Update images_test.go

When a username or model name is uppercase the registry will reject the
push. This is done for file-system compatibility. If we rely on the registry
error on push the message returned is 'file not found', which does not
convey why the push actually failed.

README.md

@@ -346,6 +346,9 @@ See the [API documentation](./docs/api.md) for all endpoints.
 - [Web management](https://github.com/lemonit-eric-mao/ollama-web-management) (Web management page)
 - [Promptery](https://github.com/promptery/promptery) (desktop client for Ollama.)
 - [Ollama App](https://github.com/JHubi1/ollama-app) (Modern and easy-to-use multi-platform client for Ollama)
+- [SpaceLlama](https://github.com/tcsenpai/spacellama) (Firefox and Chrome extension to quickly summarize web pages with ollama in a sidebar)
+- [YouLama](https://github.com/tcsenpai/youlama) (Webapp to quickly summarize any YouTube video, supporting Invidious as well)
+- [DualMind](https://github.com/tcsenpai/dualmind) (Experimental app allowing two models to talk to each other in the terminal or in a web interface)
 - [ollamarama-matrix](https://github.com/h1ddenpr0cess20/ollamarama-matrix) (Ollama chatbot for the Matrix chat protocol)
 - [ollama-chat-app](https://github.com/anan1213095357/ollama-chat-app) (Flutter-based chat app)
 - [Perfect Memory AI](https://www.perfectmemory.ai/) (Productivity AI assists personalized by what you have seen on your screen, heard and said in the meetings)
@@ -356,6 +359,7 @@ See the [API documentation](./docs/api.md) for all endpoints.
 - [Nosia](https://github.com/nosia-ai/nosia) (Easy to install and use RAG platform based on Ollama)
 - [Witsy](https://github.com/nbonamy/witsy) (An AI Desktop application avaiable for Mac/Windows/Linux)
 - [Abbey](https://github.com/US-Artificial-Intelligence/abbey) (A configurable AI interface server with notebooks, document storage, and YouTube support)
+- [Minima](https://github.com/dmayboroda/minima) (RAG with on-premises or fully local workflow)
 
 ### Cloud
 
@@ -504,7 +508,7 @@ See the [API documentation](./docs/api.md) for all endpoints.
 - [Terraform AWS Ollama & Open WebUI](https://github.com/xuyangbocn/terraform-aws-self-host-llm) (A Terraform module to deploy on AWS a ready-to-use Ollama service, together with its front end Open WebUI service.)
 - [node-red-contrib-ollama](https://github.com/jakubburkiewicz/node-red-contrib-ollama)
 - [Local AI Helper](https://github.com/ivostoykov/localAI) (Chrome and Firefox extensions that enable interactions with the active tab and customisable API endpoints. Includes secure storage for user prompts.)
-- [vnc-lm](https://github.com/jk011ru/vnc-lm) (A containerized Discord bot with support for attachments and web links)
+- [vnc-lm](https://github.com/jake83741/vnc-lm) (Discord bot for messaging with LLMs through Ollama and LiteLLM. Seamlessly move between local and flagship models.)
 - [LSP-AI](https://github.com/SilasMarvin/lsp-ai) (Open-source language server for AI-powered functionality)
 - [QodeAssist](https://github.com/Palm1r/QodeAssist) (AI-powered coding assistant plugin for Qt Creator)
 - [Obsidian Quiz Generator plugin](https://github.com/ECuiDev/obsidian-quiz-generator)

api/types.go

@@ -146,6 +146,7 @@ type ToolCall struct {
 }
 
 type ToolCallFunction struct {
+	Index     int                       `json:"index,omitempty"`
 	Name      string                    `json:"name"`
 	Arguments ToolCallFunctionArguments `json:"arguments"`
 }

cmd/cmd_test.go

@@ -8,7 +8,6 @@ import (
 	"net/http"
 	"net/http/httptest"
 	"os"
-	"path/filepath"
 	"strings"
 	"testing"
 
@@ -180,18 +179,14 @@ Weigh anchor!
 
