(cat README.md)"
+$ ollama run llama2 "Summarize this file: $(cat README.md)"
  Ollama is a lightweight, extensible framework for building and running language models on the local machine. It provides a simple API for creating, running, and managing models, as well as a library of pre-built models that can be easily used in a variety of applications.
 ```
 
@@ -170,17 +194,22 @@ ollama list
 Install `cmake` and `go`:
 
 ```
-brew install cmake
-brew install go
+brew install cmake go
 ```
 
-Then generate dependencies and build:
-
+Then generate dependencies:
 ```
 go generate ./...
+```
+Then build the binary:
+```
 go build .
 ```
 
+More detailed instructions can be found in the [developer guide](https://github.com/jmorganca/ollama/blob/main/docs/development.md)
+
+
+### Running local builds
 Next, start the server:
 
 ```
@@ -195,30 +224,98 @@ Finally, in a separate shell, run a model:
 
 ## REST API
 
-> See the [API documentation](docs/api.md) for all endpoints.
+Ollama has a REST API for running and managing models.
 
-Ollama has an API for running and managing models. For example to generate text from a model:
+### Generate a response
 
 ```
-curl -X POST http://localhost:11434/api/generate -d '{
+curl http://localhost:11434/api/generate -d '{
   "model": "llama2",
   "prompt":"Why is the sky blue?"
 }'
 ```
 
+### Chat with a model
+
+```
+curl http://localhost:11434/api/chat -d '{
+  "model": "mistral",
+  "messages": [
+    { "role": "user", "content": "why is the sky blue?" }
+  ]
+}'
+```
+
+See the [API documentation](./docs/api.md) for all endpoints.
+
 ## Community Integrations
 
+### Web & Desktop
+- [Bionic GPT](https://github.com/bionic-gpt/bionic-gpt)
+- [HTML UI](https://github.com/rtcfirefly/ollama-ui)
+- [Chatbot UI](https://github.com/ivanfioravanti/chatbot-ollama)
+- [Typescript UI](https://github.com/ollama-interface/Ollama-Gui?tab=readme-ov-file)
+- [Minimalistic React UI for Ollama Models](https://github.com/richawo/minimal-llm-ui)
+- [Web UI](https://github.com/ollama-webui/ollama-webui)
+- [Ollamac](https://github.com/kevinhermawan/Ollamac)
+- [big-AGI](https://github.com/enricoros/big-agi/blob/main/docs/config-ollama.md)
+- [Cheshire Cat assistant framework](https://github.com/cheshire-cat-ai/core)
+- [Amica](https://github.com/semperai/amica)
+- [chatd](https://github.com/BruceMacD/chatd)
+- [Ollama-SwiftUI](https://github.com/kghandour/Ollama-SwiftUI)
+
+
+### Terminal
+
+- [oterm](https://github.com/ggozad/oterm)
+- [Ellama Emacs client](https://github.com/s-kostyaev/ellama)
+- [Emacs client](https://github.com/zweifisch/ollama)
+- [gen.nvim](https://github.com/David-Kunz/gen.nvim)
+- [ollama.nvim](https://github.com/nomnivore/ollama.nvim)
+- [ogpt.nvim](https://github.com/huynle/ogpt.nvim)
+- [gptel Emacs client](https://github.com/karthink/gptel)
+- [Oatmeal](https://github.com/dustinblackman/oatmeal)
+- [cmdh](https://github.com/pgibler/cmdh)
+
+### Database
+
+- [MindsDB](https://github.com/mindsdb/mindsdb/blob/staging/mindsdb/integrations/handlers/ollama_handler/README.md)
+
+### Package managers
+
+- [Pacman](https://archlinux.org/packages/extra/x86_64/ollama/)
+
+### Libraries
+
 - [LangChain](https://python.langchain.com/docs/integrations/llms/ollama) and [LangChain.js](https://js.langchain.com/docs/modules/model_io/models/llms/integrations/ollama) with [example](https://js.langchain.com/docs/use_cases/question_answering/local_retrieval_qa)
+- [LangChainGo](https://github.com/tmc/langchaingo/) with [example](https://github.com/tmc/langchaingo/tree/main/examples/ollama-completion-example)
 - [LlamaIndex](https://gpt-index.readthedocs.io/en/stable/examples/llm/ollama.html)
+- [LiteLLM](https://github.com/BerriAI/litellm)
+- [OllamaSharp for .NET](https://github.com/awaescher/OllamaSharp)
+- [Ollama-rs for Rust](https://github.com/pepperoni21/ollama-rs)
+- [Ollama4j for Java](https://github.com/amithkoujalgi/ollama4j)
+- [ModelFusion Typescript Library](https://modelfusion.dev/integration/model-provider/ollama)
+- [OllamaKit for Swift](https://github.com/kevinhermawan/OllamaKit)
+- [Ollama for Dart](https://github.com/breitburg/dart-ollama)
+- [Ollama for Laravel](https://github.com/cloudstudio/ollama-laravel)
+- [LangChainDart](https://github.com/davidmigloz/langchain_dart)
+
+### Mobile
+
+- [Enchanted](https://github.com/AugustDev/enchanted)
+- [Maid](https://github.com/Mobile-Artificial-Intelligence/maid)
+
+### Extensions & Plugins
+
 - [Raycast extension](https://github.com/MassimilianoPasquini97/raycast_ollama)
 - [Discollama](https://github.com/mxyng/discollama) (Discord bot inside the Ollama discord channel)
 - [Continue](https://github.com/continuedev/continue)
 - [Obsidian Ollama plugin](https://github.com/hinterdupfinger/obsidian-ollama)
+- [Logseq Ollama plugin](https://github.com/omagdy7/ollama-logseq)
 - [Dagger Chatbot](https://github.com/samalba/dagger-chatbot)
-- [LiteLLM](https://github.com/BerriAI/litellm)
 - [Discord AI Bot](https://github.com/mekb-turtle/discord-ai-bot)
-- [Chatbot UI](https://github.com/ivanfioravanti/chatbot-ollama)
-- [HTML UI](https://github.com/rtcfirefly/ollama-ui)
-- [Typescript UI](https://github.com/ollama-interface/Ollama-Gui?tab=readme-ov-file)
-- [Dumbar](https://github.com/JerrySievert/Dumbar)
-- [Emacs client](https://github.com/zweifisch/ollama)
+- [Ollama Telegram Bot](https://github.com/ruecat/ollama-telegram)
+- [Hass Ollama Conversation](https://github.com/ej52/hass-ollama-conversation)
+- [Rivet plugin](https://github.com/abrenneke/rivet-plugin-ollama)
+- [Llama Coder](https://github.com/ex3ndr/llama-coder) (Copilot alternative using Ollama)
+- [Obsidian BMO Chatbot plugin](https://github.com/longy2k/obsidian-bmo-chatbot)

api/client.go

@@ -5,6 +5,7 @@ import (
 	"bytes"
 	"context"
 	"encoding/json"
+	"errors"
 	"fmt"
 	"io"
 	"net"
@@ -14,13 +15,10 @@ import (
 	"runtime"
 	"strings"
 
+	"github.com/jmorganca/ollama/format"
 	"github.com/jmorganca/ollama/version"
 )
 
-const DefaultHost = "127.0.0.1:11434"
-
-var envHost = os.Getenv("OLLAMA_HOST")
-
 type Client struct {
 	base *url.URL
 	http http.Client
@@ -43,16 +41,28 @@ func checkError(resp *http.Response, body []byte) error {
 }
 
 func ClientFromEnvironment() (*Client, error) {
+	defaultPort := "11434"
+
 	scheme, hostport, ok := strings.Cut(os.Getenv("OLLAMA_HOST"), "://")
-	if !ok {
+	switch {
+	case !ok:
 		scheme, hostport = "http", os.Getenv("OLLAMA_HOST")
+	case scheme == "http":
+		defaultPort = "80"
+	case scheme == "https":
+		defaultPort = "443"
 	}
 
+	// trim trailing slashes
+	hostport = strings.TrimRight(hostport, "/")
+
 	host, port, err := net.SplitHostPort(hostport)
 	if err != nil {
-		host, port = "127.0.0.1", "11434"
-		if ip := net.ParseIP(strings.Trim(os.Getenv("OLLAMA_HOST"), "[]")); ip != nil {
+		host, port = "127.0.0.1", defaultPort
+		if ip := net.ParseIP(strings.Trim(hostport, "[]")); ip != nil {
 			host = ip.String()
+		} else if hostport != "" {
+			host = hostport
 		}
 	}
 
@@ -63,7 +73,7 @@ func ClientFromEnvironment() (*Client, error) {
 		},
 	}
 
-	mockRequest, err := http.NewRequest("HEAD", client.base.String(), nil)
+	mockRequest, err := http.NewRequest(http.MethodHead, client.base.String(), nil)
 	if err != nil {
 		return nil, err
 	}
@@ -86,11 +96,19 @@ func (c *Client) do(ctx context.Context, method, path string, reqData, respData
 	var reqBody io.Reader
 	var data []byte
 	var err error
-	if reqData != nil {
+
+	switch reqData := reqData.(type) {
+	case io.Reader:
+		// reqData is already an io.Reader
+		reqBody = reqData
+	case nil:
+		// noop
+	default:
 		data, err = json.Marshal(reqData)
 		if err != nil {
 			return err
 		}
+
 		reqBody = bytes.NewReader(data)
 	}
 
@@ -127,7 +145,7 @@ func (c *Client) do(ctx context.Context, method, path string, reqData, respData
 	return nil
 }
 
-const maxBufferSize = 512 * 1000 // 512KB
+const maxBufferSize = 512 * format.KiloByte
 
 func (c *Client) stream(ctx context.Context, method, path string, data any, fn func([]byte) error) error {
 	var buf *bytes.Buffer
@@ -203,6 +221,19 @@ func (c *Client) Generate(ctx context.Context, req *GenerateRequest, fn Generate
 	})
 }
 
+type ChatResponseFunc func(ChatResponse) error
+
+func (c *Client) Chat(ctx context.Context, req *ChatRequest, fn ChatResponseFunc) error {
+	return c.stream(ctx, http.MethodPost, "/api/chat", req, func(bts []byte) error {
+		var resp ChatResponse
+		if err := json.Unmarshal(bts, &resp); err != nil {
+			return err
+		}
+
+		return fn(resp)
+	})
+}
+
 type PullProgressFunc func(ProgressResponse) error
 
 func (c *Client) Pull(ctx context.Context, req *PullRequest, fn PullProgressFunc) error {
@@ -278,3 +309,30 @@ func (c *Client) Heartbeat(ctx context.Context) error {
 	}
 	return nil
 }
+
+func (c *Client) CreateBlob(ctx context.Context, digest string, r io.Reader) error {
+	if err := c.do(ctx, http.MethodHead, fmt.Sprintf("/api/blobs/%s", digest), nil, nil); err != nil {
+		var statusError StatusError
+		if !errors.As(err, &statusError) || statusError.StatusCode != http.StatusNotFound {
+			return err
+		}
+
+		if err := c.do(ctx, http.MethodPost, fmt.Sprintf("/api/blobs/%s", digest), r, nil); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
+func (c *Client) Version(ctx context.Context) (string, error) {
+	var version struct {
+		Version string `json:"version"`
+	}
+
+	if err := c.do(ctx, http.MethodGet, "/api/version", nil, &version); err != nil {
+		return "", err
+	}
+
+	return version.Version, nil
+}

api/client.py

@@ -1,13 +1,17 @@
 import os
 import json
 import requests
+import os
+import hashlib
+import json
+from pathlib import Path
 
 BASE_URL = os.environ.get('OLLAMA_HOST', 'http://localhost:11434')
 
 # Generate a response for a given prompt with a provided model. This is a streaming endpoint, so will be a series of responses.
 # The final response object will include statistics and additional data from the request. Use the callback function to override
 # the default handler.
-def generate(model_name, prompt, system=None, template=None, context=None, options=None, callback=None):
+def generate(model_name, prompt, system=None, template=None, format="", context=None, options=None, callback=None):
     try:
         url = f"{BASE_URL}/api/generate"
         payload = {
@@ -16,7 +20,8 @@ def generate(model_name, prompt, system=None, template=None, context=None, optio
             "system": system, 
             "template": template, 
             "context": context, 
-            "options": options
+            "options": options,
+            "format": format,
         }
         
         # Remove keys with None values
@@ -56,30 +61,86 @@ def generate(model_name, prompt, system=None, template=None, context=None, optio
     except requests.exceptions.RequestException as e:
         print(f"An error occurred: {e}")
         return None, None
+    
+
+# Create a blob file on the server if it doesn't exist.
+def create_blob(digest, file_path):
+    url = f"{BASE_URL}/api/blobs/{digest}"
+
+    # Check if the blob exists
+    response = requests.head(url)
+    if response.status_code != 404:
+        return  # Blob already exists, no need to upload
+    response.raise_for_status()
+
+    # Upload the blob
+    with open(file_path, 'rb') as file_data:
+        requests.post(url, data=file_data)
+
 
 # Create a model from a Modelfile. Use the callback function to override the default handler.
-def create(model_name, model_path, callback=None):
+def create(model_name, filename, callback=None):
     try:
+        file_path = Path(filename).expanduser().resolve()
+        processed_lines = []
+
+        # Read and process the modelfile
+        with open(file_path, 'r') as f:
+            for line in f:            
+                # Skip empty or whitespace-only lines
+                if not line.strip():
+                    continue
+            
+                command, args = line.split(maxsplit=1)
+
+                if command.upper() in ["FROM", "ADAPTER"]:
+                    path = Path(args.strip()).expanduser()
+
+                    # Check if path is relative and resolve it
+                    if not path.is_absolute():
+                        path = (file_path.parent / path)
+
+                    # Skip if file does not exist for "model", this is handled by the server
+                    if not path.exists():
+                        processed_lines.append(line)
+                        continue
+
+                    # Calculate SHA-256 hash
+                    with open(path, 'rb') as bin_file:
+                        hash = hashlib.sha256()
+                        hash.update(bin_file.read())
+                        blob = f"sha256:{hash.hexdigest()}"
+                
+                    # Add the file to the remote server
+                    create_blob(blob, path)
+
+                    # Replace path with digest in the line
+                    line = f"{command} @{blob}\n"
+
+                processed_lines.append(line)
+
+        # Combine processed lines back into a single string
+        modelfile_content = '\n'.join(processed_lines)
+
         url = f"{BASE_URL}/api/create"
-        payload = {"name": model_name, "path": model_path}
-        
+        payload = {"name": model_name, "modelfile": modelfile_content}
+
         # Making a POST request with the stream parameter set to True to handle streaming responses
         with requests.post(url, json=payload, stream=True) as response:
             response.raise_for_status()
-
             # Iterating over the response line by line and displaying the status
             for line in response.iter_lines():
                 if line:
-                    # Parsing each line (JSON chunk) and extracting the status
                     chunk = json.loads(line)
-
                     if callback:
                         callback(chunk)
                     else:
                         print(f"Status: {chunk.get('status')}")
-    except requests.exceptions.RequestException as e:
+
+    except Exception as e:
         print(f"An error occurred: {e}")
 
+
 # Pull a model from a the model registry. Cancelled pulls are resumed from where they left off, and multiple
 # calls to will share the same download progress. Use the callback function to override the default handler.
 def pull(model_name, insecure=False, callback=None):
@@ -221,5 +282,3 @@ def heartbeat():
     except requests.exceptions.RequestException as e:
         print(f"An error occurred: {e}")
         return "Ollama is not running"
-
-

api/client_test.go

@@ -0,0 +1,43 @@
+package api
+
+import "testing"
+
+func TestClientFromEnvironment(t *testing.T) {
+	type testCase struct {
+		value  string
+		expect string
+		err    error
+	}
+
+	testCases := map[string]*testCase{
+		"empty":                      {value: "", expect: "http://127.0.0.1:11434"},
+		"only address":               {value: "1.2.3.4", expect: "http://1.2.3.4:11434"},
+		"only port":                  {value: ":1234", expect: "http://:1234"},
+		"address and port":           {value: "1.2.3.4:1234", expect: "http://1.2.3.4:1234"},
+		"scheme http and address":    {value: "http://1.2.3.4", expect: "http://1.2.3.4:80"},
+		"scheme https and address":   {value: "https://1.2.3.4", expect: "https://1.2.3.4:443"},
+		"scheme, address, and port":  {value: "https://1.2.3.4:1234", expect: "https://1.2.3.4:1234"},
+		"hostname":                   {value: "example.com", expect: "http://example.com:11434"},
+		"hostname and port":          {value: "example.com:1234", expect: "http://example.com:1234"},
+		"scheme http and hostname":   {value: "http://example.com", expect: "http://example.com:80"},
+		"scheme https and hostname":  {value: "https://example.com", expect: "https://example.com:443"},
+		"scheme, hostname, and port": {value: "https://example.com:1234", expect: "https://example.com:1234"},
+		"trailing slash":             {value: "example.com/", expect: "http://example.com:11434"},
+		"trailing slash port":        {value: "example.com:1234/", expect: "http://example.com:1234"},
+	}
+
+	for k, v := range testCases {
+		t.Run(k, func(t *testing.T) {
+			t.Setenv("OLLAMA_HOST", v.value)
+
+			client, err := ClientFromEnvironment()
+			if err != v.err {
+				t.Fatalf("expected %s, got %s", v.err, err)
+			}
+
+			if client.base.String() != v.expect {
+				t.Fatalf("expected %s, got %s", v.expect, client.base.String())
+			}
+		})
+	}
+}

api/types.go

@@ -3,10 +3,10 @@ package api
 import (
 	"encoding/json"
 	"fmt"
-	"log"
 	"math"
 	"os"
 	"reflect"
+	"strconv"
 	"strings"
 	"time"
 )
@@ -31,17 +31,100 @@ func (e StatusError) Error() string {
 	}
 }
 
+type ImageData []byte
+
 type GenerateRequest struct {
-	Model    string `json:"model"`
-	Prompt   string `json:"prompt"`
-	System   string `json:"system"`
-	Template string `json:"template"`
-	Context  []int  `json:"context,omitempty"`
-	Stream   *bool  `json:"stream,omitempty"`
+	Model    string      `json:"model"`
+	Prompt   string      `json:"prompt"`
+	System   string      `json:"system"`
+	Template string      `json:"template"`
+	Context  []int       `json:"context,omitempty"`
+	Stream   *bool       `json:"stream,omitempty"`
+	Raw      bool        `json:"raw,omitempty"`
+	Format   string      `json:"format"`
+	Images   []ImageData `json:"images,omitempty"`
 
 	Options map[string]interface{} `json:"options"`
 }
 
+type ChatRequest struct {
+	Model    string    `json:"model"`
+	Messages []Message `json:"messages"`
+	Stream   *bool     `json:"stream,omitempty"`
+	Format   string    `json:"format"`
+
+	Options map[string]interface{} `json:"options"`
+}
+
+type Message struct {
+	Role    string      `json:"role"` // one of ["system", "user", "assistant"]
+	Content string      `json:"content"`
+	Images  []ImageData `json:"images,omitempty"`
+}
+
+type ChatResponse struct {
+	Model     string    `json:"model"`
+	CreatedAt time.Time `json:"created_at"`
+	Message   Message   `json:"message"`
+
+	Done bool `json:"done"`
+
+	Metrics
+}
+
+type Metrics struct {
+	TotalDuration      time.Duration `json:"total_duration,omitempty"`
+	LoadDuration       time.Duration `json:"load_duration,omitempty"`
+	PromptEvalCount    int           `json:"prompt_eval_count,omitempty"`
+	PromptEvalDuration time.Duration `json:"prompt_eval_duration,omitempty"`
+	EvalCount          int           `json:"eval_count,omitempty"`
+	EvalDuration       time.Duration `json:"eval_duration,omitempty"`
+}
+
+// Options specfied in GenerateRequest, if you add a new option here add it to the API docs also
+type Options struct {
+	Runner
+
+	// Predict options used at runtime
+	NumKeep          int      `json:"num_keep,omitempty"`
+	Seed             int      `json:"seed,omitempty"`
+	NumPredict       int      `json:"num_predict,omitempty"`
+	TopK             int      `json:"top_k,omitempty"`
+	TopP             float32  `json:"top_p,omitempty"`
+	TFSZ             float32  `json:"tfs_z,omitempty"`
+	TypicalP         float32  `json:"typical_p,omitempty"`
+	RepeatLastN      int      `json:"repeat_last_n,omitempty"`
+	Temperature      float32  `json:"temperature,omitempty"`
+	RepeatPenalty    float32  `json:"repeat_penalty,omitempty"`
+	PresencePenalty  float32  `json:"presence_penalty,omitempty"`
+	FrequencyPenalty float32  `json:"frequency_penalty,omitempty"`
+	Mirostat         int      `json:"mirostat,omitempty"`
+	MirostatTau      float32  `json:"mirostat_tau,omitempty"`
+	MirostatEta      float32  `json:"mirostat_eta,omitempty"`
+	PenalizeNewline  bool     `json:"penalize_newline,omitempty"`
+	Stop             []string `json:"stop,omitempty"`
+}
+
+// Runner options which must be set when the model is loaded into memory
+type Runner struct {
+	UseNUMA            bool    `json:"numa,omitempty"`
+	NumCtx             int     `json:"num_ctx,omitempty"`
+	NumBatch           int     `json:"num_batch,omitempty"`
+	NumGQA             int     `json:"num_gqa,omitempty"`
+	NumGPU             int     `json:"num_gpu,omitempty"`
+	MainGPU            int     `json:"main_gpu,omitempty"`
+	LowVRAM            bool    `json:"low_vram,omitempty"`
+	F16KV              bool    `json:"f16_kv,omitempty"`
+	LogitsAll          bool    `json:"logits_all,omitempty"`
+	VocabOnly          bool    `json:"vocab_only,omitempty"`
+	UseMMap            bool    `json:"use_mmap,omitempty"`
+	UseMLock           bool    `json:"use_mlock,omitempty"`
+	EmbeddingOnly      bool    `json:"embedding_only,omitempty"`
+	RopeFrequencyBase  float32 `json:"rope_frequency_base,omitempty"`
+	RopeFrequencyScale float32 `json:"rope_frequency_scale,omitempty"`
+	NumThread          int     `json:"num_thread,omitempty"`
+}
+
 type EmbeddingRequest struct {
 	Model  string `json:"model"`
 	Prompt string `json:"prompt"`
@@ -54,9 +137,10 @@ type EmbeddingResponse struct {
 }
 
 type CreateRequest struct {
-	Name   string `json:"name"`
-	Path   string `json:"path"`
-	Stream *bool  `json:"stream,omitempty"`
+	Name      string `json:"name"`
+	Path      string `json:"path"`
+	Modelfile string `json:"modelfile"`
+	Stream    *bool  `json:"stream,omitempty"`
 }
 
 type DeleteRequest struct {
@@ -68,11 +152,12 @@ type ShowRequest struct {
 }
 
 type ShowResponse struct {
-	License    string `json:"license,omitempty"`
-	Modelfile  string `json:"modelfile,omitempty"`
-	Parameters string `json:"parameters,omitempty"`
-	Template   string `json:"template,omitempty"`
-	System     string `json:"system,omitempty"`
+	License    string       `json:"license,omitempty"`
+	Modelfile  string       `json:"modelfile,omitempty"`
+	Parameters string       `json:"parameters,omitempty"`
+	Template   string       `json:"template,omitempty"`
+	System     string       `json:"system,omitempty"`
+	Details    ModelDetails `json:"details,omitempty"`
 }
 
 type CopyRequest struct {
@@ -108,10 +193,11 @@ type ListResponse struct {
 }
 
 type ModelResponse struct {
-	Name       string    `json:"name"`
-	ModifiedAt time.Time `json:"modified_at"`
-	Size       int64     `json:"size"`
-	Digest     string    `json:"digest"`
+	Name       string       `json:"name"`
+	ModifiedAt time.Time    `json:"modified_at"`
+	Size       int64        `json:"size"`
+	Digest     string       `json:"digest"`
+	Details    ModelDetails `json:"details,omitempty"`
 }
 
 type TokenResponse struct {
@@ -126,83 +212,43 @@ type GenerateResponse struct {
 	Done    bool  `json:"done"`
 	Context []int `json:"context,omitempty"`
 
-	TotalDuration      time.Duration `json:"total_duration,omitempty"`
-	LoadDuration       time.Duration `json:"load_duration,omitempty"`
-	PromptEvalCount    int           `json:"prompt_eval_count,omitempty"`
-	PromptEvalDuration time.Duration `json:"prompt_eval_duration,omitempty"`
-	EvalCount          int           `json:"eval_count,omitempty"`
-	EvalDuration       time.Duration `json:"eval_duration,omitempty"`
+	Metrics
 }
 
-func (r *GenerateResponse) Summary() {
-	if r.TotalDuration > 0 {
-		fmt.Fprintf(os.Stderr, "total duration:       %v\n", r.TotalDuration)
-	}
-
-	if r.LoadDuration > 0 {
-		fmt.Fprintf(os.Stderr, "load duration:        %v\n", r.LoadDuration)
-	}
-
-	if r.PromptEvalCount > 0 {
-		fmt.Fprintf(os.Stderr, "prompt eval count:    %d token(s)\n", r.PromptEvalCount)
-	}
-
-	if r.PromptEvalDuration > 0 {
-		fmt.Fprintf(os.Stderr, "prompt eval duration: %s\n", r.PromptEvalDuration)
-		fmt.Fprintf(os.Stderr, "prompt eval rate:     %.2f tokens/s\n", float64(r.PromptEvalCount)/r.PromptEvalDuration.Seconds())
-	}
-
-	if r.EvalCount > 0 {
-		fmt.Fprintf(os.Stderr, "eval count:           %d token(s)\n", r.EvalCount)
-	}
-
-	if r.EvalDuration > 0 {
-		fmt.Fprintf(os.Stderr, "eval duration:        %s\n", r.EvalDuration)
-		fmt.Fprintf(os.Stderr, "eval rate:            %.2f tokens/s\n", float64(r.EvalCount)/r.EvalDuration.Seconds())
-	}
+type ModelDetails struct {
+	Format            string   `json:"format"`
+	Family            string   `json:"family"`
+	Families          []string `json:"families"`
+	ParameterSize     string   `json:"parameter_size"`
+	QuantizationLevel string   `json:"quantization_level"`
 }
 
-type Options struct {
-	Seed int `json:"seed,omitempty"`
+func (m *Metrics) Summary() {
+	if m.TotalDuration > 0 {
+		fmt.Fprintf(os.Stderr, "total duration:       %v\n", m.TotalDuration)
+	}
 
-	// Backend options
-	UseNUMA bool `json:"numa,omitempty"`
+	if m.LoadDuration > 0 {
+		fmt.Fprintf(os.Stderr, "load duration:        %v\n", m.LoadDuration)
+	}
 
-	// Model options
-	NumCtx             int     `json:"num_ctx,omitempty"`
-	NumKeep            int     `json:"num_keep,omitempty"`
-	NumBatch           int     `json:"num_batch,omitempty"`
-	NumGQA             int     `json:"num_gqa,omitempty"`
-	NumGPU             int     `json:"num_gpu,omitempty"`
-	MainGPU            int     `json:"main_gpu,omitempty"`
-	LowVRAM            bool    `json:"low_vram,omitempty"`
-	F16KV              bool    `json:"f16_kv,omitempty"`
-	LogitsAll          bool    `json:"logits_all,omitempty"`
-	VocabOnly          bool    `json:"vocab_only,omitempty"`
-	UseMMap            bool    `json:"use_mmap,omitempty"`
-	UseMLock           bool    `json:"use_mlock,omitempty"`
-	EmbeddingOnly      bool    `json:"embedding_only,omitempty"`
-	RopeFrequencyBase  float32 `json:"rope_frequency_base,omitempty"`
-	RopeFrequencyScale float32 `json:"rope_frequency_scale,omitempty"`
+	if m.PromptEvalCount > 0 {
+		fmt.Fprintf(os.Stderr, "prompt eval count:    %d token(s)\n", m.PromptEvalCount)
+	}
 
-	// Predict options
-	NumPredict       int      `json:"num_predict,omitempty"`
-	TopK             int      `json:"top_k,omitempty"`
-	TopP             float32  `json:"top_p,omitempty"`
-	TFSZ             float32  `json:"tfs_z,omitempty"`
-	TypicalP         float32  `json:"typical_p,omitempty"`
-	RepeatLastN      int      `json:"repeat_last_n,omitempty"`
-	Temperature      float32  `json:"temperature,omitempty"`
-	RepeatPenalty    float32  `json:"repeat_penalty,omitempty"`
-	PresencePenalty  float32  `json:"presence_penalty,omitempty"`
-	FrequencyPenalty float32  `json:"frequency_penalty,omitempty"`
-	Mirostat         int      `json:"mirostat,omitempty"`
-	MirostatTau      float32  `json:"mirostat_tau,omitempty"`
-	MirostatEta      float32  `json:"mirostat_eta,omitempty"`
-	PenalizeNewline  bool     `json:"penalize_newline,omitempty"`
-	Stop             []string `json:"stop,omitempty"`
+	if m.PromptEvalDuration > 0 {
+		fmt.Fprintf(os.Stderr, "prompt eval duration: %s\n", m.PromptEvalDuration)
+		fmt.Fprintf(os.Stderr, "prompt eval rate:     %.2f tokens/s\n", float64(m.PromptEvalCount)/m.PromptEvalDuration.Seconds())
+	}
 