 	t.Run("license", func(t *testing.T) {
 		var b bytes.Buffer
-		license, err := os.ReadFile(filepath.Join("..", "LICENSE"))
-		if err != nil {
-			t.Fatal(err)
-		}
-
+		license := "MIT License\nCopyright (c) Ollama\n"
 		if err := showInfo(&api.ShowResponse{
 			Details: api.ModelDetails{
 				Family:            "test",
 				ParameterSize:     "7B",
 				QuantizationLevel: "FP16",
 			},
-			License: string(license),
+			License: license,
 		}, &b); err != nil {
 			t.Fatal(err)
 		}

cmd/interactive.go

@@ -514,7 +514,7 @@ func extractFileNames(input string) []string {
 	// Regex to match file paths starting with optional drive letter, / ./ \ or .\ and include escaped or unescaped spaces (\ or %20)
 	// and followed by more characters and a file extension
 	// This will capture non filename strings, but we'll check for file existence to remove mismatches
-	regexPattern := `(?:[a-zA-Z]:)?(?:\./|/|\\)[\S\\ ]+?\.(?i:jpg|jpeg|png|svg)\b`
+	regexPattern := `(?:[a-zA-Z]:)?(?:\./|/|\\)[\S\\ ]+?\.(?i:jpg|jpeg|png)\b`
 	re := regexp.MustCompile(regexPattern)
 
 	return re.FindAllString(input, -1)

cmd/interactive_test.go

@@ -12,44 +12,45 @@ import (
 func TestExtractFilenames(t *testing.T) {
 	// Unix style paths
 	input := ` some preamble 
- ./relative\ path/one.png inbetween1 ./not a valid two.jpg inbetween2
-/unescaped space /three.jpeg inbetween3 /valid\ path/dir/four.png "./quoted with spaces/five.svg`
+ ./relative\ path/one.png inbetween1 ./not a valid two.jpg inbetween2 ./1.svg
+/unescaped space /three.jpeg inbetween3 /valid\ path/dir/four.png "./quoted with spaces/five.JPG`
 	res := extractFileNames(input)
 	assert.Len(t, res, 5)
 	assert.Contains(t, res[0], "one.png")
 	assert.Contains(t, res[1], "two.jpg")
 	assert.Contains(t, res[2], "three.jpeg")
 	assert.Contains(t, res[3], "four.png")
-	assert.Contains(t, res[4], "five.svg")
+	assert.Contains(t, res[4], "five.JPG")
 	assert.NotContains(t, res[4], '"')
-	assert.NotContains(t, res, "inbtween")
+	assert.NotContains(t, res, "inbetween1")
+	assert.NotContains(t, res, "./1.svg")
 
 	// Windows style paths
 	input = ` some preamble
  c:/users/jdoe/one.png inbetween1 c:/program files/someplace/two.jpg inbetween2 
  /absolute/nospace/three.jpeg inbetween3 /absolute/with space/four.png inbetween4
-./relative\ path/five.svg inbetween5 "./relative with/spaces/six.png inbetween6
-d:\path with\spaces\seven.svg inbetween7 c:\users\jdoe\eight.png inbetween8 
- d:\program files\someplace\nine.png inbetween9 "E:\program files\someplace\ten.svg some ending
+./relative\ path/five.JPG inbetween5 "./relative with/spaces/six.png inbetween6
+d:\path with\spaces\seven.JPEG inbetween7 c:\users\jdoe\eight.png inbetween8 
+ d:\program files\someplace\nine.png inbetween9 "E:\program files\someplace\ten.PNG some ending
 `
 	res = extractFileNames(input)
 	assert.Len(t, res, 10)
-	assert.NotContains(t, res, "inbtween")
+	assert.NotContains(t, res, "inbetween2")
 	assert.Contains(t, res[0], "one.png")
 	assert.Contains(t, res[0], "c:")
 	assert.Contains(t, res[1], "two.jpg")
 	assert.Contains(t, res[1], "c:")
 	assert.Contains(t, res[2], "three.jpeg")
 	assert.Contains(t, res[3], "four.png")
-	assert.Contains(t, res[4], "five.svg")
+	assert.Contains(t, res[4], "five.JPG")
 	assert.Contains(t, res[5], "six.png")
-	assert.Contains(t, res[6], "seven.svg")
+	assert.Contains(t, res[6], "seven.JPEG")
 	assert.Contains(t, res[6], "d:")
 	assert.Contains(t, res[7], "eight.png")
 	assert.Contains(t, res[7], "c:")
 	assert.Contains(t, res[8], "nine.png")
 	assert.Contains(t, res[8], "d:")
-	assert.Contains(t, res[9], "ten.svg")
+	assert.Contains(t, res[9], "ten.PNG")
 	assert.Contains(t, res[9], "E:")
 }
 

docs/api.md

@@ -49,10 +49,10 @@ Advanced parameters (optional):
 - `options`: additional model parameters listed in the documentation for the [Modelfile](./modelfile.md#valid-parameters-and-values) such as `temperature`
 - `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 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
 - `keep_alive`: controls how long the model will stay loaded into memory following the request (default: `5m`)
+- `context` (deprecated): the context parameter returned from a previous request to `/generate`, this can be used to keep a short conversational memory
 
 #### JSON mode
 

docs/modelfile.md

@@ -63,7 +63,7 @@ SYSTEM You are Mario from super mario bros, acting as an assistant.
 To use this:
 
 1. Save it as a file (e.g. `Modelfile`)
-2. `ollama create choose-a-model-name -f <location of the file e.g. ./Modelfile>'`
+2. `ollama create choose-a-model-name -f <location of the file e.g. ./Modelfile>`
 3. `ollama run choose-a-model-name`
 4. Start using the model!
 

llama/README.md

@@ -93,7 +93,7 @@ make -j
 
 ## Vendoring
 
-Ollama currently vendors [llama.cpp](https://github.com/ggerganov/llama.cpp/) and [ggml](https://github.com/ggerganov/ggml) through a vendoring model.  While we generally strive to contribute changes back upstream to avoid drift, we cary a small set of patches which are applied to the tracking commit.  A set of make targets are available to aid developers in updating to a newer tracking commit, or to work on changes.
+Ollama currently vendors [llama.cpp](https://github.com/ggerganov/llama.cpp/) and [ggml](https://github.com/ggerganov/ggml) through a vendoring model. While we generally strive to contribute changes back upstream to avoid drift, we cary a small set of patches which are applied to the tracking commit. A set of make targets are available to aid developers in updating to a newer tracking commit, or to work on changes.
 
 If you update the vendoring code, start by running the following command to establish the tracking llama.cpp repo in the `./vendor/` directory.
 
@@ -105,35 +105,35 @@ make apply-patches
 
 **Pin to new base commit**
 
-To update to a newer base commit, select the upstream git tag or commit and update `llama/vendoring.env`
+To update to a newer base commit, select the upstream git tag or commit and update `llama/vendoring`
 
 #### Applying patches
 
 When updating to a newer base commit, the existing patches may not apply cleanly and require manual merge resolution.
 
-Start by applying the patches.  If any of the patches have conflicts, the `git am` will stop at the first failure.
+Start by applying the patches. If any of the patches have conflicts, the `git am` will stop at the first failure.
 
 ```
 make apply-patches
 ```
 
-If you see an error message about a conflict, go into the `./vendor/` directory, and perform merge resolution using your preferred tool to the patch commit which failed.  Save the file(s) and continue the patch series with `git am --continue` .  If any additional patches fail, follow the same pattern until the full patch series is applied.  Once finished, run a final `create-patches` and `sync` target to ensure everything is updated.
+If you see an error message about a conflict, go into the `./vendor/` directory, and perform merge resolution using your preferred tool to the patch commit which failed. Save the file(s) and continue the patch series with `git am --continue` . If any additional patches fail, follow the same pattern until the full patch series is applied. Once finished, run a final `create-patches` and `sync` target to ensure everything is updated.
 
 ```
 make create-patches sync
 ```
 
-Build and test Ollama, and make any necessary changes to the Go code based on the new base commit.  Submit your PR to the Ollama repo.
+Build and test Ollama, and make any necessary changes to the Go code based on the new base commit. Submit your PR to the Ollama repo.
 
 ### Generating Patches
 
-When working on new fixes or features that impact vendored code, use the following model.  First get a clean tracking repo with all current patches applied:
+When working on new fixes or features that impact vendored code, use the following model. First get a clean tracking repo with all current patches applied:
 
 ```
 make apply-patches
 ```
 
-Now edit the upstream native code in the `./vendor/` directory.  You do not need to commit every change in order to build, a dirty working tree in the tracking repo is OK while developing.  Simply save in your editor, and run the following to refresh the vendored code with your changes, build the backend(s) and build ollama:
+Now edit the upstream native code in the `./vendor/` directory. You do not need to commit every change in order to build, a dirty working tree in the tracking repo is OK while developing. Simply save in your editor, and run the following to refresh the vendored code with your changes, build the backend(s) and build ollama:
 
 ```
 make sync
@@ -142,9 +142,9 @@ go build .
 ```
 
 > [!IMPORTANT]
-> Do **NOT** run `apply-patches` while you're iterating as that will reset the tracking repo.  It will detect a dirty tree and abort, but if your tree is clean and you accidentally ran this target, use `git reflog` to recover your commit(s).
+> Do **NOT** run `apply-patches` while you're iterating as that will reset the tracking repo. It will detect a dirty tree and abort, but if your tree is clean and you accidentally ran this target, use `git reflog` to recover your commit(s).
 