-	NumThread int `json:"num_thread,omitempty"`
+	if m.EvalCount > 0 {
+		fmt.Fprintf(os.Stderr, "eval count:           %d token(s)\n", m.EvalCount)
+	}
+
+	if m.EvalDuration > 0 {
+		fmt.Fprintf(os.Stderr, "eval duration:        %s\n", m.EvalDuration)
+		fmt.Fprintf(os.Stderr, "eval rate:            %.2f tokens/s\n", float64(m.EvalCount)/m.EvalDuration.Seconds())
+	}
 }
 
 var ErrInvalidOpts = fmt.Errorf("invalid options")
@@ -238,44 +284,39 @@ func (opts *Options) FromMap(m map[string]interface{}) error {
 						// when JSON unmarshals numbers, it uses float64, not int
 						field.SetInt(int64(t))
 					default:
-						log.Printf("could not convert model parameter %v of type %T to int, skipped", key, val)
+						return fmt.Errorf("option %q must be of type integer", key)
 					}
 				case reflect.Bool:
 					val, ok := val.(bool)
 					if !ok {
-						log.Printf("could not convert model parameter %v of type %T to bool, skipped", key, val)
-						continue
+						return fmt.Errorf("option %q must be of type boolean", key)
 					}
 					field.SetBool(val)
 				case reflect.Float32:
 					// JSON unmarshals to float64
 					val, ok := val.(float64)
 					if !ok {
-						log.Printf("could not convert model parameter %v of type %T to float32, skipped", key, val)
-						continue
+						return fmt.Errorf("option %q must be of type float32", key)
 					}
 					field.SetFloat(val)
 				case reflect.String:
 					val, ok := val.(string)
 					if !ok {
-						log.Printf("could not convert model parameter %v of type %T to string, skipped", key, val)
-						continue
+						return fmt.Errorf("option %q must be of type string", key)
 					}
 					field.SetString(val)
 				case reflect.Slice:
 					// JSON unmarshals to []interface{}, not []string
 					val, ok := val.([]interface{})
 					if !ok {
-						log.Printf("could not convert model parameter %v of type %T to slice, skipped", key, val)
-						continue
+						return fmt.Errorf("option %q must be of type array", key)
 					}
 					// convert []interface{} to []string
 					slice := make([]string, len(val))
 					for i, item := range val {
 						str, ok := item.(string)
 						if !ok {
-							log.Printf("could not convert model parameter %v of type %T to slice of strings, skipped", key, item)
-							continue
+							return fmt.Errorf("option %q must be of an array of strings", key)
 						}
 						slice[i] = str
 					}
@@ -299,7 +340,7 @@ func DefaultOptions() Options {
 	return Options{
 		// options set on request to runner
 		NumPredict:       -1,
-		NumKeep:          -1,
+		NumKeep:          0,
 		Temperature:      0.8,
 		TopK:             40,
 		TopP:             0.9,
@@ -315,20 +356,22 @@ func DefaultOptions() Options {
 		PenalizeNewline:  true,
 		Seed:             -1,
 
-		// options set when the model is loaded
-		NumCtx:             2048,
-		RopeFrequencyBase:  10000.0,
-		RopeFrequencyScale: 1.0,
-		NumBatch:           512,
-		NumGPU:             -1, // -1 here indicates that NumGPU should be set dynamically
-		NumGQA:             1,
-		NumThread:          0, // let the runtime decide
-		LowVRAM:            false,
-		F16KV:              true,
-		UseMLock:           false,
-		UseMMap:            true,
-		UseNUMA:            false,
-		EmbeddingOnly:      true,
+		Runner: Runner{
+			// options set when the model is loaded
+			NumCtx:             2048,
+			RopeFrequencyBase:  10000.0,
+			RopeFrequencyScale: 1.0,
+			NumBatch:           512,
+			NumGPU:             -1, // -1 here indicates that NumGPU should be set dynamically
+			NumGQA:             1,
+			NumThread:          0, // let the runtime decide
+			LowVRAM:            false,
+			F16KV:              true,
+			UseMLock:           false,
+			UseMMap:            true,
+			UseNUMA:            false,
+			EmbeddingOnly:      true,
+		},
 	}
 }
 
@@ -360,3 +403,63 @@ func (d *Duration) UnmarshalJSON(b []byte) (err error) {
 
 	return nil
 }
+
+// FormatParams converts specified parameter options to their correct types
+func FormatParams(params map[string][]string) (map[string]interface{}, error) {
+	opts := Options{}
+	valueOpts := reflect.ValueOf(&opts).Elem() // names of the fields in the options struct
+	typeOpts := reflect.TypeOf(opts)           // types of the fields in the options struct
+
+	// build map of json struct tags to their types
+	jsonOpts := make(map[string]reflect.StructField)
+	for _, field := range reflect.VisibleFields(typeOpts) {
+		jsonTag := strings.Split(field.Tag.Get("json"), ",")[0]
+		if jsonTag != "" {
+			jsonOpts[jsonTag] = field
+		}
+	}
+
+	out := make(map[string]interface{})
+	// iterate params and set values based on json struct tags
+	for key, vals := range params {
+		if opt, ok := jsonOpts[key]; !ok {
+			return nil, fmt.Errorf("unknown parameter '%s'", key)
+		} else {
+			field := valueOpts.FieldByName(opt.Name)
+			if field.IsValid() && field.CanSet() {
+				switch field.Kind() {
+				case reflect.Float32:
+					floatVal, err := strconv.ParseFloat(vals[0], 32)
+					if err != nil {
+						return nil, fmt.Errorf("invalid float value %s", vals)
+					}
+
+					out[key] = float32(floatVal)
+				case reflect.Int:
+					intVal, err := strconv.ParseInt(vals[0], 10, 64)
+					if err != nil {
+						return nil, fmt.Errorf("invalid int value %s", vals)
+					}
+
+					out[key] = intVal
+				case reflect.Bool:
+					boolVal, err := strconv.ParseBool(vals[0])
+					if err != nil {
+						return nil, fmt.Errorf("invalid bool value %s", vals)
+					}
+
+					out[key] = boolVal
+				case reflect.String:
+					out[key] = vals[0]
+				case reflect.Slice:
+					// TODO: only string slices are supported right now
+					out[key] = vals
+				default:
+					return nil, fmt.Errorf("unknown type %s for %s", field.Kind(), key)
+				}
+			}
+		}
+	}
+
+	return out, nil
+}

app/forge.config.ts

@@ -47,16 +47,6 @@ const config: ForgeConfig = {
   },
   rebuildConfig: {},
   makers: [new MakerSquirrel({}), new MakerZIP({}, ['darwin'])],
-  publishers: [
-    new PublisherGithub({
-      repository: {
-        name: 'ollama',
-        owner: 'jmorganca',
-      },
-      draft: false,
-      prerelease: true,
-    }),
-  ],
   hooks: {
     readPackageJson: async (_, packageJson) => {
       return { ...packageJson, version: process.env.VERSION || packageJson.version }

app/package.json

@@ -46,7 +46,7 @@
     "chmodr": "^1.2.0",
     "copy-webpack-plugin": "^11.0.0",
     "css-loader": "^6.8.1",
-    "electron": "25.2.0",
+    "electron": "25.9.2",
     "eslint": "^8.43.0",
     "eslint-plugin-import": "^2.27.5",
     "fork-ts-checker-webpack-plugin": "^7.3.0",

app/src/index.ts

@@ -162,13 +162,56 @@ app.on('before-quit', () => {
   }
 })
 
+const updateURL = `https://ollama.ai/api/update?os=${process.platform}&arch=${
+  process.arch
+}&version=${app.getVersion()}&id=${id()}`
+
+let latest = ''
+async function isNewReleaseAvailable() {
+  try {
+    const response = await fetch(updateURL)
+
+    if (!response.ok) {
+      return false
+    }
+
+    if (response.status === 204) {
+      return false
+    }
+
+    const data = await response.json()
+
+    const url = data?.url
+    if (!url) {
+      return false
+    }
+
+    if (latest === url) {
+      return false
+    }
+
+    latest = url
+
+    return true
+  } catch (error) {
+    logger.error(`update check failed - ${error}`)
+    return false
+  }
+}
+
+async function checkUpdate() {
+  const available = await isNewReleaseAvailable()
+  if (available) {
+    logger.info('checking for update')
+    autoUpdater.checkForUpdates()
+  }
+}
+
 function init() {
   if (app.isPackaged) {
-    autoUpdater.checkForUpdates()
+    checkUpdate()
     setInterval(() => {
-      if (!updateAvailable) {
-        autoUpdater.checkForUpdates()
-      }
+      checkUpdate()
     }, 60 * 60 * 1000)
   }
 
@@ -246,11 +289,7 @@ function id(): string {
   return uuid
 }
 
-autoUpdater.setFeedURL({
-  url: `https://ollama.ai/api/update?os=${process.platform}&arch=${
-    process.arch
-  }&version=${app.getVersion()}&id=${id()}`,
-})
+autoUpdater.setFeedURL({ url: updateURL })
 
 autoUpdater.on('error', e => {
   logger.error(`update check failed - ${e.message}`)

cmd/spinner.go

@@ -1,44 +0,0 @@
-package cmd
-
-import (
-	"fmt"
-	"os"
-	"time"
-
-	"github.com/jmorganca/ollama/progressbar"
-)
-
-type Spinner struct {
-	description string
-	*progressbar.ProgressBar
-}
-
-func NewSpinner(description string) *Spinner {
-	return &Spinner{
-		description: description,
-		ProgressBar: progressbar.NewOptions(-1,
-			progressbar.OptionSetWriter(os.Stderr),
-			progressbar.OptionThrottle(60*time.Millisecond),
-			progressbar.OptionSpinnerType(14),
-			progressbar.OptionSetRenderBlankState(true),
-			progressbar.OptionSetElapsedTime(false),
-			progressbar.OptionClearOnFinish(),
-			progressbar.OptionSetDescription(description),
-		),
-	}
-}
-
-func (s *Spinner) Spin(tick time.Duration) {
-	for range time.Tick(tick) {
-		if s.IsFinished() {
-			break
-		}
-
-		s.Add(1)
-	}
-}
-
-func (s *Spinner) Stop() {
-	s.Finish()
-	fmt.Println(s.description)
-}

docs/README.md

@@ -1,6 +1,25 @@
 # Documentation
 
-- [Modelfile](./modelfile.md)
-- [How to develop Ollama](./development.md)
-- [API](./api.md)
-- [Tutorials](./tutorials.md)
+To get started, see the project's **[quickstart](../README.md#quickstart)**.
+
+Ollama is a tool for running AI models on your hardware. Many users will choose to use the Command Line Interface (CLI) to work with Ollama. Learn more about all the commands in the CLI in the **[Main Readme](../README.md)**.
+
+Use the RESTful API using any language, including Python, JavaScript, Typescript, Go, Rust, and many more. Learn more about using the API in the **[API Documentation](./api.md)**.
+
+Create new models or modify models already in the library using the Modelfile. Learn more about the Modelfile syntax in the **[Modelfile Documentation](./modelfile.md)**.
+
+Import models using source model weights found on Hugging Face and similar sites by referring to the **[Import Documentation](./import.md)**.
+
+Installing on Linux in most cases is easy using the script on Ollama.ai. To get more detail about the install, including CUDA drivers, see the **[Linux Documentation](./linux.md)**.
+
+Many of our users like the flexibility of using our official Docker Image. Learn more about using Docker with Ollama using the **[Docker Documentation](./docker.md)**.
+
+It is easy to install on Linux and Mac, but many users will choose to build Ollama on their own. To do this, refer to the **[Development Documentation](./development.md)**.
+
+If encountering a problem with Ollama, the best place to start is the logs. Find more information about them here in the **[Troubleshooting Guide](./troubleshooting.md)**.
+
+Finally for all the questions that don't fit anywhere else, there is the **[FAQ](./faq.md)**
+
+[Tutorials](./tutorials.md) apply the documentation to tasks.
+
+For working code examples of using Ollama, see [Examples](../examples).

docs/api.md

@@ -3,6 +3,7 @@
 ## Endpoints
 
 - [Generate a completion](#generate-a-completion)
+- [Generate a chat completion](#generate-a-chat-completion)
 - [Create a Model](#create-a-model)
 - [List Local Models](#list-local-models)
 - [Show Model Information](#show-model-information)
@@ -16,7 +17,7 @@
 
 ### Model names
 
-Model names follow a `model:tag` format. Some examples are `orca-mini:3b-q4_1` and `llama2:70b`. The tag is optional and, if not provided, will default to `latest`. The tag is used to identify a specific version.
+Model names follow a `model:tag` format, where `model` can have an optional namespace such as `example/model`. Some examples are `orca-mini:3b-q4_1` and `llama2:70b`. The tag is optional and, if not provided, will default to `latest`. The tag is used to identify a specific version.
 
 ### Durations
 
@@ -24,7 +25,7 @@ All durations are returned in nanoseconds.
 
 ### Streaming responses
 
-Certain endpoints stream responses as JSON objects delineated with the newline (`\n`) character.
+Certain endpoints stream responses as JSON objects and can optional return non-streamed responses.
 
 ## Generate a completion
 
@@ -32,37 +33,50 @@ Certain endpoints stream responses as JSON objects delineated with the newline (
 POST /api/generate
 ```
 
-Generate a response for a given prompt with a provided model. This is a streaming endpoint, so will be a series of responses. The final response object will include statistics and additional data from the request.
+Generate a response for a given prompt with a provided model. This is a streaming endpoint, so there will be a series of responses. The final response object will include statistics and additional data from the request.
 
 ### Parameters
 
 - `model`: (required) the [model name](#model-names)
 - `prompt`: the prompt to generate a response for
+- `images`: (optional) a list of base64-encoded images (for multimodal models such as `llava`)
 
 Advanced parameters (optional):
 
+- `format`: the format to return a response in. Currently the only accepted value is `json`
 - `options`: additional model parameters listed in the documentation for the [Modelfile](./modelfile.md#valid-parameters-and-values) such as `temperature`
-- `system`: system prompt to (overrides what is defined in the `Modelfile`)
-- `template`: the full prompt or prompt template (overrides what is defined in the `Modelfile`)
+- `system`: system message to (overrides what is defined in the `Modelfile`)
+- `template`: the prompt template to use (overrides what is defined in the `Modelfile`)
 - `context`: the context parameter returned from a previous request to `/generate`, this can be used to keep a short conversational memory
-- `stream`: if `false` the response will be be returned as a single response object, rather than a stream of objects
+- `stream`: if `false` the response will be returned as a single response object, rather than a stream of objects
+- `raw`: if `true` no formatting will be applied to the prompt. You may choose to use the `raw` parameter if you are specifying a full templated prompt in your request to the API.
 
-### Request
+#### JSON mode
+
+Enable JSON mode by setting the `format` parameter to `json`. This will structure the response as a valid JSON object. See the JSON mode [example](#generate-request-json-mode) below.
+
+> Note: it's important to instruct the model to use JSON in the `prompt`. Otherwise, the model may generate large amounts whitespace.
+
+### Examples
+
+#### Generate request (Streaming)
+
+##### Request
 
 ```shell
-curl -X POST http://localhost:11434/api/generate -d '{
-  "model": "llama2:7b",
+curl http://localhost:11434/api/generate -d '{
+  "model": "llama2",
   "prompt": "Why is the sky blue?"
 }'
 ```
 
-### Response
+##### Response
 
-A stream of JSON objects:
+A stream of JSON objects is returned:
 
 ```json
 {
-  "model": "llama2:7b",
+  "model": "llama2",
   "created_at": "2023-08-04T08:52:19.385406455-07:00",
   "response": "The",
   "done": false
@@ -73,8 +87,6 @@ The final response in the stream also includes additional data about the generat
 
 - `total_duration`: time spent generating the response
 - `load_duration`: time spent in nanoseconds loading the model
-- `sample_count`: number of samples generated
-- `sample_duration`: time spent generating samples
 - `prompt_eval_count`: number of tokens in the prompt
 - `prompt_eval_duration`: time spent in nanoseconds evaluating the prompt
 - `eval_count`: number of tokens the response
@@ -86,19 +98,473 @@ To calculate how fast the response is generated in tokens per second (token/s),
 
 ```json
 {
-  "model": "llama2:7b",
+  "model": "llama2",
   "created_at": "2023-08-04T19:22:45.499127Z",
   "response": "",
-  "context": [1, 2, 3],
   "done": true,
-  "total_duration": 5589157167,
-  "load_duration": 3013701500,
-  "sample_count": 114,
-  "sample_duration": 81442000,
-  "prompt_eval_count": 46,
-  "prompt_eval_duration": 1160282000,
-  "eval_count": 113,
-  "eval_duration": 1325948000
+  "context": [1, 2, 3],
+  "total_duration": 10706818083,
+  "load_duration": 6338219291,
+  "prompt_eval_count": 26,
+  "prompt_eval_duration": 130079000,
+  "eval_count": 259,
+  "eval_duration": 4232710000
+}
+```
+
+#### Request (No streaming)
+
+##### Request
+
+A response can be received in one reply when streaming is off.
+
+```shell
+curl http://localhost:11434/api/generate -d '{
+  "model": "llama2",
+  "prompt": "Why is the sky blue?",
+  "stream": false
+}'
+```
+
+##### Response
+
+If `stream` is set to `false`, the response will be a single JSON object:
+
+```json
+{
+  "model": "llama2",
+  "created_at": "2023-08-04T19:22:45.499127Z",
+  "response": "The sky is blue because it is the color of the sky.",
+  "done": true,
+  "context": [1, 2, 3],
+  "total_duration": 5043500667,
+  "load_duration": 5025959,
+  "prompt_eval_count": 26,
+  "prompt_eval_duration": 325953000,
+  "eval_count": 290,
+  "eval_duration": 4709213000
+}
+```
+
+#### Request (JSON mode)
+
+> When `format` is set to `json`, the output will always be a well-formed JSON object. It's important to also instruct the model to respond in JSON.
+
+##### Request
+
+```shell
+curl http://localhost:11434/api/generate -d '{
+  "model": "llama2",
+  "prompt": "What color is the sky at different times of the day? Respond using JSON",
+  "format": "json",
+  "stream": false
+}'
+```
+
+##### Response
+
+```json
+{
+  "model": "llama2",
+  "created_at": "2023-11-09T21:07:55.186497Z",
+  "response": "{\n\"morning\": {\n\"color\": \"blue\"\n},\n\"noon\": {\n\"color\": \"blue-gray\"\n},\n\"afternoon\": {\n\"color\": \"warm gray\"\n},\n\"evening\": {\n\"color\": \"orange\"\n}\n}\n",
+  "done": true,
+  "context": [1, 2, 3],
+  "total_duration": 4648158584,
+  "load_duration": 4071084,
+  "prompt_eval_count": 36,
+  "prompt_eval_duration": 439038000,
+  "eval_count": 180,
+  "eval_duration": 4196918000
+}
+```
+
+The value of `response` will be a string containing JSON similar to:
+
+```json
+{
+  "morning": {
+    "color": "blue"
+  },
+  "noon": {
+    "color": "blue-gray"
+  },
+  "afternoon": {
+    "color": "warm gray"
+  },
+  "evening": {
+    "color": "orange"
+  }
+}
+```
+
+#### Request (with images)
+
+To submit images to multimodal models such as `llava` or `bakllava`, provide a list of base64-encoded `images`:
+
+#### Request
+
+```shell
+curl http://localhost:11434/api/generate -d '{
+  "model": "llava",
+  "prompt":"What is in this picture?",
+  "stream": false,
+  "images": ["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"]
+}'
+```
+
+#### Response
+
+```
+{
+  "model": "llava",
+  "created_at": "2023-11-03T15:36:02.583064Z",
+  "response": "A happy cartoon character, which is cute and cheerful.",
+  "done": true,
+  "context": [1, 2, 3],
+  "total_duration": 2938432250,
+  "load_duration": 2559292,
+  "prompt_eval_count": 1,
+  "prompt_eval_duration": 2195557000,
+  "eval_count": 44,
+  "eval_duration": 736432000
+}
+```
+
+#### Request (Raw Mode)
+
+In some cases, you may wish to bypass the templating system and provide a full prompt. In this case, you can use the `raw` parameter to disable templating. Also note that raw mode will not return a context.
+
+##### Request
+
+```shell
+curl http://localhost:11434/api/generate -d '{
+  "model": "mistral",
+  "prompt": "[INST] why is the sky blue? [/INST]",
+  "raw": true,
+  "stream": false
+}'
+```
+
+##### Response
+
+```json
+{
+  "model": "mistral",
+  "created_at": "2023-11-03T15:36:02.583064Z",
+  "response": " The sky appears blue because of a phenomenon called Rayleigh scattering.",
+  "done": true,
+  "total_duration": 8493852375,
+  "load_duration": 6589624375,
+  "prompt_eval_count": 14,
+  "prompt_eval_duration": 119039000,
+  "eval_count": 110,
+  "eval_duration": 1779061000
+}
+```
+
+#### Generate request (With options)
+
+If you want to set custom options for the model at runtime rather than in the Modelfile, you can do so with the `options` parameter. This example sets every available option, but you can set any of them individually and omit the ones you do not want to override.
+
+##### Request
+
+```shell
+curl http://localhost:11434/api/generate -d '{
+  "model": "llama2",
+  "prompt": "Why is the sky blue?",
+  "stream": false,
+  "options": {
+    "num_keep": 5,
+    "seed": 42,
+    "num_predict": 100,
+    "top_k": 20,
+    "top_p": 0.9,
+    "tfs_z": 0.5,
+    "typical_p": 0.7,
+    "repeat_last_n": 33,
+    "temperature": 0.8,
+    "repeat_penalty": 1.2,
+    "presence_penalty": 1.5,
+    "frequency_penalty": 1.0,
+    "mirostat": 1,
+    "mirostat_tau": 0.8,
+    "mirostat_eta": 0.6,
+    "penalize_newline": true,
+    "stop": ["\n", "user:"],
+    "numa": false,
+    "num_ctx": 1024,
+    "num_batch": 2,
+    "num_gqa": 1,
+    "num_gpu": 1,
+    "main_gpu": 0,
+    "low_vram": false,
+    "f16_kv": true,
+    "vocab_only": false,
+    "use_mmap": true,
+    "use_mlock": false,
+    "embedding_only": false,
+    "rope_frequency_base": 1.1,
+    "rope_frequency_scale": 0.8,
+    "num_thread": 8
+  }
+}'
+```
+
+##### Response
+
+```json
+{
+  "model": "llama2",
+  "created_at": "2023-08-04T19:22:45.499127Z",
+  "response": "The sky is blue because it is the color of the sky.",
+  "done": true,
+  "context": [1, 2, 3],
+  "total_duration": 4935886791,
+  "load_duration": 534986708,
+  "prompt_eval_count": 26,
+  "prompt_eval_duration": 107345000,
+  "eval_count": 237,
+  "eval_duration": 4289432000
+}
+```
+
+#### Load a model
+
+If an empty prompt is provided, the model will be loaded into memory.
+
+##### Request
+
+```shell
+curl http://localhost:11434/api/generate -d '{
+  "model": "llama2"
+}'
+```
+
+##### Response
+
+A single JSON object is returned:
+
+```json
+{
+  "model": "llama2",
+  "created_at": "2023-12-18T19:52:07.071755Z",
+  "response": "",
+  "done": true
+}
+```
+
+## Generate a chat completion
+
+```shell
+POST /api/chat
+```
+
+Generate the next message in a chat with a provided model. This is a streaming endpoint, so there will be a series of responses. Streaming can be disabled using `"stream": false`. The final response object will include statistics and additional data from the request.
+
+### Parameters
+
+- `model`: (required) the [model name](#model-names)
+- `messages`: the messages of the chat, this can be used to keep a chat memory
+
+The `message` object has the following fields:
+
+- `role`: the role of the message, either `system`, `user` or `assistant`
+- `content`: the content of the message
+- `images` (optional): a list of images to include in the message (for multimodal models such as `llava`)
+
+Advanced parameters (optional):
+
+- `format`: the format to return a response in. Currently the only accepted value is `json`
+- `options`: additional model parameters listed in the documentation for the [Modelfile](./modelfile.md#valid-parameters-and-values) such as `temperature`
+- `template`: the prompt template to use (overrides what is defined in the `Modelfile`)
+- `stream`: if `false` the response will be returned as a single response object, rather than a stream of objects
+
+### Examples
+
+#### Chat Request (Streaming)
+
+##### Request
+
+Send a chat message with a streaming response.
+
+```shell
+curl http://localhost:11434/api/chat -d '{
+  "model": "llama2",
+  "messages": [
+    {
+      "role": "user",
+      "content": "why is the sky blue?"
+    }
+  ]
+}'
+```
+
+##### Response
+
+A stream of JSON objects is returned:
+
+```json
+{
+  "model": "llama2",
+  "created_at": "2023-08-04T08:52:19.385406455-07:00",
+  "message": {
+    "role": "assisant",
+    "content": "The",
+    "images": null
+  },
+  "done": false
+}
+```
+
+Final response:
+
+```json
+{
+  "model": "llama2",
+  "created_at": "2023-08-04T19:22:45.499127Z",
+  "done": true,
+  "total_duration": 4883583458,
+  "load_duration": 1334875,
+  "prompt_eval_count": 26,
+  "prompt_eval_duration": 342546000,
+  "eval_count": 282,
+  "eval_duration": 4535599000
+}
+```
+
+#### Chat request (No streaming)
+
+##### Request
+
+```shell
+curl http://localhost:11434/api/chat -d '{
+  "model": "llama2",
+  "messages": [
+    {
+      "role": "user",
+      "content": "why is the sky blue?"
+    }
+  ],
+  "stream": false
+}'
+```
+
+##### Response
+
+```json
+{
+  "model": "registry.ollama.ai/library/llama2:latest",
+  "created_at": "2023-12-12T14:13:43.416799Z",
+  "message": {
+    "role": "assistant",
+    "content": "Hello! How are you today?"
+  },
+  "done": true,
+  "total_duration": 5191566416,
+  "load_duration": 2154458,
+  "prompt_eval_count": 26,
+  "prompt_eval_duration": 383809000,
+  "eval_count": 298,
+  "eval_duration": 4799921000
+}
+```
+
+#### Chat request (With History)
+
+Send a chat message with a conversation history. You can use this same approach to start the conversation using multi-shot or chain-of-thought prompting.
+
+##### Request
+
+```shell
+curl http://localhost:11434/api/chat -d '{
+  "model": "llama2",
+  "messages": [
+    {
+      "role": "user",
+      "content": "why is the sky blue?"
+    },
+    {
+      "role": "assistant",
+      "content": "due to rayleigh scattering."
+    },
+    {
+      "role": "user",
+      "content": "how is that different than mie scattering?"
+    }
+  ]
+}'
+```
+
+##### Response
+
+A stream of JSON objects is returned:
+
+```json
+{
+  "model": "llama2",
+  "created_at": "2023-08-04T08:52:19.385406455-07:00",
+  "message": {
+    "role": "assisant",
+    "content": "The"
+  },
+  "done": false
+}
+```
+
+Final response:
+
+```json
+{
+  "model": "llama2",
+  "created_at": "2023-08-04T19:22:45.499127Z",
+  "done": true,
+  "total_duration": 8113331500,
+  "load_duration": 6396458,
+  "prompt_eval_count": 61,
+  "prompt_eval_duration": 398801000,
+  "eval_count": 468,
+  "eval_duration": 7701267000
+}
+```
+
+#### Chat request (with images)
+
+##### Request
+
+Send a chat message with a conversation history.
+
+```shell
+curl http://localhost:11434/api/chat -d '{
+  "model": "llava",
+  "messages": [
+    {
+      "role": "user",
+      "content": "what is in this image?",
+      "images": ["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"]
+    },
+  ]
+}'
+```
+
+##### Response
+
+```json
+{
+  "model": "llava",
+  "created_at": "2023-12-13T22:42:50.203334Z",
+  "message": {
+    "role": "assistant",
+    "content": " The image features a cute, little pig with an angry facial expression. It's wearing a heart on its shirt and is waving in the air. This scene appears to be part of a drawing or sketching project.",
+    "images": null
+  },
+  "done": true,
+  "total_duration": 1668506709,
+  "load_duration": 1986209,
+  "prompt_eval_count": 26,
+  "prompt_eval_duration": 359682000,
+  "eval_count": 83,
+  "eval_duration": 1303285000
 }
 ```
 
@@ -108,33 +574,96 @@ To calculate how fast the response is generated in tokens per second (token/s),
 POST /api/create
 ```
 
-Create a model from a [`Modelfile`](./modelfile.md)
+Create a model from a [`Modelfile`](./modelfile.md). It is recommended to set `modelfile` to the content of the Modelfile rather than just set `path`. This is a requirement for remote create. Remote model creation must also create any file blobs, fields such as `FROM` and `ADAPTER`, explicitly with the server using [Create a Blob](#create-a-blob) and the value to the path indicated in the response.
 
 ### Parameters
 
 - `name`: name of the model to create
-- `path`: path to the Modelfile
-- `stream`: (optional) if `false` the response will be be returned as a single response object, rather than a stream of objects
+- `modelfile` (optional): contents of the Modelfile
+- `stream`: (optional) if `false` the response will be returned as a single response object, rather than a stream of objects
+- `path` (optional): path to the Modelfile
 
-### Request
+### Examples
+
+#### Create a new model
+
+Create a new model from a `Modelfile`.
+
+##### Request
 
 ```shell
-curl -X POST http://localhost:11434/api/create -d '{
+curl http://localhost:11434/api/create -d '{
   "name": "mario",
-  "path": "~/Modelfile"
+  "modelfile": "FROM llama2\nSYSTEM You are mario from Super Mario Bros."
 }'
 ```
 
-### Response
+##### Response
 
-A stream of JSON objects. When finished, `status` is `success`.
+A stream of JSON objects. Notice that the final JSON object shows a `"status": "success"`.
 