-Iterate until you're ready to submit PRs.  Once your code is ready, commit a change in the `./vendor/` directory, then generate the patches for ollama with
+Iterate until you're ready to submit PRs. Once your code is ready, commit a change in the `./vendor/` directory, then generate the patches for ollama with
 
 ```
 make create-patches
@@ -157,4 +157,4 @@ In your `./vendor/` directory, create a branch, and cherry-pick the new commit t
 
 Commit the changes in the ollama repo and submit a PR to Ollama, which will include the vendored code update with your change, along with the patches.
 
-After your PR upstream is merged, follow the **Updating Base Commit** instructions above, however first remove your patch before running `apply-patches` since the new base commit contains your change already.
+After your PR upstream is merged, follow the **Updating Base Commit** instructions above, however first remove your patch before running `apply-patches` since the new base commit contains your change already.

llama/runner/cache.go

@@ -199,6 +199,20 @@ func countCommonPrefix(a []input, b []input) int {
 	return count
 }
 
+func (c *InputCache) ShiftDiscard(inputLen int, numKeep int) int {
+	targetFree := (c.numCtx - numKeep) / 2
+	targetFree = max(targetFree, 1)
+
+	currentFree := c.numCtx - inputLen
+	discard := targetFree - currentFree
+
+	if discard < 0 {
+		discard = 0
+	}
+
+	return discard
+}
+
 // Frees up space in the KV cache by deleting the oldest half of history and shifting
 // the newest half into that space (saving numKeep inputs at the beginning).
 //
@@ -208,11 +222,7 @@ func (c *InputCache) ShiftCacheSlot(slot *InputCacheSlot, numKeep int) error {
 		return fmt.Errorf("unable to shift context - keep exceeds context (keep: %v context: %v)", numKeep, c.numCtx)
 	}
 
-	targetFree := (c.numCtx - numKeep) / 2
-	targetFree = max(targetFree, 1)
-
-	currentFree := c.numCtx - len(slot.Inputs)
-	discard := targetFree - currentFree
+	discard := c.ShiftDiscard(len(slot.Inputs), numKeep)
 
 	if discard <= 0 {
 		return nil

llama/runner/cache_test.go

@@ -227,3 +227,66 @@ func TestFindCacheSlot(t *testing.T) {
 		})
 	}
 }
+
+func TestShiftDiscard(t *testing.T) {
+	tests := []struct {
+		name     string
+		numCtx   int
+		numKeep  int
+		inputLen int
+		expected int
+	}{
+		{
+			name:     "Shift",
+			numCtx:   2048,
+			numKeep:  5,
+			inputLen: 2048,
+			expected: 1021,
+		},
+		{
+			name:     "Max Keep",
+			numCtx:   2048,
+			numKeep:  2047,
+			inputLen: 2048,
+			expected: 1,
+		},
+		{
+			name:     "No Keep",
+			numCtx:   2048,
+			numKeep:  0,
+			inputLen: 2048,
+			expected: 1024,
+		},
+		{
+			name:     "Truncate",
+			numCtx:   2048,
+			numKeep:  5,
+			inputLen: 5000,
+			expected: 3973,
+		},
+		{
+			name:     "Truncate Keep",
+			numCtx:   2048,
+			numKeep:  2047,
+			inputLen: 5000,
+			expected: 2953,
+		},
+		{
+			name:     "No Op",
+			numCtx:   2048,
+			numKeep:  5,
+			inputLen: 512,
+			expected: 0,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			c := InputCache{numCtx: tt.numCtx}
+			result := c.ShiftDiscard(tt.inputLen, tt.numKeep)
+			if result != tt.expected {
+				t.Errorf("shiftDiscard(ctx: %v, keep: %v input: %v): have %v; want %v", tt.numCtx, tt.numKeep, tt.inputLen, result, tt.expected)
+			}
+		})
+	}
+}

llama/runner/runner.go

@@ -122,9 +122,11 @@ func (s *Server) NewSequence(prompt string, images []ImageData, params NewSequen
 	params.numKeep = min(params.numKeep, s.cache.numCtx-1)
 
 	if len(inputs) > s.cache.numCtx {