 ```json
-{
-  "status": "parsing modelfile"
-}
+{"status":"reading model metadata"}
+{"status":"creating system layer"}
+{"status":"using already created layer sha256:22f7f8ef5f4c791c1b03d7eb414399294764d7cc82c7e94aa81a1feb80a983a2"}
+{"status":"using already created layer sha256:8c17c2ebb0ea011be9981cc3922db8ca8fa61e828c5d3f44cb6ae342bf80460b"}
+{"status":"using already created layer sha256:7c23fb36d80141c4ab8cdbb61ee4790102ebd2bf7aeff414453177d4f2110e5d"}
+{"status":"using already created layer sha256:2e0493f67d0c8c9c68a8aeacdf6a38a2151cb3c4c1d42accf296e19810527988"}
+{"status":"using already created layer sha256:2759286baa875dc22de5394b4a925701b1896a7e3f8e53275c36f75a877a82c9"}
+{"status":"writing layer sha256:df30045fe90f0d750db82a058109cecd6d4de9c90a3d75b19c09e5f64580bb42"}
+{"status":"writing layer sha256:f18a68eb09bf925bb1b669490407c1b1251c5db98dc4d3d81f3088498ea55690"}
+{"status":"writing manifest"}
+{"status":"success"}
 ```
 
+### Check if a Blob Exists
+
+```shell
+HEAD /api/blobs/:digest
+```
+
+Ensures that the file blob used for a FROM or ADAPTER field exists on the server. This is checking your Ollama server and not Ollama.ai.
+
+#### Query Parameters
+
+- `digest`: the SHA256 digest of the blob
+
+#### Examples
+
+##### Request
+
+```shell
+curl -I http://localhost:11434/api/blobs/sha256:29fdb92e57cf0827ded04ae6461b5931d01fa595843f55d36f5b275a52087dd2
+```
+
+##### Response
+
+Return 200 OK if the blob exists, 404 Not Found if it does not.
+
+### Create a Blob
+
+```shell
+POST /api/blobs/:digest
+```
+
+Create a blob from a file on the server. Returns the server file path.
+
+#### Query Parameters
+
+- `digest`: the expected SHA256 digest of the file
+
+#### Examples
+
+##### Request
+
+```shell
+curl -T model.bin -X POST http://localhost:11434/api/blobs/sha256:29fdb92e57cf0827ded04ae6461b5931d01fa595843f55d36f5b275a52087dd2
+```
+
+##### Response
+
+Return 201 Created if the blob was successfully created, 400 Bad Request if the digest used is not expected.
+
 ## List Local Models
 
 ```shell
@@ -143,26 +672,46 @@ GET /api/tags
 
 List models that are available locally.
 
-### Request
+### Examples
+
+#### Request
 
 ```shell
 curl http://localhost:11434/api/tags
 ```
 
-### Response
+#### Response
+
+A single JSON object will be returned.
 
 ```json
 {
   "models": [
     {
-      "name": "llama2:7b",
-      "modified_at": "2023-08-02T17:02:23.713454393-07:00",
-      "size": 3791730596
+      "name": "codellama:13b",
+      "modified_at": "2023-11-04T14:56:49.277302595-07:00",
+      "size": 7365960935,
+      "digest": "9f438cb9cd581fc025612d27f7c1a6669ff83a8bb0ed86c94fcf4c5440555697",
+      "details": {
+        "format": "gguf",
+        "family": "llama",
+        "families": null,
+        "parameter_size": "13B",
+        "quantization_level": "Q4_0"
+      }
     },
     {
-      "name": "llama2:13b",
-      "modified_at": "2023-08-08T12:08:38.093596297-07:00",
-      "size": 7323310500
+      "name": "llama2:latest",
+      "modified_at": "2023-12-07T09:32:18.757212583-08:00",
+      "size": 3825819519,
+      "digest": "fe938a131f40e6f6d40083c9f0f430a515233eb2edaa6d72eb85c50d64f2300e",
+      "details": {
+        "format": "gguf",
+        "family": "llama",
+        "families": null,
+        "parameter_size": "7B",
+        "quantization_level": "Q4_0"
+      }
     }
   ]
 }
@@ -174,28 +723,36 @@ curl http://localhost:11434/api/tags
 POST /api/show
 ```
 
-Show details about a model including modelfile, template, parameters, license, and system prompt.
+Show information about a model including details, modelfile, template, parameters, license, and system prompt.
 
 ### Parameters
 
 - `name`: name of the model to show
 
-### Request
+### Examples
+
+#### Request
 
 ```shell
 curl http://localhost:11434/api/show -d '{
-  "name": "llama2:7b"
+  "name": "llama2"
 }'
 ```
 
-### Response
+#### Response
 
 ```json
 {
-  "license": "<contents of license block>",
-  "modelfile": "# Modelfile generated by \"ollama show\"\n# To build a new Modelfile based on this one, replace the FROM line with:\n# FROM llama2:latest\n\nFROM /Users/username/.ollama/models/blobs/sha256:8daa9615cce30c259a9555b1cc250d461d1bc69980a274b44d7eda0be78076d8\nTEMPLATE \"\"\"[INST] {{ if and .First .System }}<<SYS>>{{ .System }}<</SYS>>\n\n{{ end }}{{ .Prompt }} [/INST] \"\"\"\nSYSTEM \"\"\"\"\"\"\nPARAMETER stop [INST]\nPARAMETER stop [/INST]\nPARAMETER stop <<SYS>>\nPARAMETER stop <</SYS>>\n",
-  "parameters": "stop                           [INST]\nstop                           [/INST]\nstop                           <<SYS>>\nstop                           <</SYS>>",
-  "template": "[INST] {{ if and .First .System }}<<SYS>>{{ .System }}<</SYS>>\n\n{{ end }}{{ .Prompt }} [/INST] "
+  "modelfile": "# Modelfile generated by \"ollama show\"\n# To build a new Modelfile based on this one, replace the FROM line with:\n# FROM llava:latest\n\nFROM /Users/matt/.ollama/models/blobs/sha256:200765e1283640ffbd013184bf496e261032fa75b99498a9613be4e94d63ad52\nTEMPLATE \"\"\"{{ .System }}\nUSER: {{ .Prompt }}\nASSSISTANT: \"\"\"\nPARAMETER num_ctx 4096\nPARAMETER stop \"\u003c/s\u003e\"\nPARAMETER stop \"USER:\"\nPARAMETER stop \"ASSSISTANT:\"",
+  "parameters": "num_ctx                        4096\nstop                           \u003c/s\u003e\nstop                           USER:\nstop                           ASSSISTANT:",
+  "template": "{{ .System }}\nUSER: {{ .Prompt }}\nASSSISTANT: ",
+  "details": {
+    "format": "gguf",
+    "family": "llama",
+    "families": ["llama", "clip"],
+    "parameter_size": "7B",
+    "quantization_level": "Q4_0"
+  }
 }
 ```
 
@@ -207,15 +764,21 @@ POST /api/copy
 
 Copy a model. Creates a model with another name from an existing model.
 
-### Request
+### Examples
+
+#### Request
 
 ```shell
 curl http://localhost:11434/api/copy -d '{
-  "source": "llama2:7b",
+  "source": "llama2",
   "destination": "llama2-backup"
 }'
 ```
 
+#### Response
+
+Returns a 200 OK if successful, or a 404 Not Found if the source model doesn't exist.
+
 ## Delete a Model
 
 ```shell
@@ -226,9 +789,11 @@ Delete a model and its data.
 
 ### Parameters
 
-- `model`: model name to delete
+- `name`: model name to delete
 
-### Request
+### Examples
+
+#### Request
 
 ```shell
 curl -X DELETE http://localhost:11434/api/delete -d '{
@@ -236,6 +801,10 @@ curl -X DELETE http://localhost:11434/api/delete -d '{
 }'
 ```
 
+#### Response
+
+Returns a 200 OK if successful, 404 Not Found if the model to be deleted doesn't exist.
+
 ## Pull a Model
 
 ```shell
@@ -248,23 +817,63 @@ Download a model from the ollama library. Cancelled pulls are resumed from where
 
 - `name`: name of the model to pull
 - `insecure`: (optional) allow insecure connections to the library. Only use this if you are pulling from your own library during development.
-- `stream`: (optional) if `false` the response will be be returned as a single response object, rather than a stream of objects
+- `stream`: (optional) if `false` the response will be returned as a single response object, rather than a stream of objects
 
-### Request
+### Examples
+
+#### Request
 
 ```shell
-curl -X POST http://localhost:11434/api/pull -d '{
-  "name": "llama2:7b"
+curl http://localhost:11434/api/pull -d '{
+  "name": "llama2"
 }'
 ```
 
-### Response
+#### Response
+
+If `stream` is not specified, or set to `true`, a stream of JSON objects is returned:
+
+The first object is the manifest:
+
+```json
+{
+  "status": "pulling manifest"
+}
+```
+
+Then there is a series of downloading responses. Until any of the download is completed, the `completed` key may not be included. The number of files to be downloaded depends on the number of layers specified in the manifest.
 
 ```json
 {
   "status": "downloading digestname",
   "digest": "digestname",
-  "total": 2142590208
+  "total": 2142590208,
+  "completed": 241970
+}
+```
+
+After all the files are downloaded, the final responses are:
+
+```json
+{
+    "status": "verifying sha256 digest"
+}
+{
+    "status": "writing manifest"
+}
+{
+    "status": "removing any unused layers"
+}
+{
+    "status": "success"
+}
+```
+
+if `stream` is set to false, then the response is a single JSON object:
+
+```json
+{
+  "status": "success"
 }
 ```
 
@@ -280,19 +889,21 @@ Upload a model to a model library. Requires registering for ollama.ai and adding
 
 - `name`: name of the model to push in the form of `<namespace>/<model>:<tag>`
 - `insecure`: (optional) allow insecure connections to the library. Only use this if you are pushing to your library during development.
-- `stream`: (optional) if `false` the response will be be returned as a single response object, rather than a stream of objects
+- `stream`: (optional) if `false` the response will be returned as a single response object, rather than a stream of objects
 
-### Request
+### Examples
+
+#### Request
 
 ```shell
-curl -X POST http://localhost:11434/api/push -d '{
+curl http://localhost:11434/api/push -d '{
   "name": "mattw/pygmalion:latest"
 }'
 ```
 
-### Response
+#### Response
 
-Streaming response that starts with:
+If `stream` is not specified, or set to `true`, a stream of JSON objects is returned:
 
 ```json
 { "status": "retrieving manifest" }
@@ -325,6 +936,12 @@ Finally, when the upload is complete:
 {"status":"success"}
 ```
 
+If `stream` is set to `false`, then the response is a single JSON object:
+
+```json
+{ "status": "success" }
+```
+
 ## Generate Embeddings
 
 ```shell
@@ -342,20 +959,22 @@ Advanced parameters:
 
 - `options`: additional model parameters listed in the documentation for the [Modelfile](./modelfile.md#valid-parameters-and-values) such as `temperature`
 
-### Request
+### Examples
+
+#### Request
 
 ```shell
-curl -X POST http://localhost:11434/api/embeddings -d '{
-  "model": "llama2:7b",
+curl http://localhost:11434/api/embeddings -d '{
+  "model": "llama2",
   "prompt": "Here is an article about llamas..."
 }'
 ```
 
-### Response
+#### Response
 
 ```json
 {
-  "embeddings": [
+  "embedding": [
     0.5670403838157654, 0.009260174818336964, 0.23178744316101074, -0.2916173040866852, -0.8924556970596313,
     0.8785552978515625, -0.34576427936553955, 0.5742510557174683, -0.04222835972905159, -0.137906014919281
   ]

docs/development.md

@@ -14,7 +14,13 @@ Install required tools:
 brew install go cmake gcc
 ```
 
-Get the required libraries:
+Optionally enable debugging and more verbose logging:
+
+```bash
+export CGO_CFLAGS="-g"
+```
+
+Get the required libraries and build the native LLM code:
 
 ```bash
 go generate ./...
@@ -34,6 +40,35 @@ Now you can run `ollama`:
 
 ## Building on Linux with GPU support
 
-- Install cmake and nvidia-cuda-toolkit
-- run `go generate ./...`
-- run `go build .`
+
+### Linux/Windows CUDA (NVIDIA)
+*Your operating system distribution may already have packages for NVIDIA CUDA. Distro packages are often preferable, but instructions are distro-specific. Please consult distro-specific docs for dependencies if available!*
+
+Install `cmake` and `golang` as well as [NVIDIA CUDA](https://developer.nvidia.com/cuda-downloads) development and runtime packages.
+Then generate dependencies:
+```
+go generate ./...
+```
+Then build the binary:
+```
+go build .
+```
+
+### Linux ROCm (AMD)
+*Your operating system distribution may already have packages for AMD ROCm and CLBlast. Distro packages are often preferable, but instructions are distro-specific. Please consult distro-specific docs for dependencies if available!*
+
+Install [CLBlast](https://github.com/CNugteren/CLBlast/blob/master/doc/installation.md) and [ROCm](https://rocm.docs.amd.com/en/latest/deploy/linux/quick_start.html) developement packages first, as well as `cmake` and `golang`.
+Adjust the paths below (correct for Arch) as appropriate for your distributions install locations and generate dependencies:
+```
+CLBlast_DIR=/usr/lib/cmake/CLBlast ROCM_PATH=/opt/rocm go generate ./...
+```
+Then build the binary:
+```
+go build .
+```
+
+ROCm requires elevated privileges to access the GPU at runtime.  On most distros you can add your user account to the `render` group, or run as root.
+
+## Containerized Build
+
+If you have Docker available, you can build linux binaries with `./scripts/build_linux.sh` which has the CUDA and ROCm dependencies included.

docs/faq.md

@@ -1,18 +1,137 @@
 # FAQ
 
-## How can I expose the Ollama server?
+## How can I upgrade Ollama?
 
-```bash
-OLLAMA_HOST=0.0.0.0:11435 ollama serve
+To upgrade Ollama, run the installation process again. On the Mac, click the Ollama icon in the menubar and choose the restart option if an update is available.
+
+## How can I view the logs?
+
+Review the [Troubleshooting](./troubleshooting.md) docs for more about using logs.
+
+## How do I use Ollama server environment variables on Mac
+
+On macOS, Ollama runs in the background and is managed by the menubar app. If adding environment variables, Ollama will need to be run manually.
+
+1. Click the menubar icon for Ollama and choose **Quit Ollama**.
+2. Open a new terminal window and run the following command (this example uses `OLLAMA_HOST` with an IP address of `123.1.1.1`):
+
+   ```bash
+   OLLAMA_HOST=123.1.1.1 ollama serve
+   ```
+
+## How do I use Ollama server environment variables on Linux?
+
+If Ollama is installed with the install script, a systemd service was created, running as the Ollama user. To add an environment variable, such as OLLAMA_HOST, follow these steps:
+
+1. Create a `systemd` drop-in directory and add a config file. This is only needed once.
+
+   ```bash
+   mkdir -p /etc/systemd/system/ollama.service.d
+   echo '[Service]' >>/etc/systemd/system/ollama.service.d/environment.conf
+   ```
+
+2. For each environment variable, add it to the config file:
+
+   ```bash
+   echo 'Environment="OLLAMA_HOST=0.0.0.0:11434"' >>/etc/systemd/system/ollama.service.d/environment.conf
+   ```
+
+3. Reload `systemd` and restart Ollama:
+
+   ```bash
+   systemctl daemon-reload
+   systemctl restart ollama
+   ```
+
+## How can I expose Ollama on my network?
+
+Ollama binds to 127.0.0.1 port 11434 by default. Change the bind address with the `OLLAMA_HOST` environment variable. Refer to the section above for how to use environment variables on your platform.
+
+## How can I allow additional web origins to access Ollama?
+
+Ollama allows cross-origin requests from `127.0.0.1` and `0.0.0.0` by default. Add additional origins with the `OLLAMA_ORIGINS` environment variable. For example, to add all ports on 192.168.1.1 and https://example.com, use:
+
+```shell
+OLLAMA_ORIGINS=http://192.168.1.1:*,https://example.com
 ```
 
-By default, Ollama allows cross origin requests from `127.0.0.1` and `0.0.0.0`. To support more origins, you can use the `OLLAMA_ORIGINS` environment variable:
-
-```bash
-OLLAMA_ORIGINS=http://192.168.1.1:*,https://example.com ollama serve
-```
+Refer to the section above for how to use environment variables on your platform.
 
 ## Where are models stored?
 
-* macOS: Raw model data is stored under `~/.ollama/models`.
-* Linux: Raw model data is stored under `/usr/share/ollama/.ollama/models`
+- macOS: `~/.ollama/models`.
+- Linux: `/usr/share/ollama/.ollama/models`
+
+## How do I set them to a different location?
+
+If a different directory needs to be used, set the environment variable `OLLAMA_MODELS` to the chosen directory. Refer to the section above for how to use environment variables on your platform.
+
+## Does Ollama send my prompts and answers back to Ollama.ai to use in any way?
+
+No, Ollama runs entirely locally, and conversation data will never leave your machine.
+
+## How can I use Ollama in Visual Studio Code?
+
+There is already a large collection of plugins available for VSCode as well as other editors that leverage Ollama. See the list of [extensions & plugins](https://github.com/jmorganca/ollama#extensions--plugins) at the bottom of the main repository readme.
+
+## How do I use Ollama behind a proxy?
+
+Ollama is compatible with proxy servers if `HTTP_PROXY` or `HTTPS_PROXY` are configured. When using either variables, ensure it is set where `ollama serve` can access the values. When using `HTTPS_PROXY`, ensure the proxy certificate is installed as a system certificate. Refer to the section above for how to use environment variables on your platform.
+
+### How do I use Ollama behind a proxy in Docker?
+
+The Ollama Docker container image can be configured to use a proxy by passing `-e HTTPS_PROXY=https://proxy.example.com` when starting the container.
+
+Alternatively, the Docker daemon can be configured to use a proxy. Instructions are available for Docker Desktop on [macOS](https://docs.docker.com/desktop/settings/mac/#proxies), [Windows](https://docs.docker.com/desktop/settings/windows/#proxies), and [Linux](https://docs.docker.com/desktop/settings/linux/#proxies), and Docker [daemon with systemd](https://docs.docker.com/config/daemon/systemd/#httphttps-proxy).
+
+Ensure the certificate is installed as a system certificate when using HTTPS. This may require a new Docker image when using a self-signed certificate.
+
+```dockerfile
+FROM ollama/ollama
+COPY my-ca.pem /usr/local/share/ca-certificates/my-ca.crt
+RUN update-ca-certificates
+```
+
+Build and run this image:
+
+```shell
+docker build -t ollama-with-ca .
+docker run -d -e HTTPS_PROXY=https://my.proxy.example.com -p 11434:11434 ollama-with-ca
+```
+
+## How do I use Ollama with GPU acceleration in Docker?
+
+The Ollama Docker container can be configured with GPU acceleration in Linux or Windows (with WSL2). This requires the [nvidia-container-toolkit](https://github.com/NVIDIA/nvidia-container-toolkit). See [ollama/ollama](https://hub.docker.com/r/ollama/ollama) for more details.
+
+GPU acceleration is not available for Docker Desktop in macOS due to the lack of GPU passthrough and emulation.
+
+## Why is networking slow in WSL2 on Windows 10?
+
+This can impact both installing Ollama, as well as downloading models.
+
+Open `Control Panel > Networking and Internet > View network status and tasks` and click on `Change adapter settings` on the left panel. Find the `vEthernel (WSL)` adapter, right click and select `Properties`.
+Click on `Configure` and open the `Advanced` tab. Search through each of the properties until you find `Large Send Offload Version 2 (IPv4)` and `Large Send Offload Version 2 (IPv6)`. *Disable* both of these
+properties.
+
+## What does the q in the model tag mean? What is quantization?
+
+Whenever you pull a model without a tag, Ollama will actually pull the q4_0 quantization of the model. You can verify this on the tags page. On https://ollama.ai/library/llama2/tags you can see that the hash for the latest tag matches the hash for the 7b model. ![quant hashes](https://github.com/jmorganca/ollama/assets/633681/814b1b78-8205-4845-89f9-e671b3b96085)
+
+Looking at the that page for any model, you can see several quantization options available. Quantization is a method of compression that allows the model to fit in less space and thus use less RAM and VRAM on your machine.
+
+At a high level, a model is made of an enormous collection of nodes that determine how to generate text. These nodes are connected at different levels with weights. The training process adjusts these weights to be able to output the right text every time.
+
+Most of the source models that we use start with weights that are 32bit floating-point numbers. Those weights, and another concept called biases, add up to be the parameters. So a source model with 7 billion parameters has 7 billion 32bit floating-point numbers, plus a description of all the nodes and more. That adds up to needing at least 28 Gigabytes of memory to load, if you choose to load one of those source models.
+
+Quantization turns those 32bit floating point weights into much smaller integers. The number next to the q indicates the bit size of the weights. So a q4 model converted those 32bit floats into 4bit integers. A 4bit quantization takes up the space for 7billion 4bit integers, plus a little overhead. That comes out to almost 4 Gigabytes. Obviously, there is some loss of information in this process of going from 30GB to 4GB, but it turns out in most cases it isn't really noticeable. In fact, even the 2bit quantization which fits in less than 3GB can be very useful.
+
+There are three major sets of quantizations you will see in the Ollama Library of models: **fp16**, models with just a q and a number, like **q4_0**, and then models with a **K** in the tag. The **fp16** model is one that has been converted and quantized from the source 32bit to 16bit. This will be about half the size of the 32bit source model and is the largest quantization we deliver in the library. The **q4_0**, **q4_1**, **q5_0**, etc. models use two different quantization methods that were the original methods.
+
+The models with a **K** are often referred to as K Quants. This is a method that allows for models of a similar quality but smaller than the original method used. Essentially, it finds clusters of weights and quantizes those together, allowing for higher precision while using the same bit sizes as the regular quantization options. But this requires a set of maps for the model to figure out the original values which have a computational cost. You may see some impact on the speed of models with K quants compared to the regular quantizations.
+
+## What is context, can I increase it, and why doesn't every model support a huge context?
+
+Context refers to the size of the input you can send to a model and get sensible output back. Many models have a context size of 2048 tokens. It's sometimes possible to give it more using the **num_ctx** parameter, but the answers start to degrade. This is because half of the context is "freed" up to allow for more memory. Newer models have been able to increase that context size using different methods. This increase in context size results in a corresponding increase in memory required, sometimes by orders of magnitude.
+
+> !WARNING]
+> Currently, over-allocating context size may result in model quality or stability issues.

docs/import.md

@@ -0,0 +1,195 @@
+# Import a model
+
+This guide walks through importing a GGUF, PyTorch or Safetensors model.
+
+## Importing (GGUF)
+
+### Step 1: Write a `Modelfile`
+
+Start by creating a `Modelfile`. This file is the blueprint for your model, specifying weights, parameters, prompt templates and more.
+
+```
+FROM ./mistral-7b-v0.1.Q4_0.gguf
+```
+
+(Optional) many chat models require a prompt template in order to answer correctly. A default prompt template can be specified with the `TEMPLATE` instruction in the `Modelfile`:
+
+```
+FROM ./q4_0.bin
+TEMPLATE "[INST] {{ .Prompt }} [/INST]"
+```
+
+### Step 2: Create the Ollama model
+
+Finally, create a model from your `Modelfile`:
+
+```
+ollama create example -f Modelfile
+```
+
+### Step 3: Run your model
+
+Next, test the model with `ollama run`:
+
+```
+ollama run example "What is your favourite condiment?"
+```
+
+## Importing (PyTorch & Safetensors)
+
+### Supported models
+
+Ollama supports a set of model architectures, with support for more coming soon:
+
+- Llama & Mistral
+- Falcon & RW
+- BigCode
+
+To view a model's architecture, check the `config.json` file in its HuggingFace repo. You should see an entry under `architectures` (e.g. `LlamaForCausalLM`).
+
+### Step 1: Clone the HuggingFace repository (optional)
+
+If the model is currently hosted in a HuggingFace repository, first clone that repository to download the raw model.
+
+```
+git lfs install
+git clone https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1
+cd Mistral-7B-Instruct-v0.1
+```
+
+### Step 2: Convert and quantize to a `.bin` file (optional, for PyTorch and Safetensors)
+
+If the model is in PyTorch or Safetensors format, a [Docker image](https://hub.docker.com/r/ollama/quantize) with the tooling required to convert and quantize models is available.
+
+First, Install [Docker](https://www.docker.com/get-started/).
+
+Next, to convert and quantize your model, run:
+
+```
+docker run --rm -v .:/model ollama/quantize -q q4_0 /model
+```
+
+This will output two files into the directory:
+
+- `f16.bin`: the model converted to GGUF
+- `q4_0.bin` the model quantized to a 4-bit quantization (Ollama will use this file to create the Ollama model)
+
+### Step 3: Write a `Modelfile`
+
+Next, create a `Modelfile` for your model:
+
+```
+FROM ./q4_0.bin
+```
+
+(Optional) many chat models require a prompt template in order to answer correctly. A default prompt template can be specified with the `TEMPLATE` instruction in the `Modelfile`:
+
+```
+FROM ./q4_0.bin
+TEMPLATE "[INST] {{ .Prompt }} [/INST]"
+```
+
+### Step 4: Create the Ollama model
+
+Finally, create a model from your `Modelfile`:
+
+```
+ollama create example -f Modelfile
+```
+
+### Step 5: Run your model
+
+Next, test the model with `ollama run`:
+
+```
+ollama run example "What is your favourite condiment?"
+```
+
+## Publishing your model (optional – early alpha)
+
+Publishing models is in early alpha. If you'd like to publish your model to share with others, follow these steps:
+
+1. Create [an account](https://ollama.ai/signup)
+2. Run `cat ~/.ollama/id_ed25519.pub` to view your Ollama public key. Copy this to the clipboard.
+3. Add your public key to your [Ollama account](https://ollama.ai/settings/keys)
+
+Next, copy your model to your username's namespace:
+
+```
+ollama cp example <your username>/example
+```
+
+Then push the model:
+
+```
+ollama push <your username>/example
+```
+
+After publishing, your model will be available at `https://ollama.ai/<your username>/example`.
+
+## Quantization reference
+
+The quantization options are as follow (from highest highest to lowest levels of quantization). Note: some architectures such as Falcon do not support K quants.
+
+- `q2_K`
+- `q3_K`
+- `q3_K_S`
+- `q3_K_M`
+- `q3_K_L`
+- `q4_0` (recommended)
+- `q4_1`
+- `q4_K`
+- `q4_K_S`
+- `q4_K_M`
+- `q5_0`
+- `q5_1`
+- `q5_K`
+- `q5_K_S`
+- `q5_K_M`
+- `q6_K`
+- `q8_0`
+- `f16`
+
+## Manually converting & quantizing models
+
+### Prerequisites
+
+Start by cloning the `llama.cpp` repo to your machine in another directory:
+
+```
+git clone https://github.com/ggerganov/llama.cpp.git
+cd llama.cpp
+```
+
+Next, install the Python dependencies:
+
+```
+pip install -r requirements.txt
+```
+
+Finally, build the `quantize` tool:
+
+```
+make quantize
+```
+
+### Convert the model
+
+Run the correct conversion script for your model architecture:
+
+```shell
+# LlamaForCausalLM or MistralForCausalLM
+python convert.py <path to model directory>
+
+# FalconForCausalLM
+python convert-falcon-hf-to-gguf.py <path to model directory>
+
+# GPTBigCodeForCausalLM
+python convert-starcoder-hf-to-gguf.py <path to model directory>
+```
+
+### Quantize the model
+
+```
+quantize <path to model dir>/ggml-model-f32.bin <path to model dir>/q4_0.bin q4_0
+```

docs/linux.md

@@ -1,12 +1,16 @@
-# Installing Ollama on Linux
+# Ollama on Linux
 
-> Note: A one line installer for Ollama is available by running:
+## Install
+
+Install Ollama running this one-liner:
 >
-> ```bash
-> curl https://ollama.ai/install.sh | sh
-> ```
+```bash
+curl https://ollama.ai/install.sh | sh
+```
 
-## Download the `ollama` binary
+## Manual install
+
+### Download the `ollama` binary
 
 Ollama is distributed as a self-contained binary. Download it to a directory in your PATH:
 
@@ -15,31 +19,7 @@ sudo curl -L https://ollama.ai/download/ollama-linux-amd64 -o /usr/bin/ollama
 sudo chmod +x /usr/bin/ollama
 ```
 
-## Start Ollama
-
-Start Ollama by running `ollama serve`:
-
-```bash
-ollama serve
-```
-
-Once Ollama is running, run a model in another terminal session:
-
-```bash
-ollama run llama2
-```
-
-## Install CUDA drivers (optional – for Nvidia GPUs)
-
-[Download and install](https://developer.nvidia.com/cuda-downloads) CUDA.
-
-Verify that the drivers are installed by running the following command, which should print details about your GPU:
-
-```bash
-nvidia-smi
-```
-
-## Adding Ollama as a startup service (optional)
+### Adding Ollama as a startup service (recommended)
 
 Create a user for Ollama:
 
@@ -60,7 +40,6 @@ User=ollama
 Group=ollama
 Restart=always
 RestartSec=3
-Environment="HOME=/usr/share/ollama"
 
 [Install]
 WantedBy=default.target
@@ -73,7 +52,40 @@ sudo systemctl daemon-reload
 sudo systemctl enable ollama
 ```
 
-### Viewing logs
+### Install CUDA drivers (optional – for Nvidia GPUs)
+
+[Download and install](https://developer.nvidia.com/cuda-downloads) CUDA.
+
+Verify that the drivers are installed by running the following command, which should print details about your GPU:
+
+```bash
+nvidia-smi
+```
+
+### Start Ollama
+
+Start Ollama using `systemd`:
+
+```bash
+sudo systemctl start ollama
+```
+
+## Update
+
+Update ollama by running the install script again:
+
+```bash
+curl https://ollama.ai/install.sh | sh
+```
+
+Or by downloading the ollama binary:
+
+```bash
+sudo curl -L https://ollama.ai/download/ollama-linux-amd64 -o /usr/bin/ollama
+sudo chmod +x /usr/bin/ollama
+```
+
+## Viewing logs
 
 To view logs of Ollama running as a startup service, run:
 
@@ -81,3 +93,24 @@ To view logs of Ollama running as a startup service, run:
 journalctl -u ollama
 ```
 
+## Uninstall
+
+Remove the ollama service:
+
+```bash
+sudo systemctl stop ollama
+sudo systemctl disable ollama
+sudo rm /etc/systemd/system/ollama.service
+```
+
+Remove the ollama binary from your bin directory (either `/usr/local/bin`, `/usr/bin`, or `/bin`):
+
+```bash
+sudo rm $(which ollama)
+```
+
+Remove the downloaded models and Ollama service user:
+```bash
+sudo rm -r /usr/share/ollama
+sudo userdel ollama
+```

docs/modelfile.md

@@ -1,6 +1,6 @@
 # Ollama Model File
 
-> Note: this `Modelfile` syntax is in development
+> Note: `Modelfile` syntax is in development
 
 A model file is the blueprint to create and share models with Ollama.
 
@@ -12,7 +12,6 @@ A model file is the blueprint to create and share models with Ollama.
   - [FROM (Required)](#from-required)
     - [Build from llama2](#build-from-llama2)
     - [Build from a bin file](#build-from-a-bin-file)
-  - [EMBED](#embed)
   - [PARAMETER](#parameter)
     - [Valid Parameters and Values](#valid-parameters-and-values)
   - [TEMPLATE](#template)
@@ -31,17 +30,19 @@ The format of the `Modelfile`:
 INSTRUCTION arguments
 ```
 
-| Instruction                         | Description                                                   |
-| ----------------------------------- | ------------------------------------------------------------- |
-| [`FROM`](#from-required) (required) | Defines the base model to use.                                |
-| [`PARAMETER`](#parameter)           | Sets the parameters for how Ollama will run the model.        |
-| [`TEMPLATE`](#template)             | The full prompt template to be sent to the model.             |
-| [`SYSTEM`](#system)                 | Specifies the system prompt that will be set in the template. |
-| [`ADAPTER`](#adapter)               | Defines the (Q)LoRA adapters to apply to the model.           |
-| [`LICENSE`](#license)               | Specifies the legal license.                                  |
+| Instruction                         | Description                                                    |
+| ----------------------------------- | -------------------------------------------------------------- |
+| [`FROM`](#from-required) (required) | Defines the base model to use.                                 |
+| [`PARAMETER`](#parameter)           | Sets the parameters for how Ollama will run the model.         |
+| [`TEMPLATE`](#template)             | The full prompt template to be sent to the model.              |
+| [`SYSTEM`](#system)                 | Specifies the system message that will be set in the template. |
+| [`ADAPTER`](#adapter)               | Defines the (Q)LoRA adapters to apply to the model.            |
+| [`LICENSE`](#license)               | Specifies the legal license.                                   |
 
 ## Examples
 
+### Basic `Modelfile`
+
 An example of a `Modelfile` creating a mario blueprint:
 
 ```modelfile
@@ -51,7 +52,7 @@ PARAMETER temperature 1
 # sets the context window size to 4096, this controls how many tokens the LLM can use as context to generate the next token
 PARAMETER num_ctx 4096
 
-# sets a custom system prompt to specify the behavior of the chat assistant
+# sets a custom system message to specify the behavior of the chat assistant
 SYSTEM You are Mario from super mario bros, acting as an assistant.
 ```
 
@@ -64,6 +65,35 @@ To use this:
 
 More examples are available in the [examples directory](../examples).
 
+### `Modelfile`s in [ollama.ai/library][1]
+
+There are two ways to view `Modelfile`s underlying the models in [ollama.ai/library][1]:
+
+- Option 1: view a details page from a model's tags page:
+  1.  Go to a particular model's tags (e.g. https://ollama.ai/library/llama2/tags)
+  2.  Click on a tag (e.g. https://ollama.ai/library/llama2:13b)
+  3.  Scroll down to "Layers"
+      - Note: if the [`FROM` instruction](#from-required) is not present,
+        it means the model was created from a local file
+- Option 2: use `ollama show` to print the `Modelfile` for any local models like so:
+
+  ```bash
+  > ollama show --modelfile llama2:13b
+  # Modelfile generated by "ollama show"
+  # To build a new Modelfile based on this one, replace the FROM line with:
+  # FROM llama2:13b
+
+  FROM /root/.ollama/models/blobs/sha256:123abc
+  TEMPLATE """[INST] {{ if and .First .System }}<<SYS>>{{ .System }}<</SYS>>
+
+  {{ end }}{{ .Prompt }} [/INST] """
+  SYSTEM """"""
+  PARAMETER stop [INST]
+  PARAMETER stop [/INST]
+  PARAMETER stop <<SYS>>
+  PARAMETER stop <</SYS>>
+  ```
+
 ## Instructions
 
 ### FROM (Required)
@@ -91,17 +121,6 @@ FROM ./ollama-model.bin
 
 This bin file location should be specified as an absolute path or relative to the `Modelfile` location.
 
-### EMBED
-
-The `EMBED` instruction is used to add embeddings of files to a model. This is useful for adding custom data that the model can reference when generating an answer. Note that currently only text files are supported, formatted with each line as one embedding.
-
-```modelfile
-FROM <model name>:<tag>
-EMBED <file path>.txt
-EMBED <different file path>.txt
-EMBED <path to directory>/*.txt
-```
-
 ### PARAMETER
 
 The `PARAMETER` instruction defines a parameter that can be set when the model is run.
@@ -124,8 +143,8 @@ PARAMETER <parameter> <parametervalue>
 | repeat_last_n  | Sets how far back for the model to look back to prevent repetition. (Default: 64, 0 = disabled, -1 = num_ctx)                                                                                                                                           | int        | repeat_last_n 64     |
 | repeat_penalty | Sets how strongly to penalize repetitions. A higher value (e.g., 1.5) will penalize repetitions more strongly, while a lower value (e.g., 0.9) will be more lenient. (Default: 1.1)                                                                     | float      | repeat_penalty 1.1   |
 | temperature    | The temperature of the model. Increasing the temperature will make the model answer more creatively. (Default: 0.8)                                                                                                                                     | float      | temperature 0.7      |
-| seed | Sets the random number seed to use for generation. Setting this to a specific number will make the model generate the same text for the same prompt. | int | seed 42 |
-| stop           | Sets the stop sequences to use.                                                                                                                                                                                                                         | string     | stop "AI assistant:" |
+| seed           | Sets the random number seed to use for generation. Setting this to a specific number will make the model generate the same text for the same prompt. (Default: 0)                                                                                       | int        | seed 42              |
+| stop           | Sets the stop sequences to use. When this pattern is encountered the LLM will stop generating text and return. Multiple stop patterns may be set by specifying multiple separate `stop` parameters in a modelfile.                                      | string     | stop "AI assistant:" |
 | tfs_z          | Tail free sampling is used to reduce the impact of less probable tokens from the output. A higher value (e.g., 2.0) will reduce the impact more, while a value of 1.0 disables this setting. (default: 1)                                               | float      | tfs_z 1              |
 | num_predict    | Maximum number of tokens to predict when generating text. (Default: 128, -1 = infinite generation, -2 = fill context)                                                                                                                                   | int        | num_predict 42       |
 | top_k          | Reduces the probability of generating nonsense. A higher value (e.g. 100) will give more diverse answers, while a lower value (e.g. 10) will be more conservative. (Default: 40)                                                                        | int        | top_k 40             |
@@ -133,15 +152,15 @@ PARAMETER <parameter> <parametervalue>
 
 ### TEMPLATE
 
-`TEMPLATE` of the full prompt template to be passed into the model. It may include (optionally) a system prompt and a user's prompt. This is used to create a full custom prompt, and syntax may be model specific. You can usually find the template for a given model in the readme for that model.
+`TEMPLATE` of the full prompt template to be passed into the model. It may include (optionally) a system message and a user's prompt. This is used to create a full custom prompt, and syntax may be model specific. You can usually find the template for a given model in the readme for that model.
 
 #### Template Variables
 
-| Variable        | Description                                                                                                  |
-| --------------- | ------------------------------------------------------------------------------------------------------------ |
-| `{{ .System }}` | The system prompt used to specify custom behavior, this must also be set in the Modelfile as an instruction. |
-| `{{ .Prompt }}` | The incoming prompt, this is not specified in the model file and will be set based on input.                 |
-| `{{ .First }}`  | A boolean value used to render specific template information for the first generation of a session.          |
+| Variable        | Description                                                                                                   |
+| --------------- | ------------------------------------------------------------------------------------------------------------- |
+| `{{ .System }}` | The system message used to specify custom behavior, this must also be set in the Modelfile as an instruction. |
+| `{{ .Prompt }}` | The incoming prompt, this is not specified in the model file and will be set based on input.                  |
+| `{{ .First }}`  | A boolean value used to render specific template information for the first generation of a session.           |
 
 ```modelfile
 TEMPLATE """
@@ -161,7 +180,7 @@ SYSTEM """<system message>"""
 
 ### SYSTEM
 
-The `SYSTEM` instruction specifies the system prompt to be used in the template, if applicable.
+The `SYSTEM` instruction specifies the system message to be used in the template, if applicable.
 
 ```modelfile
 SYSTEM """<system message>"""
@@ -187,5 +206,7 @@ LICENSE """
 
 ## Notes
 
-- the **`Modelfile` is not case sensitive**. In the examples, we use uppercase for instructions to make it easier to distinguish it from arguments.
-- Instructions can be in any order. In the examples, we start with FROM instruction to keep it easily readable.
+- the **`Modelfile` is not case sensitive**. In the examples, uppercase instructions are used to make it easier to distinguish it from arguments.
+- Instructions can be in any order. In the examples, the `FROM` instruction is first to keep it easily readable.
+
+[1]: https://ollama.ai/library

docs/quantize.md

@@ -1,111 +0,0 @@
-# How to Quantize a Model
-
-Sometimes the model you want to work with is not available at [https://ollama.ai/library](https://ollama.ai/library).
-
-## Figure out if we can run the model?
-
-Not all models will work with Ollama. There are a number of factors that go into whether we are able to work with the next cool model. First it has to work with llama.cpp. Then we have to have implemented the features of llama.cpp that it requires. And then, sometimes, even with both of those, the model might not work...
-
-1. What is the model you want to convert and upload?
-2. Visit the model's page on HuggingFace.
-3. Switch to the **Files and versions** tab.
-4. Click on the **config.json** file. If there is no config.json file, it may not work.
-5. Take note of the **architecture** list in the json file.
-6. Does any entry in the list match one of the following architectures?
-    1. LlamaForCausalLM
-    2. MistralForCausalLM
-    3. RWForCausalLM
-    4. FalconForCausalLM
-    5. GPTNeoXForCausalLM
-    6. GPTBigCodeForCausalLM
-7. If the answer is yes, then there is a good chance the model will run after being converted and quantized.
-8. An alternative to this process is to visit [https://caniquant.tvl.st](https://caniquant.tvl.st) and enter the org/modelname in the box and submit.
-
-At this point there are two processes you can use. You can either use a Docker container to convert and quantize, OR you can manually run the scripts. The Docker container is the easiest way to do it, but it requires you to have Docker installed on your machine. If you don't have Docker installed, you can follow the manual process.
-
-## Convert and Quantize with Docker
-
-Run `docker run --rm -v /path/to/model/repo:/repo ollama/quantize -q quantlevel /repo`. For instance, if you have downloaded the latest Mistral 7B model, then clone it to your machine. Then change into that directory and you can run:
-
-```shell
-docker run --rm -v .:/repo ollama/quantize -q q4_0 /repo
-```
-
-You can find the different quantization levels below under **Quantize the Model**.
-
-This will output two files into the directory. First is a f16.bin file that is the model converted to GGUF. The second file is a q4_0.bin file which is the model quantized to a 4 bit quantization. You should rename it to something more descriptive.
-
-You can find the repository for the Docker container here: [https://github.com/mxyng/quantize](https://github.com/mxyng/quantize)
-
-For instance, if you wanted to convert the Mistral 7B model to a Q4 quantized model, then you could go through the following steps:
-
-1. First verify the model will potentially work.
-2. Now clone Mistral 7B to your machine. You can find the command to run when you click the three vertical dots button on the model page, then click **Clone Repository**.
-   1. For this repo, the command is:
-
-      ```shell
-      git lfs install
-      git clone https://huggingface.co/mistralai/Mistral-7B-v0.1
-      ```
-
-   2. Navigate into the new directory and run `docker run --rm -v .:/repo ollama/quantize -q q4_0 /repo`
-   3. Now you can create a modelfile using the q4_0.bin file that was created.
-
-## Convert and Quantize Manually
-
-### Clone llama.cpp to your machine
-
-If we know the model has a chance of working, then we need to convert and quantize. This is a matter of running two separate scripts in the llama.cpp project.
-
-1. Decide where you want the llama.cpp repository on your machine.
-2. Navigate to that location and then run:
- [`git clone https://github.com/ggerganov/llama.cpp.git`](https://github.com/ggerganov/llama.cpp.git)
-    1. If you don't have git installed, download this zip file and unzip it to that location: https://github.com/ggerganov/llama.cpp/archive/refs/heads/master.zip
-3. Install the Python dependencies: `pip install torch transformers sentencepiece`
-4. Run 'make' to build the project and the quantize executable.
-
-### Convert the model to GGUF
-
-1. Decide on the right convert script to run. What was the model architecture you found in the first section.
-    1. LlamaForCausalLM or MistralForCausalLM:
-    run `python3 convert.py <modelfilename>`
-    No need to specify fp16 or fp32.
-    2. FalconForCausalLM or RWForCausalLM:
-    run `python3 convert-falcon-hf-to-gguf.py <modelfilename> <fpsize>`  
-    fpsize depends on the weight size. 1 for fp16, 0 for fp32
-    3. GPTNeoXForCausalLM:
-    run `python3 convert-gptneox-hf-to-gguf.py <modelfilename> <fpsize>`
-    fpsize depends on the weight size. 1 for fp16, 0 for fp32
-    4. GPTBigCodeForCausalLM:
-    run `python3 convert-starcoder-hf-to-gguf.py <modelfilename> <fpsize>`
-    fpsize depends on the weight size. 1 for fp16, 0 for fp32
-
-### Quantize the model
-
-If the model converted successfully, there is a good chance it will also quantize successfully. Now you need to decide on the quantization to use. We will always try to create all the quantizations and upload them to the library. You should decide which level is more important to you and quantize accordingly.
-
-The quantization options are as follows. Note that some architectures such as Falcon do not support K quants.
-
-- Q4_0
-- Q4_1
-- Q5_0
-- Q5_1
-- Q2_K
-- Q3_K
-- Q3_K_S
-- Q3_K_M
-- Q3_K_L
-- Q4_K
-- Q4_K_S
-- Q4_K_M
-- Q5_K
-- Q5_K_S
-- Q5_K_M
-- Q6_K
-- Q8_0
-
-Run the following command `quantize <converted model from above> <output file> <quantization type>`
-
-## Now Create the Model
-
-Now you can create the Ollama model. Refer to the [modelfile](./modelfile.md) doc for more information on doing that.

docs/troubleshooting.md

@@ -0,0 +1,22 @@
+# How to troubleshoot issues
+
+Sometimes Ollama may not perform as expected. One of the best ways to figure out what happened is to take a look at the logs. Find the logs on Mac by running the command:
+
+```shell
+cat ~/.ollama/logs/server.log
+```
+
+On Linux systems with systemd, the logs can be found with this command:
+
+```shell
+journalctl -u ollama
+```
+
+If manually running `ollama serve` in a terminal, the logs will be on that terminal.
+
+Join the [Discord](https://discord.gg/ollama) for help interpreting the logs.
+
+## Known issues
+
+
+* `signal: illegal instruction (core dumped)`: Ollama requires AVX support from the CPU. This was introduced in 2011 and CPUs started offering it in 2012. CPUs from before that and some lower end CPUs after that may not have AVX support and thus are not supported by Ollama. Some users have had luck with building Ollama on their machines disabling the need for AVX.

docs/tutorials.md

@@ -4,5 +4,6 @@ Here is a list of ways you can use Ollama with other tools to build interesting
 
 - [Using LangChain with Ollama in JavaScript](./tutorials/langchainjs.md)
 - [Using LangChain with Ollama in Python](./tutorials/langchainpy.md)
+- [Running Ollama on NVIDIA Jetson Devices](./tutorials/nvidia-jetson.md)
 
-Also be sure to check out the [examples](../examples) directory for more ways to use Ollama.
+Also be sure to check out the [examples](../examples) directory for more ways to use Ollama.

docs/tutorials/fly-gpu.md

@@ -0,0 +1,83 @@
+# Running Ollama on Fly.io GPU Instances
+
+Ollama runs with little to no configuration on [Fly.io GPU instances](https://fly.io/docs/gpus/gpu-quickstart/). If you don't have access to GPUs yet, you'll need to [apply for access](https://fly.io/gpu/) on the waitlist. Once you're accepted, you'll get an email with instructions on how to get started.
+
+Create a new app with `fly apps create`:
+
+```bash
+fly apps create
+```
+
+Then create a `fly.toml` file in a new folder that looks like this:
+
+```toml
+app = "sparkling-violet-709"
+primary_region = "ord"
+vm.size = "a100-40gb" # see https://fly.io/docs/gpus/gpu-quickstart/ for more info
+
+[build]
+  image = "ollama/ollama"
+
+[http_service]
+  internal_port = 11434
+  force_https = false
+  auto_stop_machines = true
+  auto_start_machines = true
+  min_machines_running = 0
+  processes = ["app"]
+
+[mounts]
+  source = "models"
+  destination = "/root/.ollama"
+  initial_size = "100gb"
+```
+
+Then create a [new private IPv6 address](https://fly.io/docs/reference/private-networking/#flycast-private-load-balancing) for your app:
+
+```bash
+fly ips allocate-v6 --private
+```
+
+Then deploy your app:
+
+```bash
+fly deploy
+```
+
+And finally you can access it interactively with a new Fly.io Machine:
+
+```
+fly machine run -e OLLAMA_HOST=http://your-app-name.flycast --shell ollama/ollama
+```
+
+```bash
+$ ollama run openchat:7b-v3.5-fp16
+>>> How do I bake chocolate chip cookies?
+ To bake chocolate chip cookies, follow these steps:
+
+1. Preheat the oven to 375°F (190°C) and line a baking sheet with parchment paper or silicone baking mat.
+
+2. In a large bowl, mix together 1 cup of unsalted butter (softened), 3/4 cup granulated sugar, and 3/4
+cup packed brown sugar until light and fluffy.
+
+3. Add 2 large eggs, one at a time, to the butter mixture, beating well after each addition. Stir in 1
+teaspoon of pure vanilla extract.
+
+4. In a separate bowl, whisk together 2 cups all-purpose flour, 1/2 teaspoon baking soda, and 1/2 teaspoon
+salt. Gradually add the dry ingredients to the wet ingredients, stirring until just combined.
+
+5. Fold in 2 cups of chocolate chips (or chunks) into the dough.
+
+6. Drop rounded tablespoons of dough onto the prepared baking sheet, spacing them about 2 inches apart.
+
+7. Bake for 10-12 minutes, or until the edges are golden brown. The centers should still be slightly soft.
+
+8. Allow the cookies to cool on the baking sheet for a few minutes before transferring them to a wire rack
+to cool completely.
+
+Enjoy your homemade chocolate chip cookies!
+```
+
+When you set it up like this, it will automatically turn off when you're done using it. Then when you access it again, it will automatically turn back on. This is a great way to save money on GPU instances when you're not using them. If you want a persistent wake-on-use connection to your Ollama instance, you can set up a [connection to your Fly network using WireGuard](https://fly.io/docs/reference/private-networking/#discovering-apps-through-dns-on-a-wireguard-connection). Then you can access your Ollama instance at `http://your-app-name.flycast`.
+
+And that's it!

docs/tutorials/langchainjs.md

@@ -23,13 +23,17 @@ const answer = await ollama.call(`why is the sky blue?`);
 console.log(answer);
 ```
 
-That will get us the same thing as if we ran `ollama run llama2 "why is the sky blue"` in the terminal. But we want to load a document from the web to ask a question against. **Cheerio** is a great library for ingesting a webpage, and **LangChain** uses it in their **CheerioWebBaseLoader**. So let's build that part of the app.
+That will get us the same thing as if we ran `ollama run llama2 "why is the sky blue"` in the terminal. But we want to load a document from the web to ask a question against. **Cheerio** is a great library for ingesting a webpage, and **LangChain** uses it in their **CheerioWebBaseLoader**. So let's install **Cheerio** and build that part of the app.
+
+```bash
+npm install cheerio 
+```
 
 ```javascript
 import { CheerioWebBaseLoader } from "langchain/document_loaders/web/cheerio";
 
 const loader = new CheerioWebBaseLoader("https://en.wikipedia.org/wiki/2023_Hawaii_wildfires");
-const data = loader.load();
+const data = await loader.load();
 ```
 
 That will load the document. Although this page is smaller than the Odyssey, it is certainly bigger than the context size for most LLMs. So we are going to need to split into smaller pieces, and then select just the pieces relevant to our question. This is a great use for a vector datastore. In this example, we will use the **MemoryVectorStore** that is part of **LangChain**. But there is one more thing we need to get the content into the datastore. We have to run an embeddings process that converts the tokens in the text into a series of vectors. And for that, we are going to use **Tensorflow**. There is a lot of stuff going on in this one. First, install the **Tensorflow** components that we need.

docs/tutorials/langchainpy.md

@@ -42,12 +42,13 @@ text_splitter=RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=0)
 all_splits = text_splitter.split_documents(data)
 ```
 
-It's split up, but we have to find the relevant splits and then submit those to the model. We can do this by creating embeddings and storing them in a vector database. For now, we don't have embeddings built in to Ollama, though we will be adding that soon, so for now, we can use the GPT4All library for that. We will use ChromaDB in this example for a vector database. `pip install GPT4All chromadb`
+It's split up, but we have to find the relevant splits and then submit those to the model. We can do this by creating embeddings and storing them in a vector database. We can use Ollama directly to instantiate an embedding model. We will use ChromaDB in this example for a vector database. `pip install GPT4All chromadb`
 