-		slog.Warn("truncating input prompt", "limit", s.cache.numCtx, "prompt", len(inputs), "numKeep", params.numKeep)
+		discard := len(inputs) - s.cache.numCtx
 		newInputs := inputs[:params.numKeep]
-		newInputs = append(newInputs, inputs[len(inputs)-s.cache.numCtx+params.numKeep:]...)
+		newInputs = append(newInputs, inputs[params.numKeep+discard:]...)
+
+		slog.Warn("truncating input prompt", "limit", s.cache.numCtx, "prompt", len(inputs), "keep", params.numKeep, "new", len(newInputs))
 		inputs = newInputs
 	}
 
@@ -162,10 +164,16 @@ func (s *Server) NewSequence(prompt string, images []ImageData, params NewSequen
 // generating image embeddings for each image
 func (s *Server) inputs(prompt string, images []ImageData) ([]input, error) {
 	var inputs []input
+	var parts []string
+	var matches [][]string
 
-	re := regexp.MustCompile(`\[img-(\d+)\]`)
-	parts := re.Split(prompt, -1)
-	matches := re.FindAllStringSubmatch(prompt, -1)
+	if s.image != nil {
+		re := regexp.MustCompile(`\[img-(\d+)\]`)
+		parts = re.Split(prompt, -1)
+		matches = re.FindAllStringSubmatch(prompt, -1)
+	} else {
+		parts = []string{prompt}
+	}
 
 	for i, part := range parts {
 		// text - tokenize
@@ -825,10 +833,21 @@ func (s *Server) health(w http.ResponseWriter, r *http.Request) {
 	}
 }
 
+type multiLPath []string
+
+func (m *multiLPath) Set(value string) error {
+	*m = append(*m, value)
+	return nil
+}
+
+func (m *multiLPath) String() string {
+	return strings.Join(*m, ", ")
+}
+
 func (s *Server) loadModel(
 	params llama.ModelParams,
 	mpath string,
-	lpath string,
+	lpath multiLPath,
 	ppath string,
 	kvSize int,
 	flashAttention bool,
@@ -849,10 +868,12 @@ func (s *Server) loadModel(
 		panic(err)
 	}
 
-	if lpath != "" {
-		err := s.model.ApplyLoraFromFile(s.lc, lpath, 1.0, threads)
-		if err != nil {
-			panic(err)
+	if lpath.String() != "" {
+		for _, path := range lpath {
+			err := s.model.ApplyLoraFromFile(s.lc, path, 1.0, threads)
+			if err != nil {
+				panic(err)
+			}
 		}
 	}
 
@@ -882,7 +903,6 @@ func main() {
 	mainGpu := flag.Int("main-gpu", 0, "Main GPU")
 	flashAttention := flag.Bool("flash-attn", false, "Enable flash attention")
 	kvSize := flag.Int("ctx-size", 2048, "Context (or KV cache) size")
-	lpath := flag.String("lora", "", "Path to lora layer file")
 	port := flag.Int("port", 8080, "Port to expose the server on")
 	threads := flag.Int("threads", runtime.NumCPU(), "Number of threads to use during generation")
 	verbose := flag.Bool("verbose", false, "verbose output (default: disabled)")
@@ -892,6 +912,9 @@ func main() {
 	multiUserCache := flag.Bool("multiuser-cache", false, "optimize input cache algorithm for multiple users")
 	requirements := flag.Bool("requirements", false, "print json requirement information")
 
+	var lpaths multiLPath
+	flag.Var(&lpaths, "lora", "Path to lora layer file (can be specified multiple times)")
+
 	flag.Parse()
 	if *requirements {
 		printRequirements(os.Stdout)
@@ -938,7 +961,7 @@ func main() {
 	params := llama.ModelParams{
 		NumGpuLayers: *nGpuLayers,
 		MainGpu:      *mainGpu,
-		UseMmap:      !*noMmap && *lpath == "",
+		UseMmap:      !*noMmap && lpaths.String() == "",
 		UseMlock:     *mlock,
 		TensorSplit:  tensorSplitFloats,
 		Progress: func(progress float32) {
@@ -947,7 +970,7 @@ func main() {
 	}
 
 	server.ready.Add(1)
-	go server.loadModel(params, *mpath, *lpath, *ppath, *kvSize, *flashAttention, *threads, *multiUserCache)
+	go server.loadModel(params, *mpath, lpaths, *ppath, *kvSize, *flashAttention, *threads, *multiUserCache)
 
 	server.cond = sync.NewCond(&server.mu)
 