 ```python
-from langchain.embeddings import GPT4AllEmbeddings
+from langchain.embeddings import OllamaEmbeddings
 from langchain.vectorstores import Chroma
-vectorstore = Chroma.from_documents(documents=all_splits, embedding=GPT4AllEmbeddings())
+oembed = OllamaEmbeddings(base_url="http://localhost:11434", model="llama2")
+vectorstore = Chroma.from_documents(documents=all_splits, embedding=oembed)
 ```
 
 Now let's ask a question from the document. **Who was Neleus, and who is in his family?** Neleus is a character in the Odyssey, and the answer can be found in our text.

docs/tutorials/nvidia-jetson.md

@@ -0,0 +1,38 @@
+# Running Ollama on NVIDIA Jetson Devices
+
+With some minor configuration, Ollama runs well on [NVIDIA Jetson Devices](https://www.nvidia.com/en-us/autonomous-machines/embedded-systems/). The following has been tested on [JetPack 5.1.2](https://developer.nvidia.com/embedded/jetpack).
+
+NVIDIA Jetson devices are Linux-based embedded AI computers that are purpose-built for AI applications.
+
+Jetsons have an integrated GPU that is wired directly to the memory controller of the machine. For this reason, the `nvidia-smi` command is unrecognized, and Ollama proceeds to operate in "CPU only"
+mode. This can be verified by using a monitoring tool like jtop.
+
+In order to address this, we simply pass the path to the Jetson's pre-installed CUDA libraries into `ollama serve` (while in a tmux session). We then hardcode the num_gpu parameters into a cloned
+version of our target model.
+
+Prerequisites:
+
+- curl
+- tmux
+
+Here are the steps:
+
+- Install Ollama via standard Linux command (ignore the 404 error): `curl https://ollama.ai/install.sh | sh`
+- Stop the Ollama service: `sudo systemctl stop ollama`
+- Start Ollama serve in a tmux session called ollama_jetson and reference the CUDA libraries path: `tmux has-session -t ollama_jetson 2>/dev/null || tmux new-session -d -s ollama_jetson 
+'LD_LIBRARY_PATH=/usr/local/cuda/lib64 ollama serve'`
+- Pull the model you want to use (e.g. mistral): `ollama pull mistral`
+- Create a new Modelfile specifically for enabling GPU support on the Jetson: `touch ModelfileMistralJetson`
+- In the ModelfileMistralJetson file, specify the FROM model and the num_gpu PARAMETER as shown below:
+
+```
+FROM mistral
+PARAMETER num_gpu 999
+```
+
+- Create a new model from your Modelfile: `ollama create mistral-jetson -f ./ModelfileMistralJetson`
+- Run the new model: `ollama run mistral-jetson`
+
+If you run a monitoring tool like jtop you should now see that Ollama is using the Jetson's integrated GPU.
+
+And that's it!

examples/.gitignore

@@ -1,7 +1,10 @@
 node_modules
+bun.lockb
+.vscode
 # OSX
 .DS_STORE
 
+
 # Models
 models/
 

examples/bash-comparemodels/README.md

@@ -0,0 +1,10 @@
+# Bash Shell examples
+
+When calling `ollama`, you can pass it a file to run all the prompts in the file, one after the other:
+
+`ollama run llama2 < sourcequestions.txt`
+
+This concept is used in the following example.
+
+## Compare Models
+`comparemodels.sh` is a script that runs all the questions in `sourcequestions.txt` using any 4 models you choose that you have already pulled from the Ollama library or have created locally.

examples/bash-comparemodels/comparemodels.sh

@@ -0,0 +1,64 @@
+#! /usr/bin/env bash
+# Compare multiple models by running them with the same questions
+
+NUMBEROFCHOICES=4
+SELECTIONS=()
+declare -a SUMS=()
+
+# Get the list of models
+CHOICES=$(ollama list | awk '{print $1}')
+
+# Select which models to run as a comparison
+echo "Select $NUMBEROFCHOICES models to compare:"
+select ITEM in $CHOICES; do
+    if [[ -n $ITEM ]]; then
+        echo "You have selected $ITEM"
+        SELECTIONS+=("$ITEM")
+        ((COUNT++))
+        if [[ $COUNT -eq $NUMBEROFCHOICES ]]; then
+            break
+        fi
+    else
+        echo "Invalid selection"
+    fi
+done
+
+# Loop through each of the selected models
+for ITEM in "${SELECTIONS[@]}"; do
+    echo "--------------------------------------------------------------"
+    echo "Loading the model $ITEM into memory"
+    ollama run "$ITEM" ""
+    echo "--------------------------------------------------------------"
+    echo "Running the questions through the model $ITEM"
+    COMMAND_OUTPUT=$(ollama run "$ITEM" --verbose < sourcequestions.txt 2>&1| tee /dev/stderr)
+
+    # eval duration is sometimes listed in seconds and sometimes in milliseconds. 
+    # Add up the values for each model
+    SUM=$(echo "$COMMAND_OUTPUT" | awk '
+    /eval duration:/ {
+        value = $3
+        if (index(value, "ms") > 0) {
+            gsub("ms", "", value)
+            value /= 1000
+        } else {
+            gsub("s", "", value)
+        }
+        sum += value
+    }
+    END { print sum }')
+
+
+    SUMS+=("All questions for $ITEM completed in $SUM seconds")
+done
+
+echo ""
+echo "--------------------------------------------------------------"
+echo -e "Sums of eval durations for each run:"
+for val in "${SUMS[@]}"; do
+    echo "$val"
+done
+
+echo "--------------------------------------------------------------"
+echo "Comparison complete. Now you can decide"
+echo "which model is best."
+echo "--------------------------------------------------------------"

examples/bash-comparemodels/sourcequestions.txt

@@ -0,0 +1,7 @@
+Why is the sky blue
+What is a black hole
+Explain the big bang theory like I am 5?
+What is the quickest way to win a game of Monopoly with 3 others?
+Why does a vacuum bottle keep my coffee hot and my milkshake cold?
+What is the difference between a meteor, a meteorite, and a meteoroid?
+Create an array with 5 items and print to the console. Do this in Python, C#, Typescript, and Rust.

examples/golang-simplegenerate/main.go

@@ -3,10 +3,10 @@ package main
 import (
 	"bytes"
 	"fmt"
-	"net/http"
-	"os"
 	"io"
 	"log"
+	"net/http"
+	"os"
 )
 
 func main() {
@@ -16,8 +16,10 @@ func main() {
 	if err != nil {
 		fmt.Print(err.Error())
 		os.Exit(1)
-	} 
+	}
 
+	defer resp.Body.Close()
+	
 	responseData, err := io.ReadAll(resp.Body)
 	if err != nil {
 		log.Fatal(err)

examples/jupyter-notebook/README.md

@@ -0,0 +1,5 @@
+# Ollama Jupyter Notebook
+
+This example downloads and installs Ollama in a Jupyter instance such as Google Colab. It will start the Ollama service and expose an endpoint using `ngrok` which can be used to communicate with the Ollama instance remotely.
+
+For best results, use an instance with GPU accelerator.

examples/jupyter-notebook/ollama.ipynb

@@ -0,0 +1,102 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "93f59dcb-c588-41b8-a792-55d88ade739c",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Download and run the Ollama Linux install script\n",
+    "!curl https://ollama.ai/install.sh | sh\n",
+    "!command -v systemctl >/dev/null && sudo systemctl stop ollama"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "658c147e-c7f8-490e-910e-62b80f577dda",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "!pip install aiohttp pyngrok\n",
+    "\n",
+    "import os\n",
+    "import asyncio\n",
+    "from aiohttp import ClientSession\n",
+    "\n",
+    "# Set LD_LIBRARY_PATH so the system NVIDIA library becomes preferred\n",
+    "# over the built-in library. This is particularly important for \n",
+    "# Google Colab which installs older drivers\n",
+    "os.environ.update({'LD_LIBRARY_PATH': '/usr/lib64-nvidia'})\n",
+    "\n",
+    "async def run(cmd):\n",
+    "  '''\n",
+    "  run is a helper function to run subcommands asynchronously.\n",
+    "  '''\n",
+    "  print('>>> starting', *cmd)\n",
+    "  p = await asyncio.subprocess.create_subprocess_exec(\n",
+    "      *cmd,\n",
+    "      stdout=asyncio.subprocess.PIPE,\n",
+    "      stderr=asyncio.subprocess.PIPE,\n",
+    "  )\n",
+    "\n",
+    "  async def pipe(lines):\n",
+    "    async for line in lines:\n",
+    "      print(line.strip().decode('utf-8'))\n",
+    "\n",
+    "  await asyncio.gather(\n",
+    "      pipe(p.stdout),\n",
+    "      pipe(p.stderr),\n",
+    "  )\n",
+    "\n",
+    "\n",
+    "await asyncio.gather(\n",
+    "    run(['ollama', 'serve']),\n",
+    "    run(['ngrok', 'http', '--log', 'stderr', '11434']),\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "e7735a55-9aad-4caf-8683-52e2163ba53b",
+   "metadata": {},
+   "source": [
+    "The previous cell starts two processes, `ollama` and `ngrok`. The log output will show a line like the following which describes the external address.\n",
+    "\n",
+    "```\n",
+    "t=2023-11-12T22:55:56+0000 lvl=info msg=\"started tunnel\" obj=tunnels name=command_line addr=http://localhost:11434 url=https://8249-34-125-179-11.ngrok.io\n",
+    "```\n",
+    "\n",
+    "The external address in this case is `https://8249-34-125-179-11.ngrok.io` which can be passed into `OLLAMA_HOST` to access this instance.\n",
+    "\n",
+    "```bash\n",
+    "export OLLAMA_HOST=https://8249-34-125-179-11.ngrok.io\n",
+    "ollama list\n",
+    "ollama run mistral\n",
+    "```"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.11.6"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

examples/kubernetes/README.md

@@ -0,0 +1,36 @@
+# Deploy Ollama to Kubernetes
+
+## Prerequisites
+
+- Ollama: https://ollama.ai/download
+- Kubernetes cluster. This example will use Google Kubernetes Engine.
+
+## Steps
+
+1. Create the Ollama namespace, daemon set, and service
+
+    ```bash
+    kubectl apply -f cpu.yaml
+    ```
+
+1. Port forward the Ollama service to connect and use it locally
+
+    ```bash
+    kubectl -n ollama port-forward service/ollama 11434:80
+    ```
+
+1. Pull and run a model, for example `orca-mini:3b`
+
+    ```bash
+    ollama run orca-mini:3b
+    ```
+
+## (Optional) Hardware Acceleration
+
+Hardware acceleration in Kubernetes requires NVIDIA's [`k8s-device-plugin`](https://github.com/NVIDIA/k8s-device-plugin). Follow the link for more details.
+
+Once configured, create a GPU enabled Ollama deployment.
+
+```bash
+kubectl apply -f gpu.yaml
+```

examples/kubernetes/cpu.yaml

@@ -0,0 +1,42 @@
+---
+apiVersion: v1
+kind: Namespace
+metadata:
+  name: ollama
+---
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: ollama
+  namespace: ollama
+spec:
+  selector:
+    matchLabels:
+      name: ollama
+  template:
+    metadata:
+      labels:
+        name: ollama
+    spec:
+      containers:
+      - name: ollama
+        image: ollama/ollama:latest
+        ports:
+        - name: http
+          containerPort: 11434
+          protocol: TCP
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: ollama
+  namespace: ollama
+spec:
+  type: ClusterIP
+  selector:
+    name: ollama
+  ports:
+  - port: 80
+    name: http
+    targetPort: http
+    protocol: TCP

examples/kubernetes/gpu.yaml

@@ -0,0 +1,58 @@
+---
+apiVersion: v1
+kind: Namespace
+metadata:
+  name: ollama
+---
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: ollama
+  namespace: ollama
+spec:
+  strategy:
+    type: Recreate
+  selector:
+    matchLabels:
+      name: ollama
+  template:
+    metadata:
+      labels:
+        name: ollama
+    spec:
+      containers:
+      - name: ollama
+        image: ollama/ollama:latest
+        env:
+        - name: PATH
+          value: /usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
+        - name: LD_LIBRARY_PATH
+          value: /usr/local/nvidia/lib:/usr/local/nvidia/lib64
+        - name: NVIDIA_DRIVER_CAPABILITIES
+          value: compute,utility
+        ports:
+        - name: http
+          containerPort: 11434
+          protocol: TCP
+        resources:
+          limits:
+            nvidia.com/gpu: 1
+      tolerations:
+      - key: nvidia.com/gpu
+        operator: Exists
+        effect: NoSchedule
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: ollama
+  namespace: ollama
+spec:
+  type: ClusterIP
+  selector:
+    name: ollama
+  ports:
+  - port: 80
+    name: http
+    targetPort: http
+    protocol: TCP

examples/langchain-python-rag-privategpt/constants.py

@@ -6,7 +6,6 @@ PERSIST_DIRECTORY = os.environ.get('PERSIST_DIRECTORY', 'db')
 
 # Define the Chroma settings
 CHROMA_SETTINGS = Settings(
-        chroma_db_impl='duckdb+parquet',
         persist_directory=PERSIST_DIRECTORY,
         anonymized_telemetry=False
 )

examples/langchain-python-rag-privategpt/ingest.py

@@ -150,7 +150,7 @@ def main():
         print("Creating new vectorstore")
         texts = process_documents()
         print(f"Creating embeddings. May take some minutes...")
-        db = Chroma.from_documents(texts, embeddings, persist_directory=persist_directory, client_settings=CHROMA_SETTINGS)
+        db = Chroma.from_documents(texts, embeddings, persist_directory=persist_directory)
     db.persist()
     db = None
 

examples/langchain-python-rag-privategpt/privateGPT.py

@@ -4,6 +4,7 @@ from langchain.embeddings import HuggingFaceEmbeddings
 from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler
 from langchain.vectorstores import Chroma
 from langchain.llms import Ollama
+import chromadb
 import os
 import argparse
 import time
@@ -22,7 +23,9 @@ def main():
     # Parse the command line arguments
     args = parse_arguments()
     embeddings = HuggingFaceEmbeddings(model_name=embeddings_model_name)
-    db = Chroma(persist_directory=persist_directory, embedding_function=embeddings, client_settings=CHROMA_SETTINGS)
+
+    db = Chroma(persist_directory=persist_directory, embedding_function=embeddings)
+
     retriever = db.as_retriever(search_kwargs={"k": target_source_chunks})
     # activate/deactivate the streaming StdOut callback for LLMs
     callbacks = [] if args.mute_stream else [StreamingStdOutCallbackHandler()]

examples/langchain-python-rag-websummary/README.md

@@ -1,15 +1,23 @@
 # LangChain Web Summarization
 
-This example summarizes a website
+This example summarizes the website, [https://ollama.ai/blog/run-llama2-uncensored-locally](https://ollama.ai/blog/run-llama2-uncensored-locally)
 
-## Setup
+## Running the Example
 
-```
-pip install -r requirements.txt
-```
+1. Ensure you have the `llama2` model installed:
 
-## Run
+   ```bash
+   ollama pull llama2
+   ```
 
-```
-python main.py
-```
+2. Install the Python Requirements.
+
+   ```bash
+   pip install -r requirements.txt
+   ```
+
+3. Run the example:
+
+   ```bash
+   python main.py
+   ```

examples/langchain-python-rag-websummary/requirements.txt

@@ -1,2 +1 @@
 langchain==0.0.259
-bs4==0.0.1

examples/langchain-python-simple/README.md

@@ -2,20 +2,23 @@
 
 This example is a basic "hello world" of using LangChain with Ollama.
 
-## Setup
+## Running the Example
 
-```
-pip install -r requirements.txt
-```
+1. Ensure you have the `llama2` model installed:
 
-## Run
+   ```bash
+   ollama pull llama2
+   ```
 
-```
-python main.py
-```
+2. Install the Python Requirements.
 
-Running this example will print the response for "hello":
+   ```bash
+   pip install -r requirements.txt
+   ```
 
-```
-Hello! It's nice to meet you. hopefully you are having a great day! Is there something I can help you with or would you like to chat?
-```
+3. Run the example:
+
+   ```bash
+   python main.py
+   ```
+  

examples/langchain-python-simple/main.py

@@ -1,4 +1,6 @@
 from langchain.llms import Ollama
+
+input = input("What is your question?")
 llm = Ollama(model="llama2")
-res = llm.predict("hello")
+res = llm.predict(input)
 print (res)

examples/langchain-typescript-simple/README.md

@@ -2,20 +2,22 @@
 
 This example is a basic "hello world" of using LangChain with Ollama using Node.js and Typescript.
 
-## Setup
+## Running the Example
 
-```shell
-npm install
-```
+1. Install the prerequisites:
 
-## Run
+   ```bash
+   npm install
+   ```
 
-```shell
-ts-node main.ts
-```
+2. Ensure the `mistral` model is available:
 
-Running this example will print the response for "hello":
+   ```bash
+   ollama pull mistral
+   ```
 
-```plaintext
-Hello! It's nice to meet you. hopefully you are having a great day! Is there something I can help you with or would you like to chat?
-```
+3. Run the example:
+
+   ```bash
+   npm start
+   ```

examples/langchain-typescript-simple/main.ts

@@ -1,15 +1,25 @@
-import { Ollama} from 'langchain/llms/ollama';
+import { Ollama } from 'langchain/llms/ollama';
+import * as readline from "readline";
 
 async function main() {
   const ollama = new Ollama({
     model: 'mistral'    
     // other parameters can be found at https://js.langchain.com/docs/api/llms_ollama/classes/Ollama
-  })
-  const stream = await ollama.stream("Hello");
+  });
 
-  for await (const chunk of stream) {
-    process.stdout.write(chunk);
-  }
+  const rl = readline.createInterface({
+    input: process.stdin,
+    output: process.stdout,
+  });
+
+  rl.question("What is your question: \n", async (user_input) => {
+    const stream = await ollama.stream(user_input);
+  
+    for await (const chunk of stream) {
+      process.stdout.write(chunk);
+    }
+    rl.close();
+  })
 }
 
 main();

examples/langchain-typescript-simple/package-lock.json

@@ -1,5 +1,5 @@
 {
-  "name": "with-langchain-typescript-simplegenerate",
+  "name": "langchain-typescript-simple",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {

examples/langchain-typescript-simple/package.json

@@ -1,8 +1,13 @@
 {
+  "scripts": {
+    "start": "tsx main.ts"
+  },
   "devDependencies": {
-    "typescript": "^5.2.2"
+    "tsx": "^4.6.2",
+    "typescript": "^5.3.3"
   },
   "dependencies": {
-    "langchain": "^0.0.165"
+    "langchain": "^0.0.165",
+    "readline": "^1.3.0"
   }
 }

examples/modelfile-10tweets/Modelfile

@@ -1,7 +0,0 @@
-# Modelfile for creating a list of ten tweets from a topic
-# Run `ollama create 10tweets -f ./Modelfile` and then `ollama run 10tweets` and enter a topic
-
-FROM llama2
-SYSTEM """
-You are a content marketer who needs to come up with 10 short but succinct tweets. The answer should be a list of ten tweets. Each tweet can have a maximum of 280 characters and should include hashtags. Each user input will be a subject and you should expand it in ten creative ways. Never stop after just one tweet. Always include ten. 
-"""

examples/modelfile-10tweets/README.md

@@ -1,23 +0,0 @@
-# Ten Tweets Modelfile
-
-This is a simple modelfile that generates ten tweets based off any topic.
-
-```bash
-ollama create tentweets
-
-ollama run tentweets
->>> underwater basketweaving
- Great! Here are ten creative tweets about underwater basketweaving:
-
-1. "Just discovered the ultimate stress-reliever: Underwater basketweaving! 🌊🧵 #UnderwaterBasketweaving #StressRelief"
-2. "Who needs meditation when you can do underwater basketweaving? 😴👀 #PeacefulDistraction #UnderwaterBasketweaving"
-3. "Just spent an hour in the pool and still managed to knot my basket. Goal: untangle it before next session. 💪🏽 #ChallengeAccepted #UnderwaterBasketweaving"
-4. "When life gives you lemons, make underwater basketweaving! 🍋🧵 #LemonadeLife #UnderwaterBasketweaving"
-5. "Just realized my underwater basketweaving skills could come in handy during a zombie apocalypse. 😂🧡 #SurvivalTips #UnderwaterBasketweaving"
-6. "I'm not lazy, I'm just conserving energy for my next underwater basketweaving session. 😴💤 #LazyDay #UnderwaterBasketweaving"
-7. "Just found my inner peace while doing underwater basketweaving. It's like meditation, but with knots! 🙏🧵 #Mindfulness #UnderwaterBasketweaving"
-8. "Why study for exams when you can do underwater basketweaving and forget all your worries? 😜🧵 #ProcrastinationStation #UnderwaterBasketweaving"
-9. "Just had to cut my underwater basketweaving session short due to a sudden urge to breathe. 🤯🌊 #AquaticAdventures #UnderwaterBasketweaving"
-10. "I'm not sure what's more impressive: my underwater basketweaving skills or the fact that I didn't drown trying to make this tweet. 😅🧵 #Accomplishment 
-#UnderwaterBasketweaving"
-```

examples/modelfile-devopsengineer/Modelfile

@@ -1,8 +0,0 @@
-# Modelfile for creating a devops engineer assistant
-# Run `ollama create devops-engineer -f ./Modelfile` and then `ollama run devops-engineer` and enter a topic
-
-FROM mistral
-PARAMETER temperature 1
-SYSTEM """
-You are a senior devops engineer, acting as an assistant. You offer help with cloud technologies like: Terraform, AWS, kubernetes, python. You answer with code examples when possible
-"""

examples/modelfile-devopsengineer/README.md

@@ -1,40 +0,0 @@
-# DevOps Engineer Modelfile
-
-This is a simple modelfile that takes will respond to any questions that a senior DevOps engineer should know how to solve.
-
-````text
-ollama create devopsengineer
-
-ollama run devopsengineer
->>> how can i proxy all connections from inside the office through a single server that caches large files. Provide a config file for nginx to achieve this?
-
-Here is an example configuration file for Nginx that can be used to proxy all connections from inside the office network to a centralized cache server:
-```perl
-http {
-    upstream cache_server {
-        server <cache_server_ip>;
-        server <cache_server_ip>;
-        server <cache_server_ip>;
-    }
-    
-    server {
-        listen 80;
-        server_name localhost;
-        
-        location / {
-            proxy_pass http://cache_server;
-            proxy_set_header Host $host;
-            proxy_set_header X-Real-IP $remote_addr;
-            proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
-        }
-    }
-}
-```
-In this configuration, we define an `upstream` block that lists the IP addresses of the cache servers. We then define a server block that listens on port 80 and routes all 
-traffic to the `/` location to the cache servers using the `proxy_pass` directive. The `proxy_set_header` directives are used to preserve the source IP address of the client
-request when forwarding it to the cache server.
-
-To use this configuration, you would need to replace the placeholder `<cache_server_ip>` with the actual IP addresses of your cache servers. You would also need to make sure
-that the cache servers are configured to accept incoming connections from the Nginx server and handle requests for files.
-
-````

examples/modelfile-midjourney/Modelfile

@@ -1,11 +0,0 @@
-# Modelfile for creating a Midjourney prompts from a topic
-# This prompt was adapted from the original at https://www.greataiprompts.com/guide/midjourney/best-chatgpt-prompt-for-midjourney/
-# Run `ollama create mj -f ./Modelfile` and then `ollama run mj` and enter a topic
-
-FROM zephyr
-PARAMETER temperature 0.8
-PARAMETER top_k 500
-PARAMETER top_p 0.9
-SYSTEM """
-Embrace your role as a creative illustrator. Based on a concept provided, you must produce a single paragraph with a multifaceted description of an image, ensuring significant details of the concept and more is represented in your instructions. You do not need to write complete sentences but rather short concepts with the following information: the level of detail that should be represented, an artistic style and maybe a specific name of a painter or illustrator, the ideal color pallete, lighting, mood, perspective, the setting, time of day, weather, the season, the time period, location, materials, the textures, patterns, lines, brushstrokes, techniques, the medium, the genre, the rendering style. Don't include everything and keep the description length under 250 words. 
-"""

examples/modelfile-midjourney/README.md

@@ -1,11 +0,0 @@
-# Midjourney Prompt Generator Modelfile
-
-This simple modelfile will help create a prompt to feed to Midjourney.
-
-```text
-ollama create midjourney
-
-ollama run midjourney
->>> a sports car in the mountains. 
-A sleek, high-performance automobile cuts through a serpentine mountain landscape. The concept is a classic illustration of speed and power, depicted in the style of pop art by Andy Warhol. The color palette is dominated by bold, primary hues of red, blue, and yellow, with striking accent colors of white, black, and metallic shades. The lighting is bright and focused, casting sharp shadows on the rugged terrain. A sense of excitement and anticipation permeates throughout the scene, as the car navigates a treacherous course through the winding road. The perspective is low, allowing for a full view of the vehicle's sleek lines and intricate details. The setting takes place in the afternoon during a sunny day in autumn, as evidenced by the vibrant foliage on the mountainside. The time period is modern, with nods to classic car design. The materials are primarily digital, allowing for smooth curves and sharp contrasts. The textures are sleek and polished, with meticulously detailed lines and brushstrokes that accentuate the car's aerodynamic design. The patterns consist of geometric shapes and bold stripes, adding to the car's dynamic appeal. The genre is modern realism, with a focus on precision and detail. The rendering style is highly technical, capturing the nuances and subtleties of the vehicle and its surroundings in breathtaking detail.
-```

examples/modelfile-recipemaker/Modelfile

@@ -1,6 +0,0 @@
-# Modelfile for creating a recipe from a list of ingredients
-# Run `ollama create recipemaker -f ./Modelfile` and then `ollama run recipemaker` and feed it lists of ingredients to create recipes around.
-FROM nous-hermes
-SYSTEM """
-The instruction will be a list of ingredients. You should generate a recipe that can be made in less than an hour. You can also include ingredients that most people will find in their pantry every day. The recipe should be 4 people and you should include a description of what the meal will taste like
-"""

examples/modelfile-recipemaker/README.md

@@ -1,20 +0,0 @@
-# Recipe Maker Modelfile 
-
-Simple modelfile to generate a recipe from a short list of ingredients.
-
-```
-ollama create recipemaker
-
-ollama run recipemaker
->>> chilli pepper, white chocolate, kale
- Ingredients:
-- 1 small chili pepper
-- 4 squares of white chocolate
-- handful of kale leaves
-
-Instructions:
-1. In a blender or food processor, puree the chilies and white chocolate until smooth.
-2. Add the chopped kale leaves to the blender and pulse until well combined.
-3. Serve immediately as a dip for crackers or use it as an ingredient in your favorite recipe. The mixture of spicy chili pepper with sweet white chocolate and nutritious 
-kale will make your taste buds dance with delight!
-```

examples/modelfile-sentiments/Modelfile

@@ -1,28 +0,0 @@
-# Modelfile for creating a sentiment analyzer. 
-# Run `ollama create sentiments -f pathtofile` and then `ollama run sentiments` and enter a topic
-
-FROM orca
-TEMPLATE """
-{{- if .First }}
-### System:
-{{ .System }}
-{{- end }}
-### User: 
-I hate it when my phone dies
-### Response: 
-NEGATIVE
-### User: 
-He is awesome
-### Response: 
-POSITIVE
-### User: 
-This is the link to the article
-### Response: 
-NEUTRAL
-### User:
-{{ .Prompt }}
-
-### Response:
-"""
-
-SYSTEM """You are a sentiment analyzer. You will receive text and output only one word, either POSITIVE or NEGATIVE or NEUTRAL, depending on the sentiment of the text."""

examples/modelfile-sentiments/Readme.md

@@ -1,25 +0,0 @@
-# Sentiments Modelfile
-
-This is a simple sentiments analyzer using the Orca model. When you pull Orca from the registry, it has a Template already defined that looks like this:
-
-```Modelfile
-{{- if .First }}
-### System:
-{{ .System }}
-{{- end }}
-
-### User:
-{{ .Prompt }}
-
-### Response:
-```
-
-If we just wanted to have the text:
-
-```Plaintext
-You are a sentiment analyzer. You will receive text and output only one word, either POSITIVE or NEGATIVE or NEUTRAL, depending on the sentiment of the text.
-```
-
-then we could have put this in a SYSTEM block. But we want to provide examples which require updating the full Template. Any Modelfile you create will inherit all the settings from the source model. But in this example, we are overriding the Template.
-
-When providing examples for the input and output, you should include the way the model usually provides information. Since the Orca model expects a user prompt to appear after ### User: and the response is after ### Response, we should format our examples like that as well. If we were using the Llama 2 model, the format would be a bit different.

examples/modelfile-tweetwriter/Modelfile

@@ -1,7 +0,0 @@
-# Modelfile for creating a tweet from a topic
-# Run `ollama create tweetwriter -f ./Modelfile` and then `ollama run tweetwriter` and enter a topic
-
-FROM nous-hermes
-SYSTEM """
-You are a content marketer who needs to come up with a short but succinct tweet. Make sure to include the appropriate hashtags and links. Sometimes when appropriate, describe a meme that can be included as well. All answers should be in the form of a tweet which has a max size of 280 characters. Every instruction will be the topic to create a tweet about.
-"""

examples/modelfile-tweetwriter/readme.md

@@ -0,0 +1,23 @@
+# Example Modelfile - Tweetwriter
+
+This simple examples shows what you can do without any code, simply relying on a Modelfile. The file has two instructions:
+
+1. FROM - The From instructions defines the parent model to use for this one. If you choose a model from the library, you can enter just the model name. For all other models, you need to specify the namespace as well. You could also use a local file. Just include the relative path to the converted, quantized model weights file. To learn more about creating that file, see the `import.md` file in the docs folder of this repository.
+2. SYSTEM - This defines the system prompt for the model and overrides the system prompt from the parent model.
+
+## Running the Example
+
+1. Create the model:
+
+   ```bash
+   ollama create tweetwriter
+   ```
+
+2. Enter a topic to generate a tweet about.
+3. Show the Modelfile in the REPL.
+
+   ```bash
+   /show modelfile
+   ```
+
+   Notice that the FROM and SYSTEM match what was in the file. But there is also a TEMPLATE and PARAMETER. These are inherited from the parent model.

examples/python-dockerit/README.md

@@ -1,15 +1,31 @@
 # DockerIt
 
-DockerIt is a tool to help you build and run your application in a Docker container. It consists of a model that defines the system prompt and model weights to use, along with a python script to then build the container and run the image automatically. 
+DockerIt is a tool to help you build and run your application in a Docker container. It consists of a model that defines the system prompt and model weights to use, along with a python script to then build the container and run the image automatically.
+
+## Running the Example
+
+1. Ensure you have the `mattw/dockerit` model installed:
+
+   ```bash
+   ollama pull mattw/dockerit
+   ```
+
+2. Make sure Docker is running on your machine.
+
+3. Install the Python Requirements.
+
+   ```bash
+   pip install -r requirements.txt
+   ```
+
+4. Run the example:
+
+   ```bash
+   python dockerit.py "simple postgres server with admin password set to 123"
+   ```
+
+5. Enter the name you would like to use for your container image.
 
 ## Caveats
 
-This is an simple example. It's assuming the Dockerfile content generated is going to work. In many cases, even with simple web servers, it fails when trying to copy files that don't exist. It's simply an example of what you could possibly do.
-
-## Example Usage
-
-```bash
-> python3 ./dockerit.py "simple postgres server with admin password set to 123"
-Enter the name of the image: matttest
-Container named happy_keller  started with id:  7c201bb6c30f02b356ddbc8e2a5af9d7d7d7b8c228519c9a501d15c0bd9d6b3e
-```
+This is a simple example. It's assuming the Dockerfile content generated is going to work. In many cases, even with simple web servers, it fails when trying to copy files that don't exist. It's simply an example of what you could possibly do.

examples/python-json-datagenerator/predefinedschema.py

@@ -0,0 +1,31 @@
+import requests
+import json
+import random
+
+model = "llama2"
+template = {
+  "firstName": "", 
+  "lastName": "", 
+  "address": {
+    "street": "", 
+    "city": "", 
+    "state": "", 
+    "zipCode": ""
+  }, 
+  "phoneNumber": ""
+}
+
+prompt = f"generate one realistically believable sample data set of a persons first name, last name, address in the US, and  phone number. \nUse the following template: {json.dumps(template)}."
+
+data = {
+    "prompt": prompt,
+    "model": model,
+    "format": "json",
+    "stream": False,
+    "options": {"temperature": 2.5, "top_p": 0.99, "top_k": 100},
+}
+
+print(f"Generating a sample user")
+response = requests.post("http://localhost:11434/api/generate", json=data, stream=False)
+json_data = json.loads(response.text)
+print(json.dumps(json.loads(json_data["response"]), indent=2))

examples/python-json-datagenerator/randomaddresses.py

@@ -0,0 +1,31 @@
+import requests
+import json
+import random
+
+countries = [
+    "United States",
+    "United Kingdom",
+    "the Netherlands",
+    "Germany",
+    "Mexico",
+    "Canada",
+    "France",
+]
+country = random.choice(countries)
+model = "llama2"
+
+prompt = f"generate one realistically believable sample data set of a persons first name, last name, address in {country}, and phone number. Do not use common names. Respond using JSON. Key names should have no backslashes, values should use plain ascii with no special characters."
+
+data = {
+    "prompt": prompt,
+    "model": model,
+    "format": "json",
+    "stream": False,
+    "options": {"temperature": 2.5, "top_p": 0.99, "top_k": 100},
+}
+
+print(f"Generating a sample user in {country}")
+response = requests.post("http://localhost:11434/api/generate", json=data, stream=False)
+json_data = json.loads(response.text)
+
+print(json.dumps(json.loads(json_data["response"]), indent=2))

examples/python-json-datagenerator/readme.md

@@ -0,0 +1,60 @@
+# JSON Output Example
+
+![llmjson 2023-11-10 15_31_31](https://github.com/jmorganca/ollama/assets/633681/e599d986-9b4a-4118-81a4-4cfe7e22da25)
+
+There are two python scripts in this example. `randomaddresses.py` generates random addresses from different countries. `predefinedschema.py` sets a template for the model to fill in.
+
+## Running the Example
+
+1. Ensure you have the `llama2` model installed:
+
+   ```bash
+   ollama pull llama2
+   ```
+
+2. Install the Python Requirements.
+
+   ```bash
+   pip install -r requirements.txt
+   ```
+
+3. Run the Random Addresses example:
+
+   ```bash
+   python randomaddresses.py
+   ```
+
+4. Run the Predefined Schema example:
+
+   ```bash
+   python predefinedschema.py
+   ```
+
+## Review the Code
+
+Both programs are basically the same, with a different prompt for each, demonstrating two different ideas. The key part of getting JSON out of a model is to state in the prompt or system prompt that it should respond using JSON, and specifying the `format` as `json` in the data body.
+
+```python
+prompt = f"generate one realistically believable sample data set of a persons first name, last name, address in {country}, and  phone number. Do not use common names. Respond using JSON. Key names should with no backslashes, values should use plain ascii with no special characters."
+
+data = {
+    "prompt": prompt,
+    "model": model,
+    "format": "json",
+    "stream": False,
+    "options": {"temperature": 2.5, "top_p": 0.99, "top_k": 100},
+}
+```
+
+When running `randomaddresses.py` you will see that the schema changes and adapts to the chosen country.
+
+In `predefinedschema.py`, a template has been specified in the prompt as well. It's been defined as JSON and then dumped into the prompt string to make it easier to work with.
+
+Both examples turn streaming off so that we end up with the completed JSON all at once. We need to convert the `response.text` to JSON so that when we output it as a string we can set the indent spacing to make the output easy to read.
+
+```python
+response = requests.post("http://localhost:11434/api/generate", json=data, stream=False)
+json_data = json.loads(response.text)
+
+print(json.dumps(json.loads(json_data["response"]), indent=2))
+```

examples/python-json-datagenerator/requirements.txt

@@ -0,0 +1 @@
+Requests==2.31.0

examples/python-loganalysis/Modelfile

@@ -0,0 +1,8 @@
+FROM codebooga:latest
+
+SYSTEM """
+You are a log file analyzer. You will receive a set of lines from a log file for some software application, find the errors and other interesting aspects of the logs, and explain them so a new user can understand what they mean. If there are any steps they can do to resolve them, list the steps in your answer.
+"""
+
+PARAMETER TEMPERATURE 0.3
+

examples/python-loganalysis/loganalysis.py

@@ -0,0 +1,41 @@
+import sys
+import re
+import requests
+import json
+
+# prelines and postlines represent the number of lines of context to include in the output around the error
+prelines = 10
+postlines = 10
+
+def find_errors_in_log_file():
+  if len(sys.argv) < 2:
+    print("Usage: python loganalysis.py <filename>")
+    return
+
+  log_file_path = sys.argv[1]
+  with open(log_file_path, 'r') as log_file:
+    log_lines = log_file.readlines()
+
+  error_logs = []
+  for i, line in enumerate(log_lines):
+      if "error" in line.lower():
+          start_index = max(0, i - prelines)
+          end_index = min(len(log_lines), i + postlines + 1)
+          error_logs.extend(log_lines[start_index:end_index])
+
+  return error_logs
+
+error_logs = find_errors_in_log_file()
+
+data = {
+  "prompt": "\n".join(error_logs), 
+  "model": "mattw/loganalyzer"
+}
+
+response = requests.post("http://localhost:11434/api/generate", json=data, stream=True)
+for line in response.iter_lines():
+  if line:
+    json_data = json.loads(line)
+    if json_data['done'] == False:
+      print(json_data['response'], end='', flush=True)
+

examples/python-loganalysis/logtest.logfile

@@ -0,0 +1,32 @@
+2023-11-10 07:17:40 /docker-entrypoint.sh: /docker-entrypoint.d/ is not empty, will attempt to perform configuration
+2023-11-10 07:17:40 /docker-entrypoint.sh: Looking for shell scripts in /docker-entrypoint.d/
+2023-11-10 07:17:40 /docker-entrypoint.sh: Launching /docker-entrypoint.d/10-listen-on-ipv6-by-default.sh
+2023-11-10 07:17:40 10-listen-on-ipv6-by-default.sh: info: Getting the checksum of /etc/nginx/conf.d/default.conf
+2023-11-10 07:17:40 10-listen-on-ipv6-by-default.sh: info: Enabled listen on IPv6 in /etc/nginx/conf.d/default.conf
+2023-11-10 07:17:40 /docker-entrypoint.sh: Sourcing /docker-entrypoint.d/15-local-resolvers.envsh
+2023-11-10 07:17:40 /docker-entrypoint.sh: Launching /docker-entrypoint.d/20-envsubst-on-templates.sh
+2023-11-10 07:17:40 /docker-entrypoint.sh: Launching /docker-entrypoint.d/30-tune-worker-processes.sh
+2023-11-10 07:17:40 /docker-entrypoint.sh: Configuration complete; ready for start up
+2023-11-10 07:17:40 2023/11/10 13:17:40 [notice] 1#1: using the "epoll" event method
+2023-11-10 07:17:40 2023/11/10 13:17:40 [notice] 1#1: nginx/1.25.3
+2023-11-10 07:17:40 2023/11/10 13:17:40 [notice] 1#1: built by gcc 12.2.0 (Debian 12.2.0-14) 
+2023-11-10 07:17:40 2023/11/10 13:17:40 [notice] 1#1: OS: Linux 6.4.16-linuxkit
+2023-11-10 07:17:40 2023/11/10 13:17:40 [notice] 1#1: getrlimit(RLIMIT_NOFILE): 1048576:1048576
+2023-11-10 07:17:40 2023/11/10 13:17:40 [notice] 1#1: start worker processes
+2023-11-10 07:17:40 2023/11/10 13:17:40 [notice] 1#1: start worker process 29
+2023-11-10 07:17:40 2023/11/10 13:17:40 [notice] 1#1: start worker process 30
+2023-11-10 07:17:40 2023/11/10 13:17:40 [notice] 1#1: start worker process 31
+2023-11-10 07:17:40 2023/11/10 13:17:40 [notice] 1#1: start worker process 32
+2023-11-10 07:17:40 2023/11/10 13:17:40 [notice] 1#1: start worker process 33
+2023-11-10 07:17:40 2023/11/10 13:17:40 [notice] 1#1: start worker process 34
+2023-11-10 07:17:40 2023/11/10 13:17:40 [notice] 1#1: start worker process 35
+2023-11-10 07:17:40 2023/11/10 13:17:40 [notice] 1#1: start worker process 36
+2023-11-10 07:17:40 2023/11/10 13:17:40 [notice] 1#1: start worker process 37
+2023-11-10 07:17:40 2023/11/10 13:17:40 [notice] 1#1: start worker process 38
+2023-11-10 07:17:44 192.168.65.1 - - [10/Nov/2023:13:17:43 +0000] "GET / HTTP/1.1" 200 615 "-" "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/119.0.0.0 Safari/537.36" "-"
+2023-11-10 07:17:44 2023/11/10 13:17:44 [error] 29#29: *1 open() "/usr/share/nginx/html/favicon.ico" failed (2: No such file or directory), client: 192.168.65.1, server: localhost, request: "GET /favicon.ico HTTP/1.1", host: "localhost:8080", referrer: "http://localhost:8080/"
+2023-11-10 07:17:44 192.168.65.1 - - [10/Nov/2023:13:17:44 +0000] "GET /favicon.ico HTTP/1.1" 404 555 "http://localhost:8080/" "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/119.0.0.0 Safari/537.36" "-"
+2023-11-10 07:17:50 2023/11/10 13:17:50 [error] 29#29: *1 open() "/usr/share/nginx/html/ahstat" failed (2: No such file or directory), client: 192.168.65.1, server: localhost, request: "GET /ahstat HTTP/1.1", host: "localhost:8080"
+2023-11-10 07:17:50 192.168.65.1 - - [10/Nov/2023:13:17:50 +0000] "GET /ahstat HTTP/1.1" 404 555 "-" "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/119.0.0.0 Safari/537.36" "-"
+2023-11-10 07:18:53 2023/11/10 13:18:53 [error] 29#29: *1 open() "/usr/share/nginx/html/ahstat" failed (2: No such file or directory), client: 192.168.65.1, server: localhost, request: "GET /ahstat HTTP/1.1", host: "localhost:8080"
+2023-11-10 07:18:53 192.168.65.1 - - [10/Nov/2023:13:18:53 +0000] "GET /ahstat HTTP/1.1" 404 555 "-" "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/119.0.0.0 Safari/537.36" "-"

examples/python-loganalysis/readme.md

@@ -0,0 +1,70 @@
+# Log Analysis example
+
+![loganalyzer 2023-11-10 08_53_29](https://github.com/jmorganca/ollama/assets/633681/ad30f1fc-321f-4953-8914-e30e24db9921)
+
+This example shows one possible way to create a log file analyzer. It uses the model **mattw/loganalyzer** which is based on **codebooga**, a 34b parameter model.
+
+To use it, run:
+
+`python loganalysis.py <logfile>`
+
+You can try this with the `logtest.logfile` file included in this directory.
+
+## Running the Example
+
+1. Ensure you have the `mattw/loganalyzer` model installed:
+
+   ```bash
+   ollama pull mattw/loganalyzer
+   ```
+
+2. Install the Python Requirements.
+
+   ```bash
+   pip install -r requirements.txt
+   ```
+
+3. Run the example:
+
+   ```bash
+   python loganalysis.py logtest.logfile
+   ```
+
+## Review the code
+
+The first part of this example is a Modelfile that takes `codebooga` and applies a new System Prompt:
+
+```plaintext
+SYSTEM """
+You are a log file analyzer. You will receive a set of lines from a log file for some software application, find the errors and other interesting aspects of the logs, and explain them so a new user can understand what they mean. If there are any steps they can do to resolve them, list the steps in your answer.
+"""
+```
+
+This model is available at https://ollama.ai/mattw/loganalyzer. You can customize it and add to your own namespace using the command `ollama create <namespace/modelname> -f <path-to-modelfile>` then `ollama push <namespace/modelname>`.
+
+Then loganalysis.py scans all the lines in the given log file and searches for the word 'error'. When the word is found, the 10 lines before and after are set as the prompt for a call to the Generate API.
+
+```python
+data = {
+  "prompt": "\n".join(error_logs), 
+  "model": "mattw/loganalyzer"
+}
+```
+
+Finally, the streamed output is parsed and the response field in the output is printed to the line.
+
+```python
+response = requests.post("http://localhost:11434/api/generate", json=data, stream=True)
+for line in response.iter_lines():
+  if line:
+    json_data = json.loads(line)
+    if json_data['done'] == False:
+      print(json_data['response'], end='')
+
+```
+
+## Next Steps
+
+There is a lot more that can be done here. This is a simple way to detect errors, looking for the word error. Perhaps it would be interesting to find anomalous activity in the logs. It could be interesting to create embeddings for each line and compare them, looking for similar lines. Or look into applying Levenshtein Distance algorithms to find similar lines to help identify the anomalous lines.
+
+Try different models and different prompts to analyze the data. You could consider adding retrieval augmented generation (RAG) to this to help understand newer log formats.

examples/python-loganalysis/requirements.txt

@@ -0,0 +1 @@
+Requests==2.31.0

examples/python-rag-newssummary/README.md

@@ -0,0 +1,35 @@
+# News Summarizer
+
+This example goes through a series of steps:
+
+  1. You choose a topic area (e.g., "news", "NVidia", "music", etc.).
+  2. Gets the most recent articles on that topic from various sources.
+  3. Uses Ollama to summarize each article.
+  4. Creates chunks of sentences from each article.
+  5. Uses Sentence Transformers to generate embeddings for each of those chunks.
+  6. You enter a question regarding the summaries shown.
+  7. Uses Sentence Transformers to generate an embedding for that question.
+  8. Uses the embedded question to find the most similar chunks.
+  9. Feeds all that to Ollama to generate a good answer to your question based on these news articles.
+
+This example lets you pick from a few different topic areas, then summarize the most recent x articles for that topic. It then creates chunks of sentences from each article and then generates embeddings for each of those chunks.
+
+## Running the Example
+
+1. Ensure you have the `mistral-openorca` model installed:
+
+   ```bash
+   ollama pull mistral-openorca
+   ```
+
+2. Install the Python Requirements.
+
+   ```bash
+   pip install -r requirements.txt
+   ```
+
+3. Run the example:
+
+   ```bash
+   python summ.py
+   ```

examples/python-rag-newssummary/requirements.txt

@@ -0,0 +1,9 @@
+beautifulsoup4==4.12.2
+feedparser==6.0.10
+mattsollamatools==0.0.8
+newspaper3k==0.2.8
+nltk==3.8.1
+numpy==1.24.3
+Requests==2.31.0
+scikit_learn==1.3.0
+sentence_transformers==2.2.2

examples/python-rag-newssummary/summ.py

@@ -0,0 +1,86 @@
+import curses
+import json
+from utils import get_url_for_topic, topic_urls, menu, getUrls, get_summary, getArticleText, knn_search
+import requests
+from sentence_transformers import SentenceTransformer
+from mattsollamatools import chunker
+
+if __name__ == "__main__":
+    chosen_topic = curses.wrapper(menu)
+    print("Here is your news summary:\n")
+    urls = getUrls(chosen_topic, n=5)
+    model = SentenceTransformer('all-MiniLM-L6-v2')
+    allEmbeddings = []
+
+    for url in urls:
+      article={}
+      article['embeddings'] = []
+      article['url'] = url
+      text = getArticleText(url)
+      summary = get_summary(text)
+      chunks = chunker(text)  # Use the chunk_text function from web_utils
+      embeddings = model.encode(chunks)
+      for (chunk, embedding) in zip(chunks, embeddings):
+        item = {}
+        item['source'] = chunk
+        item['embedding'] = embedding.tolist()  # Convert NumPy array to list
+        item['sourcelength'] = len(chunk)
+        article['embeddings'].append(item)
+    
+      allEmbeddings.append(article)
+
+      print(f"{summary}\n")
+
+    
+    while True:
+      context = []
+      # Input a question from the user
+      question = input("Enter your question about the news, or type quit: ")
+
+      if question.lower() == 'quit':
+        break
+
+      # Embed the user's question
+      question_embedding = model.encode([question])
+
+      # Perform KNN search to find the best matches (indices and source text)
+      best_matches = knn_search(question_embedding, allEmbeddings, k=10)
+
+
+      sourcetext=""
+      for i, (index, source_text) in enumerate(best_matches, start=1):
+          sourcetext += f"{i}. Index: {index}, Source Text: {source_text}"
+
+      systemPrompt = f"Only use the following information to answer the question. Do not use anything else: {sourcetext}"
+
+      url = "http://localhost:11434/api/generate"
+
+      payload = {
+      "model": "mistral-openorca",
+      "prompt": question, 
+      "system": systemPrompt,
+      "stream": False, 
+      "context": context
+      }
+
+      # Convert the payload to a JSON string
+      payload_json = json.dumps(payload)
+
+      # Set the headers to specify JSON content
+      headers = {
+          "Content-Type": "application/json"
+      }
+
+      # Send the POST request
+      response = requests.post(url, data=payload_json, headers=headers)
+
+      # Check the response
+      if response.status_code == 200:
+          output = json.loads(response.text)
+          context = output['context']
+          print(output['response']+ "\n")
+          
+
+      else:
+          print(f"Request failed with status code {response.status_code}")
+

examples/python-rag-newssummary/utils.py

@@ -0,0 +1,108 @@
+import curses
+import feedparser
+import requests
+import unicodedata
+import json
+from newspaper import Article
+from bs4 import BeautifulSoup
+from nltk.tokenize import sent_tokenize, word_tokenize
+import numpy as np
+from sklearn.neighbors import NearestNeighbors
+from mattsollamatools import chunker
+
+# Create a dictionary to store topics and their URLs
+topic_urls = {
+    "Mac": "https://9to5mac.com/guides/mac/feed",
+    "News": "http://www.npr.org/rss/rss.php?id=1001",
+    "Nvidia": "https://nvidianews.nvidia.com/releases.xml",
+    "Raspberry Pi": "https://www.raspberrypi.com/news/feed/", 
+    "Music": "https://www.billboard.com/c/music/music-news/feed/"
+}
+
+# Use curses to create a menu of topics
+def menu(stdscr):
+    chosen_topic = get_url_for_topic(stdscr)  
+    url = topic_urls[chosen_topic] if chosen_topic in topic_urls else "Topic not found"
+    
+    stdscr.addstr(len(topic_urls) + 3, 0, f"Selected URL for {chosen_topic}: {url}")
+    stdscr.refresh()
+    
+    return chosen_topic
+
+# You have chosen a topic. Now return the url for that topic
+def get_url_for_topic(stdscr):
+    curses.curs_set(0)  # Hide the cursor
+    stdscr.clear()
+
+    stdscr.addstr(0, 0, "Choose a topic using the arrow keys (Press Enter to select):")
+
+    # Create a list of topics
+    topics = list(topic_urls.keys())
+    current_topic = 0
+
+    while True:
+        for i, topic in enumerate(topics):
+            if i == current_topic:
+                stdscr.addstr(i + 2, 2, f"> {topic}")
+            else:
+                stdscr.addstr(i + 2, 2, f"  {topic}")
+
+        stdscr.refresh()
+
+        key = stdscr.getch()
+
+        if key == curses.KEY_DOWN and current_topic < len(topics) - 1:
+            current_topic += 1
+        elif key == curses.KEY_UP and current_topic > 0:
+            current_topic -= 1
+        elif key == 10:  # Enter key
+            return topic_urls[topics[current_topic]]
+
+# Get the last N URLs from an RSS feed
+def getUrls(feed_url, n=20):
+    feed = feedparser.parse(feed_url)
+    entries = feed.entries[-n:]
+    urls = [entry.link for entry in entries]
+    return urls
+
+# Often there are a bunch of ads and menus on pages for a news article. This uses newspaper3k to get just the text of just the article.
+def getArticleText(url):
+  article = Article(url)
+  article.download()
+  article.parse()
+  return article.text
+
+def get_summary(text):
+  systemPrompt = "Write a concise summary of the text, return your responses with 5 lines that cover the key points of the text given."
+  prompt = text
+  
+  url = "http://localhost:11434/api/generate"
+
+  payload = {
+    "model": "mistral-openorca",
+    "prompt": prompt, 
+    "system": systemPrompt,
+    "stream": False
+  }
+  payload_json = json.dumps(payload)
+  headers = {"Content-Type": "application/json"}
+  response = requests.post(url, data=payload_json, headers=headers)
+
+  return json.loads(response.text)["response"]
+
+# Perform K-nearest neighbors (KNN) search
+def knn_search(question_embedding, embeddings, k=5):
+    X = np.array([item['embedding'] for article in embeddings for item in article['embeddings']])
+    source_texts = [item['source'] for article in embeddings for item in article['embeddings']]
+    
+    # Fit a KNN model on the embeddings
+    knn = NearestNeighbors(n_neighbors=k, metric='cosine')
+    knn.fit(X)
+    
+    # Find the indices and distances of the k-nearest neighbors
+    distances, indices = knn.kneighbors(question_embedding, n_neighbors=k)
+    
+    # Get the indices and source texts of the best matches
+    best_matches = [(indices[0][i], source_texts[indices[0][i]]) for i in range(k)]
+    
+    return best_matches

examples/python-simplechat/client.py

@@ -0,0 +1,47 @@
+import json
+import requests
+
+# NOTE: ollama must be running for this to work, start the ollama app or run `ollama serve`
+model = "llama2"  # TODO: update this for whatever model you wish to use
+
+
+def chat(messages):
+    r = requests.post(
+        "http://0.0.0.0:11434/api/chat",
+        json={"model": model, "messages": messages, "stream": True},
+    )
+    r.raise_for_status()
+    output = ""
+
+    for line in r.iter_lines():
+        body = json.loads(line)
+        if "error" in body:
+            raise Exception(body["error"])
+        if body.get("done") is False:
+            message = body.get("message", "")
+            content = message.get("content", "")
+            output += content
+            # the response streams one token at a time, print that as we receive it
+            print(content, end="", flush=True)
+
+        if body.get("done", False):
+            message["content"] = output
+            return message
+
+
+def main():
+    messages = []
+
+    while True:
+        user_input = input("Enter a prompt: ")
+        if not user_input:
+            exit()
+        print()
+        messages.append({"role": "user", "content": user_input})
+        message = chat(messages)
+        messages.append(message)
+        print("\n\n")
+
+
+if __name__ == "__main__":
+    main()

examples/python-simplechat/readme.md

@@ -0,0 +1,44 @@
+# Simple Chat Example
+
+The **chat** endpoint is one of two ways to generate text from an LLM with Ollama, and is introduced in version 0.1.14. At a high level, you provide the endpoint an array of objects with a role and content specified. Then with each output and prompt, you add more of those role/content objects, which builds up the history.
+
+## Running the Example
+
+1. Ensure you have the `llama2` model installed:
+
+   ```bash
+   ollama pull llama2
+   ```
+
+2. Install the Python Requirements.
+
+   ```bash
+   pip install -r requirements.txt
+   ```
+
+3. Run the example:
+
+   ```bash
+   python client.py
+   ```
+
+## Review the Code
+
+You can see in the **chat** function that actually calling the endpoint is done simply with:
+
+```python
+r = requests.post(
+  "http://0.0.0.0:11434/api/chat",
+  json={"model": model, "messages": messages, "stream": True},
+)
+```
+
+With the **generate** endpoint, you need to provide a `prompt`. But with **chat**, you provide `messages`. And the resulting stream of responses includes a `message` object with a `content` field.
+
+The final JSON object doesn't provide the full content, so you will need to build the content yourself.
+
+In the **main** function, we collect `user_input` and add it as a message to our messages and that is passed to the chat function. When the LLM is done responding the output is added as another message.
+
+## Next Steps
+
+In this example, all generations are kept. You might want to experiment with summarizing everything older than 10 conversations to enable longer history with less context being used.

examples/python-simplechat/requirements.txt

@@ -0,0 +1 @@
+Requests==2.31.0

examples/python-simplegenerate/README.md

@@ -0,0 +1,29 @@
+# Simple Generate Example
+
+This is a simple example using the **Generate** endpoint.
+
+## Running the Example
+
+1. Ensure you have the `stablelm-zephyr` model installed:
+
+   ```bash
+   ollama pull stablelm-zephyr
+   ```
+
+2. Install the Python Requirements.
+
+   ```bash
+   pip install -r requirements.txt
+   ```
+
+3. Run the example:
+
+   ```bash
+   python client.py
+   ```
+
+## Review the Code
+
+The **main** function simply asks for input, then passes that to the generate function. The output from generate is then passed back to generate on the next run.
+
+The **generate** function uses `requests.post` to call `/api/generate`, passing the model, prompt, and context. The `generate` endpoint returns a stream of JSON blobs that are then iterated through, looking for the response values. That is then printed out. The final JSON object includes the full context of the conversation so far, and that is the return value from the function.

examples/python-simplegenerate/client.py

@@ -2,7 +2,7 @@ import json
 import requests
 
 # NOTE: ollama must be running for this to work, start the ollama app or run `ollama serve`
-model = 'llama2' # TODO: update this for whatever model you wish to use
+model = 'stablelm-zephyr' # TODO: update this for whatever model you wish to use
 
 def generate(prompt, context):
     r = requests.post('http://localhost:11434/api/generate',
@@ -17,7 +17,7 @@ def generate(prompt, context):
     for line in r.iter_lines():
         body = json.loads(line)
         response_part = body.get('response', '')
-        # the response streams one token at a time, print that as we recieve it
+        # the response streams one token at a time, print that as we receive it
         print(response_part, end='', flush=True)
 
         if 'error' in body:
@@ -30,9 +30,11 @@ def main():
     context = [] # the context stores a conversation history, you can use this to make the model more context aware
     while True:
         user_input = input("Enter a prompt: ")
+        if not user_input:
+            exit()
         print()
         context = generate(user_input, context)
         print()
 
 if __name__ == "__main__":
-    main()
+    main()

examples/python-simplegenerate/requirements.txt

@@ -0,0 +1 @@
+Requests==2.31.0

examples/typescript-functioncalling/extractemail.ts

@@ -0,0 +1,118 @@
+import { Ollama } from "ollama-node";
+import { readFile } from "fs/promises";
+
+// function to be called on events
+function reportEvents(name: string, date: string, location: string) {
+  const nameString = name ? `${name}` : `an event`;
+  const dateString = date ? ` on ${date}` : ``;
+  const locationString = location ? ` at ${location}` : ``;
+  console.log(`You have an event: ${nameString}${dateString}${locationString}`)
+}
+
+// function to be called on addresses
+function reportAddresses(address) {
+  for (const field in address) {
+    if (address[field]) {
+      if (field === "city") {
+        const city = address.city;
+        const state = address.state ? `, ${address.state}` : '';
+        const zip = address.zip ? ` ${address.zip}` : '';
+        console.log(`${city}${state}${zip}`);
+        break;
+      } else {
+        console.log(`${address[field]}`);
+      }
+    }
+  }
+  console.log(``);
+}
+
+async function main() {
+
+  const ollama = new Ollama();
+
+  const systemprompt = `You will be given a text along with a prompt and a schema. You will have to extract the information requested in the prompt from the text and generate output in JSON observing the schema provided. If the schema shows a type of integer or number, you must only show a integer for that field. A string should always be a valid string. If a value is unknown, leave it empty. Output the JSON with extra spaces to ensure that it pretty prints.`
+
+  const schema = {
+    "eventsQuantity": {
+      "type": "integer",
+      "description": "The number of events in the source text"
+    },
+    "addressesQuantity": {
+      "type": "integer",
+      "description": "The number of addresses in the source text"
+    },
+    "events": [{
+      name: {
+        "type": "string",
+        description: "Name of the event"
+      },
+      "date": {
+        "type": "string",
+        "description": "Date of the event"
+      },
+      "location": {
+        "type": "string",
+        "description": "Location of the event"
+      },
+      "extraInfo": {
+        "type": "string",
+        "description": "Any extra information that is provided about the event."
+      }
+    }],
+    "people": [{
+      "name": {
+        "type": "string",
+        "description": "Name of the person"
+      },
+      "company": {
+        "type": "string",
+        "description": "Name of the company where they work"
+      },
+      "street": {
+        "type": "string",
+        "description": "Street address of the person or company. This is only the street name and the numerical address. Do not include city, state, or zip of the address in this field."
+      },
+      "city": {
+        "type": "string",
+        "description": "City portion of the address of the person or company"
+      },
+      "state": {
+        "type": "string",
+        "description": "State portion of the address of the person or company"
+      },
+      "zip": {
+        "type": "string",
+        "description": "Zip code of the person or company"
+      },
+      "extraInfo": {
+        "type": "string",
+        "description": "Any extra information that is provided about the location."
+      }
+    }]
+  }
+
+  const textcontent = await readFile("./info.txt", "utf-8").then((text) => text.split(" ").slice(0, 2000).join(" "));