llm/server.go

@@ -144,10 +144,6 @@ func NewLlamaServer(gpus discover.GpuInfoList, model string, ggml *GGML, adapter
 	// Loop through potential servers
 	finalErr := errors.New("no suitable llama servers found")
 
-	if len(adapters) > 1 {
-		return nil, errors.New("ollama supports only one lora adapter, but multiple were provided")
-	}
-
 	rDir, err := runners.Refresh(build.EmbedFS)
 	if err != nil {
 		return nil, err
@@ -201,8 +197,9 @@ func NewLlamaServer(gpus discover.GpuInfoList, model string, ggml *GGML, adapter
 	}
 
 	if len(adapters) > 0 {
-		// TODO: applying multiple adapters is not supported by the llama.cpp server yet
-		params = append(params, "--lora", adapters[0])
+		for _, adapter := range adapters {
+			params = append(params, "--lora", adapter)
+		}
 	}
 
 	if len(projectors) > 0 {

openai/openai.go

@@ -140,6 +140,7 @@ type CompletionChunk struct {
 
 type ToolCall struct {
 	ID       string `json:"id"`
+	Index    int    `json:"index"`
 	Type     string `json:"type"`
 	Function struct {
 		Name      string `json:"name"`
@@ -200,12 +201,13 @@ func toolCallId() string {
 	return "call_" + strings.ToLower(string(b))
 }
 
-func toChatCompletion(id string, r api.ChatResponse) ChatCompletion {
-	toolCalls := make([]ToolCall, len(r.Message.ToolCalls))
-	for i, tc := range r.Message.ToolCalls {
+func toToolCalls(tc []api.ToolCall) []ToolCall {
+	toolCalls := make([]ToolCall, len(tc))
+	for i, tc := range tc {
 		toolCalls[i].ID = toolCallId()
 		toolCalls[i].Type = "function"
 		toolCalls[i].Function.Name = tc.Function.Name
+		toolCalls[i].Index = tc.Function.Index
 
 		args, err := json.Marshal(tc.Function.Arguments)
 		if err != nil {
@@ -215,7 +217,11 @@ func toChatCompletion(id string, r api.ChatResponse) ChatCompletion {
 
 		toolCalls[i].Function.Arguments = string(args)
 	}
+	return toolCalls
+}
 
+func toChatCompletion(id string, r api.ChatResponse) ChatCompletion {
+	toolCalls := toToolCalls(r.Message.ToolCalls)
 	return ChatCompletion{
 		Id:                id,
 		Object:            "chat.completion",
@@ -244,6 +250,7 @@ func toChatCompletion(id string, r api.ChatResponse) ChatCompletion {
 }
 
 func toChunk(id string, r api.ChatResponse) ChatCompletionChunk {
+	toolCalls := toToolCalls(r.Message.ToolCalls)
 	return ChatCompletionChunk{
 		Id:                id,
 		Object:            "chat.completion.chunk",
@@ -252,7 +259,7 @@ func toChunk(id string, r api.ChatResponse) ChatCompletionChunk {
 		SystemFingerprint: "fp_ollama",
 		Choices: []ChunkChoice{{
 			Index: 0,
-			Delta: Message{Role: "assistant", Content: r.Message.Content},
+			Delta: Message{Role: "assistant", Content: r.Message.Content, ToolCalls: toolCalls},
 			FinishReason: func(reason string) *string {
 				if len(reason) > 0 {
 					return &reason
@@ -571,7 +578,7 @@ type EmbedWriter struct {
 	model string
 }
 
-func (w *BaseWriter) writeError(code int, data []byte) (int, error) {
+func (w *BaseWriter) writeError(data []byte) (int, error) {
 	var serr api.StatusError
 	err := json.Unmarshal(data, &serr)
 	if err != nil {
@@ -630,7 +637,7 @@ func (w *ChatWriter) writeResponse(data []byte) (int, error) {
 func (w *ChatWriter) Write(data []byte) (int, error) {
 	code := w.ResponseWriter.Status()
 	if code != http.StatusOK {
-		return w.writeError(code, data)
+		return w.writeError(data)
 	}
 
 	return w.writeResponse(data)
@@ -679,7 +686,7 @@ func (w *CompleteWriter) writeResponse(data []byte) (int, error) {
 func (w *CompleteWriter) Write(data []byte) (int, error) {
 	code := w.ResponseWriter.Status()