+
+  const prompt = `The source text is a series of emails that have been put into a single file. They are separated by three dashes. Review the source text and determine the full address of the person sending each of the emails as well as any events that we need to track. If they provide a company address use that. If any extra info is provided, such as a description of the place, or a floor, add it to extraInfo. The first field in the address JSON is quantity of events and should be set to the number of events tracked and the second field should be set to the number of addresses tracked in the file. Don't stuff an event into the output that isn't an event. Only add data to the mostly appropriate field. Don't make up fields that aren't in the schema. If there isn't a value for a field, use null. Output should be in JSON.\n\nSchema: \n${JSON.stringify(schema, null, 2)}\n\nSource Text:\n${textcontent}`
+
+  await ollama.setModel("neural-chat");
+  ollama.setSystemPrompt(systemprompt);
+  ollama.setJSONFormat(true);
+  const data = await ollama.generate(prompt);
+  const output = JSON.parse(data.output);
+  const events = output.events;
+  const addresses = output.people;
+
+  console.log(`Here are your ${output.eventsQuantity} events:`);
+  for (const event of events) {
+    reportEvents(event.name, event.date, event.location);
+  }
+
+  console.log(`\n\nHere are your ${output.addressesQuantity} addresses:`);
+  for (const address of addresses) {
+    reportAddresses(address);
+  }
+}
+
+main();

examples/typescript-functioncalling/extractwp.ts

@@ -0,0 +1,38 @@
+import { Ollama } from "ollama-node";
+import { readFile } from "fs/promises";
+
+async function main() {
+
+  const ollama = new Ollama();
+
+  // Set the system prompt to prepare the model to receive a prompt and a schema and set some rules for the output.
+  const systemprompt = `You will be given a text along with a prompt and a schema. You will have to extract the information requested in the prompt from the text and generate output in JSON observing the schema provided. If the schema shows a type of integer or number, you must only show a integer for that field. A string should always be a valid string. If a value is unknown, leave it empty. Output the JSON with extra spaces to ensure that it pretty prints.`
+
+  const schema = {
+    "people": [{
+      "name": {
+        "type": "string",
+        "description": "Name of the person"
+      },
+      "title": {
+        "type": "string",
+        "description": "Title of the person"
+      }
+    }],
+  }
+
+  // Depending on the model chosen, you may be limited by the size of the context window, so limit the context to 2000 words.
+  const textcontent = await readFile("./wp.txt", "utf-8").then((text) => text.split(" ").slice(0, 2000).join(" "));
+
+  // Specific instructions for this task
+  const prompt = `Review the source text and determine the 10 most important people to focus on. Then extract the name and title for those people. Output should be in JSON.\n\nSchema: \n${JSON.stringify(schema, null, 2)}\n\nSource Text:\n${textcontent}`
+
+  await ollama.setModel("neural-chat");
+  ollama.setSystemPrompt(systemprompt);
+
+  // setJSONFormat is the equivalent of setting 'format: json' in the API
+  ollama.setJSONFormat(true);
+  await ollama.streamingGenerate(prompt, (word) => { process.stdout.write(word) })
+}
+
+main();

examples/typescript-functioncalling/info.txt

@@ -0,0 +1,17 @@
+---
+Hi matt, 
+
+thanks for letting me know that you are going to come today, November 16, for my tea party. My address is 123 Falk St on Bainbridge Island. I live in the house with the red door. I will be home all day so just come by whenever you want.
+
+Fred
+
+---
+Great, send the check to our office at 1917 1st St, Seattle, WA 98101. I will let you know when we receive it.
+
+Mark Richardson
+Big Corp
+---
+We are looking forward to seeing you at our Local AI Meetup. It will be held on December 3. It will be at the offices of Enormous Co. Our address is 344 1st Ave, Seattle, WA 98101. We will be meeting in the conference room on the 3rd floor.
+
+Barbara Reilly
+Enormous Co.

examples/typescript-functioncalling/package-lock.json

@@ -0,0 +1,519 @@
+{
+  "name": "typescript-functioncalling",
+  "lockfileVersion": 3,
+  "requires": true,
+  "packages": {
+    "": {
+      "dependencies": {
+        "ollama-node": "^0.1.27"
+      },
+      "devDependencies": {
+        "tsx": "^4.1.2",
+        "typescript": "^5.2.2"
+      }
+    },
+    "node_modules/@esbuild/android-arm": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/android-arm/-/android-arm-0.18.20.tgz",
+      "integrity": "sha512-fyi7TDI/ijKKNZTUJAQqiG5T7YjJXgnzkURqmGj13C6dCqckZBLdl4h7bkhHt/t0WP+zO9/zwroDvANaOqO5Sw==",
+      "cpu": [
+        "arm"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "android"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/android-arm64": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/android-arm64/-/android-arm64-0.18.20.tgz",
+      "integrity": "sha512-Nz4rJcchGDtENV0eMKUNa6L12zz2zBDXuhj/Vjh18zGqB44Bi7MBMSXjgunJgjRhCmKOjnPuZp4Mb6OKqtMHLQ==",
+      "cpu": [
+        "arm64"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "android"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/android-x64": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/android-x64/-/android-x64-0.18.20.tgz",
+      "integrity": "sha512-8GDdlePJA8D6zlZYJV/jnrRAi6rOiNaCC/JclcXpB+KIuvfBN4owLtgzY2bsxnx666XjJx2kDPUmnTtR8qKQUg==",
+      "cpu": [
+        "x64"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "android"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/darwin-arm64": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/darwin-arm64/-/darwin-arm64-0.18.20.tgz",
+      "integrity": "sha512-bxRHW5kHU38zS2lPTPOyuyTm+S+eobPUnTNkdJEfAddYgEcll4xkT8DB9d2008DtTbl7uJag2HuE5NZAZgnNEA==",
+      "cpu": [
+        "arm64"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "darwin"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/darwin-x64": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/darwin-x64/-/darwin-x64-0.18.20.tgz",
+      "integrity": "sha512-pc5gxlMDxzm513qPGbCbDukOdsGtKhfxD1zJKXjCCcU7ju50O7MeAZ8c4krSJcOIJGFR+qx21yMMVYwiQvyTyQ==",
+      "cpu": [
+        "x64"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "darwin"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/freebsd-arm64": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/freebsd-arm64/-/freebsd-arm64-0.18.20.tgz",
+      "integrity": "sha512-yqDQHy4QHevpMAaxhhIwYPMv1NECwOvIpGCZkECn8w2WFHXjEwrBn3CeNIYsibZ/iZEUemj++M26W3cNR5h+Tw==",
+      "cpu": [
+        "arm64"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "freebsd"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/freebsd-x64": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/freebsd-x64/-/freebsd-x64-0.18.20.tgz",
+      "integrity": "sha512-tgWRPPuQsd3RmBZwarGVHZQvtzfEBOreNuxEMKFcd5DaDn2PbBxfwLcj4+aenoh7ctXcbXmOQIn8HI6mCSw5MQ==",
+      "cpu": [
+        "x64"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "freebsd"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/linux-arm": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/linux-arm/-/linux-arm-0.18.20.tgz",
+      "integrity": "sha512-/5bHkMWnq1EgKr1V+Ybz3s1hWXok7mDFUMQ4cG10AfW3wL02PSZi5kFpYKrptDsgb2WAJIvRcDm+qIvXf/apvg==",
+      "cpu": [
+        "arm"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "linux"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/linux-arm64": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/linux-arm64/-/linux-arm64-0.18.20.tgz",
+      "integrity": "sha512-2YbscF+UL7SQAVIpnWvYwM+3LskyDmPhe31pE7/aoTMFKKzIc9lLbyGUpmmb8a8AixOL61sQ/mFh3jEjHYFvdA==",
+      "cpu": [
+        "arm64"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "linux"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/linux-ia32": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/linux-ia32/-/linux-ia32-0.18.20.tgz",
+      "integrity": "sha512-P4etWwq6IsReT0E1KHU40bOnzMHoH73aXp96Fs8TIT6z9Hu8G6+0SHSw9i2isWrD2nbx2qo5yUqACgdfVGx7TA==",
+      "cpu": [
+        "ia32"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "linux"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/linux-loong64": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/linux-loong64/-/linux-loong64-0.18.20.tgz",
+      "integrity": "sha512-nXW8nqBTrOpDLPgPY9uV+/1DjxoQ7DoB2N8eocyq8I9XuqJ7BiAMDMf9n1xZM9TgW0J8zrquIb/A7s3BJv7rjg==",
+      "cpu": [
+        "loong64"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "linux"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/linux-mips64el": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/linux-mips64el/-/linux-mips64el-0.18.20.tgz",
+      "integrity": "sha512-d5NeaXZcHp8PzYy5VnXV3VSd2D328Zb+9dEq5HE6bw6+N86JVPExrA6O68OPwobntbNJ0pzCpUFZTo3w0GyetQ==",
+      "cpu": [
+        "mips64el"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "linux"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/linux-ppc64": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/linux-ppc64/-/linux-ppc64-0.18.20.tgz",
+      "integrity": "sha512-WHPyeScRNcmANnLQkq6AfyXRFr5D6N2sKgkFo2FqguP44Nw2eyDlbTdZwd9GYk98DZG9QItIiTlFLHJHjxP3FA==",
+      "cpu": [
+        "ppc64"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "linux"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/linux-riscv64": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/linux-riscv64/-/linux-riscv64-0.18.20.tgz",
+      "integrity": "sha512-WSxo6h5ecI5XH34KC7w5veNnKkju3zBRLEQNY7mv5mtBmrP/MjNBCAlsM2u5hDBlS3NGcTQpoBvRzqBcRtpq1A==",
+      "cpu": [
+        "riscv64"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "linux"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/linux-s390x": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/linux-s390x/-/linux-s390x-0.18.20.tgz",
+      "integrity": "sha512-+8231GMs3mAEth6Ja1iK0a1sQ3ohfcpzpRLH8uuc5/KVDFneH6jtAJLFGafpzpMRO6DzJ6AvXKze9LfFMrIHVQ==",
+      "cpu": [
+        "s390x"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "linux"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/linux-x64": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/linux-x64/-/linux-x64-0.18.20.tgz",
+      "integrity": "sha512-UYqiqemphJcNsFEskc73jQ7B9jgwjWrSayxawS6UVFZGWrAAtkzjxSqnoclCXxWtfwLdzU+vTpcNYhpn43uP1w==",
+      "cpu": [
+        "x64"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "linux"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/netbsd-x64": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/netbsd-x64/-/netbsd-x64-0.18.20.tgz",
+      "integrity": "sha512-iO1c++VP6xUBUmltHZoMtCUdPlnPGdBom6IrO4gyKPFFVBKioIImVooR5I83nTew5UOYrk3gIJhbZh8X44y06A==",
+      "cpu": [
+        "x64"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "netbsd"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/openbsd-x64": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/openbsd-x64/-/openbsd-x64-0.18.20.tgz",
+      "integrity": "sha512-e5e4YSsuQfX4cxcygw/UCPIEP6wbIL+se3sxPdCiMbFLBWu0eiZOJ7WoD+ptCLrmjZBK1Wk7I6D/I3NglUGOxg==",
+      "cpu": [
+        "x64"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "openbsd"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/sunos-x64": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/sunos-x64/-/sunos-x64-0.18.20.tgz",
+      "integrity": "sha512-kDbFRFp0YpTQVVrqUd5FTYmWo45zGaXe0X8E1G/LKFC0v8x0vWrhOWSLITcCn63lmZIxfOMXtCfti/RxN/0wnQ==",
+      "cpu": [
+        "x64"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "sunos"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/win32-arm64": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/win32-arm64/-/win32-arm64-0.18.20.tgz",
+      "integrity": "sha512-ddYFR6ItYgoaq4v4JmQQaAI5s7npztfV4Ag6NrhiaW0RrnOXqBkgwZLofVTlq1daVTQNhtI5oieTvkRPfZrePg==",
+      "cpu": [
+        "arm64"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "win32"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/win32-ia32": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/win32-ia32/-/win32-ia32-0.18.20.tgz",
+      "integrity": "sha512-Wv7QBi3ID/rROT08SABTS7eV4hX26sVduqDOTe1MvGMjNd3EjOz4b7zeexIR62GTIEKrfJXKL9LFxTYgkyeu7g==",
+      "cpu": [
+        "ia32"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "win32"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@esbuild/win32-x64": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/@esbuild/win32-x64/-/win32-x64-0.18.20.tgz",
+      "integrity": "sha512-kTdfRcSiDfQca/y9QIkng02avJ+NCaQvrMejlsB3RRv5sE9rRoeBPISaZpKxHELzRxZyLvNts1P27W3wV+8geQ==",
+      "cpu": [
+        "x64"
+      ],
+      "dev": true,
+      "optional": true,
+      "os": [
+        "win32"
+      ],
+      "engines": {
+        "node": ">=12"
+      }
+    },
+    "node_modules/@types/node": {
+      "version": "20.9.0",
+      "resolved": "https://registry.npmjs.org/@types/node/-/node-20.9.0.tgz",
+      "integrity": "sha512-nekiGu2NDb1BcVofVcEKMIwzlx4NjHlcjhoxxKBNLtz15Y1z7MYf549DFvkHSId02Ax6kGwWntIBPC3l/JZcmw==",
+      "dependencies": {
+        "undici-types": "~5.26.4"
+      }
+    },
+    "node_modules/buffer-from": {
+      "version": "1.1.2",
+      "resolved": "https://registry.npmjs.org/buffer-from/-/buffer-from-1.1.2.tgz",
+      "integrity": "sha512-E+XQCRwSbaaiChtv6k6Dwgc+bx+Bs6vuKJHHl5kox/BaKbhiXzqQOwK4cO22yElGp2OCmjwVhT3HmxgyPGnJfQ==",
+      "dev": true
+    },
+    "node_modules/esbuild": {
+      "version": "0.18.20",
+      "resolved": "https://registry.npmjs.org/esbuild/-/esbuild-0.18.20.tgz",
+      "integrity": "sha512-ceqxoedUrcayh7Y7ZX6NdbbDzGROiyVBgC4PriJThBKSVPWnnFHZAkfI1lJT8QFkOwH4qOS2SJkS4wvpGl8BpA==",
+      "dev": true,
+      "hasInstallScript": true,
+      "bin": {
+        "esbuild": "bin/esbuild"
+      },
+      "engines": {
+        "node": ">=12"
+      },
+      "optionalDependencies": {
+        "@esbuild/android-arm": "0.18.20",
+        "@esbuild/android-arm64": "0.18.20",
+        "@esbuild/android-x64": "0.18.20",
+        "@esbuild/darwin-arm64": "0.18.20",
+        "@esbuild/darwin-x64": "0.18.20",
+        "@esbuild/freebsd-arm64": "0.18.20",
+        "@esbuild/freebsd-x64": "0.18.20",
+        "@esbuild/linux-arm": "0.18.20",
+        "@esbuild/linux-arm64": "0.18.20",
+        "@esbuild/linux-ia32": "0.18.20",
+        "@esbuild/linux-loong64": "0.18.20",
+        "@esbuild/linux-mips64el": "0.18.20",
+        "@esbuild/linux-ppc64": "0.18.20",
+        "@esbuild/linux-riscv64": "0.18.20",
+        "@esbuild/linux-s390x": "0.18.20",
+        "@esbuild/linux-x64": "0.18.20",
+        "@esbuild/netbsd-x64": "0.18.20",
+        "@esbuild/openbsd-x64": "0.18.20",
+        "@esbuild/sunos-x64": "0.18.20",
+        "@esbuild/win32-arm64": "0.18.20",
+        "@esbuild/win32-ia32": "0.18.20",
+        "@esbuild/win32-x64": "0.18.20"
+      }
+    },
+    "node_modules/fsevents": {
+      "version": "2.3.3",
+      "resolved": "https://registry.npmjs.org/fsevents/-/fsevents-2.3.3.tgz",
+      "integrity": "sha512-5xoDfX+fL7faATnagmWPpbFtwh/R77WmMMqqHGS65C3vvB0YHrgF+B1YmZ3441tMj5n63k0212XNoJwzlhffQw==",
+      "dev": true,
+      "hasInstallScript": true,
+      "optional": true,
+      "os": [
+        "darwin"
+      ],
+      "engines": {
+        "node": "^8.16.0 || ^10.6.0 || >=11.0.0"
+      }
+    },
+    "node_modules/get-tsconfig": {
+      "version": "4.7.2",
+      "resolved": "https://registry.npmjs.org/get-tsconfig/-/get-tsconfig-4.7.2.tgz",
+      "integrity": "sha512-wuMsz4leaj5hbGgg4IvDU0bqJagpftG5l5cXIAvo8uZrqn0NJqwtfupTN00VnkQJPcIRrxYrm1Ue24btpCha2A==",
+      "dev": true,
+      "dependencies": {
+        "resolve-pkg-maps": "^1.0.0"
+      },
+      "funding": {
+        "url": "https://github.com/privatenumber/get-tsconfig?sponsor=1"
+      }
+    },
+    "node_modules/ollama-node": {
+      "version": "0.1.27",
+      "resolved": "https://registry.npmjs.org/ollama-node/-/ollama-node-0.1.27.tgz",
+      "integrity": "sha512-tFABPf5P0sXCR5USA31E3tqbge5h/4uf/t5j8/rPvHDo0SDwXeN0kah2J7hIqqkYlO1vLRs0uLC1/Mprgv9t2g==",
+      "dependencies": {
+        "@types/node": "^20.8.4"
+      }
+    },
+    "node_modules/resolve-pkg-maps": {
+      "version": "1.0.0",
+      "resolved": "https://registry.npmjs.org/resolve-pkg-maps/-/resolve-pkg-maps-1.0.0.tgz",
+      "integrity": "sha512-seS2Tj26TBVOC2NIc2rOe2y2ZO7efxITtLZcGSOnHHNOQ7CkiUBfw0Iw2ck6xkIhPwLhKNLS8BO+hEpngQlqzw==",
+      "dev": true,
+      "funding": {
+        "url": "https://github.com/privatenumber/resolve-pkg-maps?sponsor=1"
+      }
+    },
+    "node_modules/source-map": {
+      "version": "0.6.1",
+      "resolved": "https://registry.npmjs.org/source-map/-/source-map-0.6.1.tgz",
+      "integrity": "sha512-UjgapumWlbMhkBgzT7Ykc5YXUT46F0iKu8SGXq0bcwP5dz/h0Plj6enJqjz1Zbq2l5WaqYnrVbwWOWMyF3F47g==",
+      "dev": true,
+      "engines": {
+        "node": ">=0.10.0"
+      }
+    },
+    "node_modules/source-map-support": {
+      "version": "0.5.21",
+      "resolved": "https://registry.npmjs.org/source-map-support/-/source-map-support-0.5.21.tgz",
+      "integrity": "sha512-uBHU3L3czsIyYXKX88fdrGovxdSCoTGDRZ6SYXtSRxLZUzHg5P/66Ht6uoUlHu9EZod+inXhKo3qQgwXUT/y1w==",
+      "dev": true,
+      "dependencies": {
+        "buffer-from": "^1.0.0",
+        "source-map": "^0.6.0"
+      }
+    },
+    "node_modules/tsx": {
+      "version": "4.1.2",
+      "resolved": "https://registry.npmjs.org/tsx/-/tsx-4.1.2.tgz",
+      "integrity": "sha512-1spM1bFV6MP2s4tO4tDC7g52fsaFdtEWdO4GfGdqi20qUgPbnAJqixOyIAvCSx1DDj3YIUB4CD06owTWUsOAuQ==",
+      "dev": true,
+      "dependencies": {
+        "esbuild": "~0.18.20",
+        "get-tsconfig": "^4.7.2",
+        "source-map-support": "^0.5.21"
+      },
+      "bin": {
+        "tsx": "dist/cli.mjs"
+      },
+      "engines": {
+        "node": ">=18.0.0"
+      },
+      "optionalDependencies": {
+        "fsevents": "~2.3.3"
+      }
+    },
+    "node_modules/typescript": {
+      "version": "5.2.2",
+      "resolved": "https://registry.npmjs.org/typescript/-/typescript-5.2.2.tgz",
+      "integrity": "sha512-mI4WrpHsbCIcwT9cF4FZvr80QUeKvsUsUvKDoR+X/7XHQH98xYD8YHZg7ANtz2GtZt/CBq2QJ0thkGJMHfqc1w==",
+      "dev": true,
+      "bin": {
+        "tsc": "bin/tsc",
+        "tsserver": "bin/tsserver"
+      },
+      "engines": {
+        "node": ">=14.17"
+      }
+    },
+    "node_modules/undici-types": {
+      "version": "5.26.5",
+      "resolved": "https://registry.npmjs.org/undici-types/-/undici-types-5.26.5.tgz",
+      "integrity": "sha512-JlCMO+ehdEIKqlFxk6IfVoAUVmgz7cU7zD/h9XZ0qzeosSHmUJVOzSQvvYSYWXkFXC+IfLKSIffhv0sVZup6pA=="
+    }
+  }
+}

examples/typescript-functioncalling/package.json

@@ -0,0 +1,9 @@
+{
+  "dependencies": {
+    "ollama-node": "^0.1.27"
+  },
+  "devDependencies": {
+    "tsx": "^4.1.2",
+    "typescript": "^5.2.2"
+  }
+}

examples/typescript-functioncalling/readme.md

@@ -0,0 +1,28 @@
+# Function calling
+
+![function calling 2023-11-16 16_12_58](https://github.com/jmorganca/ollama/assets/633681/a0acc247-9746-45ab-b325-b65dfbbee4fb)
+
+One of the features added to some models is 'function calling'. It's a bit of a confusing name. It's understandable if you think that means the model can call functions, but that's not what it means. Function calling simply means that the output of the model is formatted in JSON, using a preconfigured schema, and uses the expected types. Then your code can use the output of the model and call functions with it. Using the JSON format in Ollama, you can use any model for function calling. 
+
+The two examples provided can extract information out of the provided texts. The first example uses the first couple of chapters from War and Peace by Lev Nikolayevich Tolstoy, and extracts the names and titles of the characters introduced in the story. The second example uses a more complicated schema to pull out addresses and event information from a series of emails.
+
+## Running the examples
+
+1. Clone this repo and navigate to the `examples/typescript-functioncalling` directory.
+2. Install the dependencies with `npm install`.
+3. Review the `wp.txt` file.
+4. Run `tsx extractwp.ts`.
+5. Review the `info.txt` file.
+6. Run `tsx extractemail.ts`.
+
+## Review the Code
+
+Both examples do roughly the same thing with different source material. They both use the same system prompt, which tells the model to expect some instructions and a schema. Then we inject the schema into the prompt and generate an answer.
+
+The first example, `extractwp.ts`, outputs the resulting JSON to the console, listing the characters introduced at the start of War and Peace. The second example, `extractemail.ts`, is a bit more complicated, extracting two different types of information: addresses and events. It outputs the results to a JSON blob, then the addresses are handed off to one function called `reportAddresses` and the events are handed off to another function called `reportEvents`.
+
+Notice that both examples are using the model from Intel called `neural-chat`. This is not a model tuned for function calling, yet it performs very well at this task.
+
+## Next Steps
+
+Try exporting some of your real emails to the input file and seeing how well the model does. Try pointing the first example at other books. You could even have it cycle through all the sections and maybe add up the number of times any character is seen throughout the book, determining the most important characters. You can also try out different models.

examples/typescript-functioncalling/wp.txt

@@ -0,0 +1,183 @@
+"Well, Prince, so Genoa and Lucca are now just family estates of the Buonapartes. But I warn you, if you don't tell me that this means war, if you still try to defend the infamies and horrors perpetrated by that Antichrist - I really believe he is Antichrist - I will have nothing more to do with you and you are no longer my friend, no longer my 'faithful slave,' as you call yourself! But how do you do? I see I have frightened you - sit down and tell me all the news."
+
+It was in July, 1805, and the speaker was the well-known Anna Pavlovna Scherer, maid of honor and favorite of the Empress Marya Fedorovna. With these words she greeted Prince Vasili Kuragin, a man of high rank and importance, who was the first to arrive at her reception. Anna Pavlovna had had a cough for some days. She was, as she said, suffering from la grippe; grippe being then a new word in St. Petersburg, used only by the elite.
+
+All her invitations without exception, written in French, and delivered by a scarlet-liveried footman that morning, ran as follows:
+
+"If you have nothing better to do, Count (or Prince), and if the prospect of spending an evening with a poor invalid is not too terrible, I shall be very charmed to see you tonight between 7 and 10 - Annette Scherer."
+
+"Heavens! what a virulent attack!" replied the prince, not in the least disconcerted by this reception. He had just entered, wearing an embroidered court uniform, knee breeches, and shoes, and had stars on his breast and a serene expression on his flat face. He spoke in that refined French in which our grandfathers not only spoke but thought, and with the gentle, patronizing intonation natural to a man of importance who had grown old in society and at court. He went up to Anna Pavlovna, kissed her hand, presenting to her his bald, scented, and shining head, and complacently seated himself on the sofa.
+
+"First of all, dear friend, tell me how you are. Set your friend's mind at rest," said he without altering his tone, beneath the politeness and affected sympathy of which indifference and even irony could be discerned.
+
+"Can one be well while suffering morally? Can one be calm in times like these if one has any feeling?" said Anna Pavlovna. "You are staying the whole evening, I hope?"
+
+"And the fete at the English ambassador's? Today is Wednesday. I must put in an appearance there," said the prince. "My daughter is coming for me to take me there."
+
+"I thought today's fete had been canceled. I confess all these festivities and fireworks are becoming wearisome."
+
+"If they had known that you wished it, the entertainment would have been put off," said the prince, who, like a wound-up clock, by force of habit said things he did not even wish to be believed.
+
+"Don't tease! Well, and what has been decided about Novosiltsev's dispatch? You know everything."
+
+"What can one say about it?" replied the prince in a cold, listless tone. "What has been decided? They have decided that Buonaparte has burnt his boats, and I believe that we are ready to burn ours."
+
+Prince Vasili always spoke languidly, like an actor repeating a stale part. Anna Pavlovna Scherer on the contrary, despite her forty years, overflowed with animation and impulsiveness. To be an enthusiast had become her social vocation and, sometimes even when she did not feel like it, she became enthusiastic in order not to disappoint the expectations of those who knew her. The subdued smile which, though it did not suit her faded features, always played round her lips expressed, as in a spoiled child, a continual consciousness of her charming defect, which she neither wished, nor could, nor considered it necessary, to correct.
+
+In the midst of a conversation on political matters Anna Pavlovna burst out:
+
+"Oh, don't speak to me of Austria. Perhaps I don't understand things, but Austria never has wished, and does not wish, for war. She is betraying us! Russia alone must save Europe. Our gracious sovereign recognizes his high vocation and will be true to it. That is the one thing I have faith in! Our good and wonderful sovereign has to perform the noblest role on earth, and he is so virtuous and noble that God will not forsake him. He will fulfill his vocation and crush the hydra of revolution, which has become more terrible than ever in the person of this murderer and villain! We alone must avenge the blood of the just one.... Whom, I ask you, can we rely on?... England with her commercial spirit will not and cannot understand the Emperor Alexander's loftiness of soul. She has refused to evacuate Malta. She wanted to find, and still seeks, some secret motive in our actions. What answer did Novosiltsev get? None. The English have not understood and cannot understand the self-abnegation of our Emperor who wants nothing for himself, but only desires the good of mankind. And what have they promised? Nothing! And what little they have promised they will not perform! Prussia has always declared that Buonaparte is invincible, and that all Europe is powerless before him.... And I don't believe a word that Hardenburg says, or Haugwitz either. This famous Prussian neutrality is just a trap. I have faith only in God and the lofty destiny of our adored monarch. He will save Europe!"
+
+She suddenly paused, smiling at her own impetuosity.
+
+"I think," said the prince with a smile, "that if you had been sent instead of our dear Wintzingerode you would have captured the King of Prussia's consent by assault. You are so eloquent. Will you give me a cup of tea?"
+
+"In a moment. A propos," she added, becoming calm again, "I am expecting two very interesting men tonight, le Vicomte de Mortemart, who is connected with the Montmorencys through the Rohans, one of the best French families. He is one of the genuine emigres, the good ones. And also the Abbe Morio. Do you know that profound thinker? He has been received by the Emperor. Had you heard?"
+
+"I shall be delighted to meet them," said the prince. "But tell me," he added with studied carelessness as if it had only just occurred to him, though the question he was about to ask was the chief motive of his visit, "is it true that the Dowager Empress wants Baron Funke to be appointed first secretary at Vienna? The baron by all accounts is a poor creature."
+
+Prince Vasili wished to obtain this post for his son, but others were trying through the Dowager Empress Marya Fedorovna to secure it for the baron.
+
+Anna Pavlovna almost closed her eyes to indicate that neither she nor anyone else had a right to criticize what the Empress desired or was pleased with.
+
+"Baron Funke has been recommended to the Dowager Empress by her sister," was all she said, in a dry and mournful tone.
+
+As she named the Empress, Anna Pavlovna's face suddenly assumed an expression of profound and sincere devotion and respect mingled with sadness, and this occurred every time she mentioned her illustrious patroness. She added that Her Majesty had deigned to show Baron Funke beaucoup d'estime, and again her face clouded over with sadness.
+
+The prince was silent and looked indifferent. But, with the womanly and courtierlike quickness and tact habitual to her, Anna Pavlovna wished both to rebuke him (for daring to speak as he had done of a man recommended to the Empress) and at the same time to console him, so she said:
+
+"Now about your family. Do you know that since your daughter came out everyone has been enraptured by her? They say she is amazingly beautiful."
+
+The prince bowed to signify his respect and gratitude.
+
+"I often think," she continued after a short pause, drawing nearer to the prince and smiling amiably at him as if to show that political and social topics were ended and the time had come for intimate conversation - "I often think how unfairly sometimes the joys of life are distributed. Why has fate given you two such splendid children? I don't speak of Anatole, your youngest. I don't like him," she added in a tone admitting of no rejoinder and raising her eyebrows. "Two such charming children. And really you appreciate them less than anyone, and so you don't deserve to have them."
+
+And she smiled her ecstatic smile.
+
+"I can't help it," said the prince. "Lavater would have said I lack the bump of paternity."
+
+"Don't joke; I mean to have a serious talk with you. Do you know I am dissatisfied with your younger son? Between ourselves" (and her face assumed its melancholy expression), "he was mentioned at Her Majesty's and you were pitied...."
+
+The prince answered nothing, but she looked at him significantly, awaiting a reply. He frowned.
+
+"What would you have me do?" he said at last. "You know I did all a father could for their education, and they have both turned out fools. Hippolyte is at least a quiet fool, but Anatole is an active one. That is the only difference between them." He said this smiling in a way more natural and animated than usual, so that the wrinkles round his mouth very clearly revealed something unexpectedly coarse and unpleasant.
+
+"And why are children born to such men as you? If you were not a father there would be nothing I could reproach you with," said Anna Pavlovna, looking up pensively.
+
+"I am your faithful slave and to you alone I can confess that my children are the bane of my life. It is the cross I have to bear. That is how I explain it to myself. It can't be helped!"
+
+He said no more, but expressed his resignation to cruel fate by a gesture. Anna Pavlovna meditated.
+
+"Have you never thought of marrying your prodigal son Anatole?" she asked. "They say old maids have a mania for matchmaking, and though I don't feel that weakness in myself as yet, I know a little person who is very unhappy with her father. She is a relation of yours, Princess Mary Bolkonskaya."
+
+Prince Vasili did not reply, though, with the quickness of memory and perception befitting a man of the world, he indicated by a movement of the head that he was considering this information.
+
+"Do you know," he said at last, evidently unable to check the sad current of his thoughts, "that Anatole is costing me forty thousand rubles a year? And," he went on after a pause, "what will it be in five years, if he goes on like this?" Presently he added: "That's what we fathers have to put up with.... Is this princess of yours rich?"
+
+"Her father is very rich and stingy. He lives in the country. He is the well-known Prince Bolkonski who had to retire from the army under the late Emperor, and was nicknamed 'the King of Prussia.' He is very clever but eccentric, and a bore. The poor girl is very unhappy. She has a brother; I think you know him, he married Lise Meinen lately. He is an aide-de-camp of Kutuzov's and will be here tonight."
+
+"Listen, dear Annette," said the prince, suddenly taking Anna Pavlovna's hand and for some reason drawing it downwards. "Arrange that affair for me and I shall always be your most devoted slave-slafe with an f, as a village elder of mine writes in his reports. She is rich and of good family and that's all I want."
+
+And with the familiarity and easy grace peculiar to him, he raised the maid of honor's hand to his lips, kissed it, and swung it to and fro as he lay back in his armchair, looking in another direction.
+
+"Attendez," said Anna Pavlovna, reflecting, "I'll speak to Lise, young Bolkonski's wife, this very evening, and perhaps the thing can be arranged. It shall be on your family's behalf that I'll start my apprenticeship as old maid."
+
+Anna Pavlovna's drawing room was gradually filling. The highest Petersburg society was assembled there: people differing widely in age and character but alike in the social circle to which they belonged. Prince Vasili's daughter, the beautiful Helene, came to take her father to the ambassador's entertainment; she wore a ball dress and her badge as maid of honor. The youthful little Princess Bolkonskaya, known as la femme la plus seduisante de Petersbourg, * was also there. She had been married during the previous winter, and being pregnant did not go to any large gatherings, but only to small receptions. Prince Vasili's son, Hippolyte, had come with Mortemart, whom he introduced. The Abbe Morio and many others had also come.
+
+* The most fascinating woman in Petersburg.
+
+To each new arrival Anna Pavlovna said, "You have not yet seen my aunt," or "You do not know my aunt?" and very gravely conducted him or her to a little old lady, wearing large bows of ribbon in her cap, who had come sailing in from another room as soon as the guests began to arrive; and slowly turning her eyes from the visitor to her aunt, Anna Pavlovna mentioned each one's name and then left them.
+
+Each visitor performed the ceremony of greeting this old aunt whom not one of them knew, not one of them wanted to know, and not one of them cared about; Anna Pavlovna observed these greetings with mournful and solemn interest and silent approval. The aunt spoke to each of them in the same words, about their health and her own, and the health of Her Majesty, "who, thank God, was better today." And each visitor, though politeness prevented his showing impatience, left the old woman with a sense of relief at having performed a vexatious duty and did not return to her the whole evening.
+
+The young Princess Bolkonskaya had brought some work in a gold-embroidered velvet bag. Her pretty little upper lip, on which a delicate dark down was just perceptible, was too short for her teeth, but it lifted all the more sweetly, and was especially charming when she occasionally drew it down to meet the lower lip. As is always the case with a thoroughly attractive woman, her defect - the shortness of her upper lip and her half-open mouth - seemed to be her own special and peculiar form of beauty. Everyone brightened at the sight of this pretty young woman, so soon to become a mother, so full of life and health, and carrying her burden so lightly. Old men and dull dispirited young ones who looked at her, after being in her company and talking to her a little while, felt as if they too were becoming, like her, full of life and health. All who talked to her, and at each word saw her bright smile and the constant gleam of her white teeth, thought that they were in a specially amiable mood that day.
+
+The little princess went round the table with quick, short, swaying steps, her workbag on her arm, and gaily spreading out her dress sat down on a sofa near the silver samovar, as if all she was doing was a pleasure to herself and to all around her. "I have brought my work," said she in French, displaying her bag and addressing all present. "Mind, Annette, I hope you have not played a wicked trick on me," she added, turning to her hostess. "You wrote that it was to be quite a small reception, and just see how badly I am dressed." And she spread out her arms to show her short-waisted, lace-trimmed, dainty gray dress, girdled with a broad ribbon just below the breast.
+
+"Soyez tranquille, Lise, you will always be prettier than anyone else," replied Anna Pavlovna.
+
+"You know," said the princess in the same tone of voice and still in French, turning to a general, "my husband is deserting me? He is going to get himself killed. Tell me what this wretched war is for?" she added, addressing Prince Vasili, and without waiting for an answer she turned to speak to his daughter, the beautiful Helene.
+
+"What a delightful woman this little princess is!" said Prince Vasili to Anna Pavlovna.
+
+One of the next arrivals was a stout, heavily built young man with close-cropped hair, spectacles, the light-colored breeches fashionable at that time, a very high ruffle, and a brown dress coat. This stout young man was an illegitimate son of Count Bezukhov, a well-known grandee of Catherine's time who now lay dying in Moscow. The young man had not yet entered either the military or civil service, as he had only just returned from abroad where he had been educated, and this was his first appearance in society. Anna Pavlovna greeted him with the nod she accorded to the lowest hierarchy in her drawing room. But in spite of this lowest-grade greeting, a look of anxiety and fear, as at the sight of something too large and unsuited to the place, came over her face when she saw Pierre enter. Though he was certainly rather bigger than the other men in the room, her anxiety could only have reference to the clever though shy, but observant and natural, expression which distinguished him from everyone else in that drawing room.
+
+"It is very good of you, Monsieur Pierre, to come and visit a poor invalid," said Anna Pavlovna, exchanging an alarmed glance with her aunt as she conducted him to her.
+
+Pierre murmured something unintelligible, and continued to look round as if in search of something. On his way to the aunt he bowed to the little princess with a pleased smile, as to an intimate acquaintance.
+
+Anna Pavlovna's alarm was justified, for Pierre turned away from the aunt without waiting to hear her speech about Her Majesty's health. Anna Pavlovna in dismay detained him with the words: "Do you know the Abbe Morio? He is a most interesting man."
+
+"Yes, I have heard of his scheme for perpetual peace, and it is very interesting but hardly feasible."
+
+"You think so?" rejoined Anna Pavlovna in order to say something and get away to attend to her duties as hostess. But Pierre now committed a reverse act of impoliteness. First he had left a lady before she had finished speaking to him, and now he continued to speak to another who wished to get away. With his head bent, and his big feet spread apart, he began explaining his reasons for thinking the abbe's plan chimerical.
+
+"We will talk of it later," said Anna Pavlovna with a smile.
+
+And having got rid of this young man who did not know how to behave, she resumed her duties as hostess and continued to listen and watch, ready to help at any point where the conversation might happen to flag. As the foreman of a spinning mill, when he has set the hands to work, goes round and notices here a spindle that has stopped or there one that creaks or makes more noise than it should, and hastens to check the machine or set it in proper motion, so Anna Pavlovna moved about her drawing room, approaching now a silent, now a too-noisy group, and by a word or slight rearrangement kept the conversational machine in steady, proper, and regular motion. But amid these cares her anxiety about Pierre was evident. She kept an anxious watch on him when he approached the group round Mortemart to listen to what was being said there, and again when he passed to another group whose center was the abbe.
+
+Pierre had been educated abroad, and this reception at Anna Pavlovna's was the first he had attended in Russia. He knew that all the intellectual lights of Petersburg were gathered there and, like a child in a toyshop, did not know which way to look, afraid of missing any clever conversation that was to be heard. Seeing the self-confident and refined expression on the faces of those present he was always expecting to hear something very profound. At last he came up to Morio. Here the conversation seemed interesting and he stood waiting for an opportunity to express his own views, as young people are fond of doing.
+
+CHAPTER III
+Anna Pavlovna's reception was in full swing. The spindles hummed steadily and ceaselessly on all sides. With the exception of the aunt, beside whom sat only one elderly lady, who with her thin careworn face was rather out of place in this brilliant society, the whole company had settled into three groups. One, chiefly masculine, had formed round the abbe. Another, of young people, was grouped round the beautiful Princess Helene, Prince Vasili's daughter, and the little Princess Bolkonskaya, very pretty and rosy, though rather too plump for her age. The third group was gathered round Mortemart and Anna Pavlovna.
+
+The vicomte was a nice-looking young man with soft features and polished manners, who evidently considered himself a celebrity but out of politeness modestly placed himself at the disposal of the circle in which he found himself. Anna Pavlovna was obviously serving him up as a treat to her guests. As a clever maitre d'hotel serves up as a specially choice delicacy a piece of meat that no one who had seen it in the kitchen would have cared to eat, so Anna Pavlovna served up to her guests, first the vicomte and then the abbe, as peculiarly choice morsels. The group about Mortemart immediately began discussing the murder of the Duc d'Enghien. The vicomte said that the Duc d'Enghien had perished by his own magnanimity, and that there were particular reasons for Buonaparte's hatred of him.
+
+"Ah, yes! Do tell us all about it, Vicomte," said Anna Pavlovna, with a pleasant feeling that there was something A la Louis XV in the sound of that sentence: "Contez nous cela, Vicomte."
+
+The vicomte bowed and smiled courteously in token of his willingness to comply. Anna Pavlovna arranged a group round him, inviting everyone to listen to his tale.
+
+"The vicomte knew the duc personally," whispered Anna Pavlovna to one of the guests. "The vicomte is a wonderful raconteur," said she to another. "How evidently he belongs to the best society," said she to a third; and the vicomte was served up to the company in the choicest and most advantageous style, like a well-garnished joint of roast beef on a hot dish.
+
+The vicomte wished to begin his story and gave a subtle smile.
+
+"Come over here, Helene, dear," said Anna Pavlovna to the beautiful young princess who was sitting some way off, the center of another group.
+
+The princess smiled. She rose with the same unchanging smile with which she had first entered the room - the smile of a perfectly beautiful woman. With a slight rustle of her white dress trimmed with moss and ivy, with a gleam of white shoulders, glossy hair, and sparkling diamonds, she passed between the men who made way for her, not looking at any of them but smiling on all, as if graciously allowing each the privilege of admiring her beautiful figure and shapely shoulders, back, and bosom - which in the fashion of those days were very much exposed - and she seemed to bring the glamour of a ballroom with her as she moved toward Anna Pavlovna. Helene was so lovely that not only did she not show any trace of coquetry, but on the contrary she even appeared shy of her unquestionable and all too victorious beauty. She seemed to wish, but to be unable, to diminish its effect.
+
+"How lovely!" said everyone who saw her; and the vicomte lifted his shoulders and dropped his eyes as if startled by something extraordinary when she took her seat opposite and beamed upon him also with her unchanging smile.
+
+"Madame, I doubt my ability before such an audience," said he, smilingly inclining his head.
+
+The princess rested her bare round arm on a little table and considered a reply unnecessary. She smilingly waited. All the time the story was being told she sat upright, glancing now at her beautiful round arm, altered in shape by its pressure on the table, now at her still more beautiful bosom, on which she readjusted a diamond necklace. From time to time she smoothed the folds of her dress, and whenever the story produced an effect she glanced at Anna Pavlovna, at once adopted just the expression she saw on the maid of honor's face, and again relapsed into her radiant smile.
+
+The little princess had also left the tea table and followed Helene.
+
+"Wait a moment, I'll get my work.... Now then, what are you thinking of?" she went on, turning to Prince Hippolyte. "Fetch me my workbag."
+
+There was a general movement as the princess, smiling and talking merrily to everyone at once, sat down and gaily arranged herself in her seat.
+
+"Now I am all right," she said, and asking the vicomte to begin, she took up her work.
+
+Prince Hippolyte, having brought the workbag, joined the circle and moving a chair close to hers seated himself beside her.
+
+Le charmant Hippolyte was surprising by his extraordinary resemblance to his beautiful sister, but yet more by the fact that in spite of this resemblance he was exceedingly ugly. His features were like his sister's, but while in her case everything was lit up by a joyous, self-satisfied, youthful, and constant smile of animation, and by the wonderful classic beauty of her figure, his face on the contrary was dulled by imbecility and a constant expression of sullen self-confidence, while his body was thin and weak. His eyes, nose, and mouth all seemed puckered into a vacant, wearied grimace, and his arms and legs always fell into unnatural positions.
+
+"It's not going to be a ghost story?" said he, sitting down beside the princess and hastily adjusting his lorgnette, as if without this instrument he could not begin to speak.
+
+"Why no, my dear fellow," said the astonished narrator, shrugging his shoulders.
+
+"Because I hate ghost stories," said Prince Hippolyte in a tone which showed that he only understood the meaning of his words after he had uttered them.
+
+He spoke with such self-confidence that his hearers could not be sure whether what he said was very witty or very stupid. He was dressed in a dark-green dress coat, knee breeches of the color of cuisse de nymphe effrayee, as he called it, shoes, and silk stockings.
+
+The vicomte told his tale very neatly. It was an anecdote, then current, to the effect that the Duc d'Enghien had gone secretly to Paris to visit Mademoiselle George; that at her house he came upon Bonaparte, who also enjoyed the famous actress' favors, and that in his presence Napoleon happened to fall into one of the fainting fits to which he was subject, and was thus at the duc's mercy. The latter spared him, and this magnanimity Bonaparte subsequently repaid by death.
+
+The story was very pretty and interesting, especially at the point where the rivals suddenly recognized one another; and the ladies looked agitated.
+
+"Charming!" said Anna Pavlovna with an inquiring glance at the little princess.
+
+"Charming!" whispered the little princess, sticking the needle into her work as if to testify that the interest and fascination of the story prevented her from going on with it.
+
+The vicomte appreciated this silent praise and smiling gratefully prepared to continue, but just then Anna Pavlovna, who had kept a watchful eye on the young man who so alarmed her, noticed that he was talking too loudly and vehemently with the abbe, so she hurried to the rescue. Pierre had managed to start a conversation with the abbe about the balance of power, and the latter, evidently interested by the young man's simple-minded eagerness, was explaining his pet theory. Both were talking and listening too eagerly and too naturally, which was why Anna Pavlovna disapproved.
+
+"The means are ... the balance of power in Europe and the rights of the people," the abbe was saying. "It is only necessary for one powerful nation like Russia - barbaric as she is said to be - to place herself disinterestedly at the head of an alliance having for its object the maintenance of the balance of power of Europe, and it would save the world!"
+
+"But how are you to get that balance?" Pierre was beginning.
+
+At that moment Anna Pavlovna came up and, looking severely at Pierre, asked the Italian how he stood Russian climate. The Italian's face instantly changed and assumed an offensively affected, sugary expression, evidently habitual to him when conversing with women.
+
+"I am so enchanted by the brilliancy of the wit and culture of the society, more especially of the feminine society, in which I have had the honor of being received, that I have not yet had time to think of the climate," said he.
+
+Not letting the abbe and Pierre escape, Anna Pavlovna, the more conveniently to keep them under observation, brought them into the larger circle.
+

examples/typescript-mentors/README.md

@@ -4,18 +4,62 @@ This example demonstrates how one would create a set of 'mentors' you can have a
 
 ## Usage
 
-```bash
-ts-node ./character-generator.ts "Lorne Greene"
-```
+1. Add llama2 to have the mentors ask your questions:
 
-This will create `lornegreene/Modelfile`. Now you can create a model with this command:
+   ```bash
+   ollama pull llama2
+   ```
 
-```bash
-ollama create lornegreene -f lornegreene/Modelfile
-```
+2. Install prerequisites:
 
-If you want to add your own mentors, you will have to update the code to look at your namespace instead of **mattw**. Also set the list of mentors to include yours.
+   ```bash
+   npm install
+   ```
 
-```bash
-ts-node ./mentors.ts "What is a Jackalope?"
-```
+3. Ask a question:
+
+   ```bash
+   npm start "what is a jackalope"
+   ```
+
+You can also add your own character to be chosen at random when you ask a question.
+
+1. Make sure you have the right model installed:
+
+   ```bash
+   ollama pull stablebeluga2:70b-q4_K_M
+   ```
+  
+2. Create a new character:
+  
+   ```bash
+   npm run charactergen "Lorne Greene"
+   ```
+
+   You can choose any well-known person you like. This example will create `lornegreene/Modelfile`.
+
+3. Now you can create a model with this command:
+
+   ```bash
+   ollama create <YourNamespace>/lornegreene -f lornegreene/Modelfile
+   ```
+
+   `YourNamespace` is whatever name you set up when you signed up at [https://ollama.ai/signup](https://ollama.ai/signup).
+
+4. To add this to your mentors, you will have to update the code as follows. On line 8 of `mentors.ts`, add an object to the array, replacing `<YourNamespace>` with the namespace you used above.
+
+   ```bash
+   {ns: "<YourNamespace>", char: "Lorne Greene"}
+   ```
+
+## Review the Code
+
+There are two scripts you can run in this example. The first is the main script to ask the mentors a question. The other one lets you generate a character to add to the mentors. Both scripts are mostly about adjusting the prompts at each inference stage.
+
+### mentors.ts
+
+In the **main** function, it starts by generating a list of mentors. This chooses 3 from a list of interesting characters. Then we ask for a question, and then things get interesting. We set the prompt for each of the 3 mentors a little differently. And the 2nd and 3rd mentors see what the previous folks said. The other functions in mentors sets the prompts for each mentor.
+
+### character-generator.ts
+
+**Character Generator** simply customizes the prompt to build a character profile for any famous person. And most of the script is just tweaking the prompt. This uses Stable Beluga 2 70b parameters. The 70b models tend to do better writing a bio about a character than smaller models, and Stable Beluga seemed to do better than Llama 2. Since this is used at development time for the characters, it doesn't affect the runtime of asking the mentors for their input.

examples/typescript-mentors/mentors.ts

@@ -2,10 +2,11 @@ import { Ollama } from 'ollama-node';
 
 const mentorCount = 3;
 const ollama = new Ollama();
+type Mentor = { ns: string, char: string };
 
-function getMentors(): string[] {
-  const mentors = ['Gary Vaynerchuk', 'Kanye West', 'Martha Stewart', 'Neil deGrasse Tyson', 'Owen Wilson', 'Ronald Reagan', 'Donald Trump', 'Barack Obama', 'Jeff Bezos'];
-  const chosenMentors: string[] = [];
+function getMentors(): Mentor[] {
+  const mentors = [{ ns: 'mattw', char: 'Gary Vaynerchuk' }, { ns: 'mattw', char: 'Kanye West'}, {ns: 'mattw', char: 'Martha Stewart'}, {ns: 'mattw', char: 'Neil deGrasse Tyson'}, {ns: 'mattw', char: 'Owen Wilson'}, {ns: 'mattw', char: 'Ronald Reagan'}, {ns: 'mattw', char: 'Donald Trump'}, {ns: 'mattw', char: 'Barack Obama'}, {ns: 'mattw', char: 'Jeff Bezos'}];
+  const chosenMentors: Mentor[] = [];
   for (let i = 0; i < mentorCount; i++) {
     const mentor = mentors[Math.floor(Math.random() * mentors.length)];
     chosenMentors.push(mentor);
@@ -14,12 +15,12 @@ function getMentors(): string[] {
   return chosenMentors;
 }
 
-function getMentorFileName(mentor: string): string {
-  const model = mentor.toLowerCase().replace(/\s/g, '');
-  return `mattw/${model}`;
+function getMentorFileName(mentor: Mentor): string {
+  const model = mentor.char.toLowerCase().replace(/\s/g, '');
+  return `${mentor.ns}/${model}`;
 }
 
-async function getSystemPrompt(mentor: string, isLast: boolean, question: string): Promise<string> {
+async function getSystemPrompt(mentor: Mentor, isLast: boolean, question: string): Promise<string> {
   ollama.setModel(getMentorFileName(mentor));
   const info = await ollama.showModelInfo()
   let SystemPrompt = info.system || '';
@@ -43,8 +44,8 @@ async function main() {
     ollama.setModel(getMentorFileName(mentor));
     ollama.setSystemPrompt(SystemPrompt);
     let output = '';
-    process.stdout.write(`\n${mentor}: `);
-    for await (const chunk of ollama.streamingGenerate(theConversation + `Continue the conversation as if you were ${mentor} on the question "${question}".`)) {
+    process.stdout.write(`\n${mentor.char}: `);
+    for await (const chunk of ollama.streamingGenerate(theConversation + `Continue the conversation as if you were ${mentor.char} on the question "${question}".`)) {
       if (chunk.response) {
         output += chunk.response;
         process.stdout.write(chunk.response);
@@ -52,7 +53,7 @@ async function main() {
         process.stdout.write('\n');
       }
     }
-    theConversation += `${mentor}: ${output}\n\n`
+    theConversation += `${mentor.char}: ${output}\n\n`
   }
 }
 

examples/typescript-mentors/package.json

@@ -1,7 +1,15 @@
 {
+  "scripts": {
+    "charactergen": "tsx character-generator.ts", 
+    "start": "tsx mentors.ts"
+  },
   "dependencies": {
     "fs": "^0.0.1-security",
     "ollama-node": "^0.0.3",
     "path": "^0.12.7"
+  },
+  "devDependencies": {
+    "tsx": "^4.6.2",
+    "typescript": "^5.3.3"
   }
 }

examples/typescript-simplechat/client.ts

@@ -0,0 +1,77 @@
+import * as readline from "readline";
+
+const model = "llama2";
+type Message = {
+  role: "assistant" | "user" | "system";
+  content: string;
+}
+const messages: Message[] = [{
+  role: "system",
+  content: "You are a helpful AI agent."
+}]
+
+const rl = readline.createInterface({
+  input: process.stdin,
+  output: process.stdout
+})
+
+async function chat(messages: Message[]): Promise<Message> {
+  const body = {
+    model: model,
+    messages: messages
+  }
+
+  const response = await fetch("http://localhost:11434/api/chat", {
+    method: "POST",
+    body: JSON.stringify(body)
+  })
+
+  const reader = response.body?.getReader()
+  if (!reader) {
+    throw new Error("Failed to read response body")
+  }
+  let content = ""
+  while (true) {
+    const { done, value } = await reader.read()
+    if (done) {
+      break;
+    }
+    const rawjson = new TextDecoder().decode(value);
+    const json = JSON.parse(rawjson)
+
+    if (json.done === false) {
+      process.stdout.write(json.message.content);
+      content += json.message.content
+    }
+
+  }
+  return { role: "assistant", content: content };
+}
+
+async function askQuestion(): Promise<void> {
+  return new Promise<void>((resolve) => {
+    rl.question("\n\nAsk a question: (press enter alone to quit)\n\n", async (user_input) => {
+      if (user_input.trim() === "") {
+        rl.close();
+        console.log("Thankyou. Goodbye.\n")
+        console.log("=======\nHere is the message history that was used in this conversation.\n=======\n")
+        messages.forEach(message => {
+          console.log(message)
+        })
+        resolve();
+      } else {
+        console.log();
+        messages.push({ role: "user", content: user_input });
+        messages.push(await chat(messages));
+        await askQuestion(); // Ask the next question
+      }
+    });
+  });
+}
+
+async function main() {
+  await askQuestion();
+
+}
+
+main();

examples/typescript-simplechat/package.json

@@ -0,0 +1,12 @@
+{ 
+  "scripts": {
+    "start": "tsx client.ts"
+  }, 
+  "dependencies": {
+     "@types/node": "^20.10.4", 
+     "prompt-sync": "^4.2.0", 
+     "readline": "^1.3.0", 
+     "tsx": "^4.6.2", 
+     "typescript": "^5.3.3" 
+     } 
+    }

examples/typescript-simplechat/readme.md

@@ -0,0 +1,35 @@
+# Simple Chat Example
+
+The **chat** endpoint, available as of v0.1.14, is one of two ways to generate text from an LLM with Ollama. At a high level, you provide the endpoint an array of message objects with a role and content specified. Then with each output and prompt, you add more messages, which builds up the history.
+
+## Run the Example
+
+`npm start`
+
+## Review the Code
+
+You can see in the **chat** function that is actually calling the endpoint is simply done with:
+
+```typescript
+const body = {
+  model: model,
+  messages: messages
+}
+
+const response = await fetch("http://localhost:11434/api/chat", {
+  method: "POST",
+  body: JSON.stringify(body)
+})
+```
+
+With the **generate** endpoint, you need to provide a `prompt`. But with **chat**, you provide `messages`. And the resulting stream of responses includes a `message` object with a `content` field.
+
+The final JSON object doesn't provide the full content, so you will need to build the content yourself. In this example, **chat** takes the full array of messages and outputs the resulting message from this call of the chat endpoint.
+
+In the **askQuestion** function, we collect `user_input` and add it as a message to our messages, and that is passed to the chat function. When the LLM is done responding, the output is added as another message to the messages array.
+
+At the end, you will see a printout of all the messages.
+
+## Next Steps
+
+In this example, all generations are kept. You might want to experiment with summarizing everything older than 10 conversations to enable longer history with less context being used.

format/bytes.go

@@ -1,16 +1,47 @@
 package format
 
-import "fmt"
+import (
+	"fmt"
+	"math"
+)
+
+const (
+	Byte     = 1
+	KiloByte = Byte * 1000
+	MegaByte = KiloByte * 1000
+	GigaByte = MegaByte * 1000
+	TeraByte = GigaByte * 1000
+)
 
 func HumanBytes(b int64) string {
+	var value float64
+	var unit string
+
 	switch {
-	case b > 1000*1000*1000:
-		return fmt.Sprintf("%d GB", b/1000/1000/1000)
-	case b > 1000*1000:
-		return fmt.Sprintf("%d MB", b/1000/1000)
-	case b > 1000:
-		return fmt.Sprintf("%d KB", b/1000)
+	case b >= TeraByte:
+		value = float64(b) / TeraByte
+		unit = "TB"
+	case b >= GigaByte:
+		value = float64(b) / GigaByte
+		unit = "GB"
+	case b >= MegaByte:
+		value = float64(b) / MegaByte
+		unit = "MB"
+	case b >= KiloByte:
+		value = float64(b) / KiloByte
+		unit = "KB"
 	default:
 		return fmt.Sprintf("%d B", b)
 	}
+
+	switch {
+	case value >= 100:
+		return fmt.Sprintf("%d %s", int(value), unit)
+	case value >= 10:
+		return fmt.Sprintf("%d %s", int(value), unit)
+	case value != math.Trunc(value):
+		return fmt.Sprintf("%.1f %s", value, unit)
+	default:
+		return fmt.Sprintf("%d %s", int(value), unit)
+	}
 }

format/format.go

@@ -0,0 +1,25 @@
+package format
+
+import (
+	"fmt"
+	"math"
+)
+
+const (
+	Thousand = 1000
+	Million  = Thousand * 1000
+	Billion  = Million * 1000
+)
+
+func HumanNumber(b uint64) string {
+	switch {
+	case b > Billion:
+		return fmt.Sprintf("%.0fB", math.Round(float64(b)/Billion))
+	case b > Million:
+		return fmt.Sprintf("%.0fM", math.Round(float64(b)/Million))
+	case b > Thousand:
+		return fmt.Sprintf("%.0fK", math.Round(float64(b)/Thousand))
+	default:
+		return fmt.Sprintf("%d", b)
+	}
+}

format/time_test.go

@@ -29,7 +29,7 @@ func TestHumanTime(t *testing.T) {
 	})
 
 	t.Run("soon", func(t *testing.T) {
-		v := now.Add(800*time.Millisecond)
+		v := now.Add(800 * time.Millisecond)
 		assertEqual(t, HumanTime(v, ""), "Less than a second from now")
 	})
 }