 	if code != http.StatusOK {
-		return w.writeError(code, data)
+		return w.writeError(data)
 	}
 
 	return w.writeResponse(data)
@@ -704,7 +711,7 @@ func (w *ListWriter) writeResponse(data []byte) (int, error) {
 func (w *ListWriter) Write(data []byte) (int, error) {
 	code := w.ResponseWriter.Status()
 	if code != http.StatusOK {
-		return w.writeError(code, data)
+		return w.writeError(data)
 	}
 
 	return w.writeResponse(data)
@@ -730,7 +737,7 @@ func (w *RetrieveWriter) writeResponse(data []byte) (int, error) {
 func (w *RetrieveWriter) Write(data []byte) (int, error) {
 	code := w.ResponseWriter.Status()
 	if code != http.StatusOK {
-		return w.writeError(code, data)
+		return w.writeError(data)
 	}
 
 	return w.writeResponse(data)
@@ -755,7 +762,7 @@ func (w *EmbedWriter) writeResponse(data []byte) (int, error) {
 func (w *EmbedWriter) Write(data []byte) (int, error) {
 	code := w.ResponseWriter.Status()
 	if code != http.StatusOK {
-		return w.writeError(code, data)
+		return w.writeError(data)
 	}
 
 	return w.writeResponse(data)

openai/openai_test.go

@@ -195,7 +195,86 @@ func TestChatMiddleware(t *testing.T) {
 				Stream: &False,
 			},
 		},
-
+		{
+			name: "chat handler with streaming tools",
+			body: `{
+				"model": "test-model",
+				"messages": [
+					{"role": "user", "content": "What's the weather like in Paris?"}
+				],
+				"stream": true,
+				"tools": [{
+					"type": "function",
+					"function": {
+						"name": "get_weather",
+						"description": "Get the current weather",
+						"parameters": {
+							"type": "object",
+							"required": ["location"],
+							"properties": {
+								"location": {
+									"type": "string",
+									"description": "The city and state"
+								},
+								"unit": {
+									"type": "string",
+									"enum": ["celsius", "fahrenheit"]
+								}
+							}
+						}
+					}
+				}]
+			}`,
+			req: api.ChatRequest{
+				Model: "test-model",
+				Messages: []api.Message{
+					{
+						Role:    "user",
+						Content: "What's the weather like in Paris?",
+					},
+				},
+				Tools: []api.Tool{
+					{
+						Type: "function",
+						Function: api.ToolFunction{
+							Name:        "get_weather",
+							Description: "Get the current weather",
+							Parameters: struct {
+								Type       string   `json:"type"`
+								Required   []string `json:"required"`
+								Properties map[string]struct {
+									Type        string   `json:"type"`
+									Description string   `json:"description"`
+									Enum        []string `json:"enum,omitempty"`
+								} `json:"properties"`
+							}{
+								Type:     "object",
+								Required: []string{"location"},
+								Properties: map[string]struct {
+									Type        string   `json:"type"`
+									Description string   `json:"description"`
+									Enum        []string `json:"enum,omitempty"`
+								}{
+									"location": {
+										Type:        "string",
+										Description: "The city and state",
+									},
+									"unit": {
+										Type: "string",
+										Enum: []string{"celsius", "fahrenheit"},
+									},
+								},
+							},
+						},
+					},
+				},
+				Options: map[string]any{
+					"temperature": 1.0,
+					"top_p":       1.0,
+				},
+				Stream: &True,
+			},
+		},
 		{
 			name: "chat handler error forwarding",
 			body: `{

server/images.go

@@ -802,6 +802,12 @@ func PushModel(ctx context.Context, name string, regOpts *registryOptions, fn fu
 	if mp.ProtocolScheme == "http" && !regOpts.Insecure {
 		return errors.New("insecure protocol http")
 	}
+	if mp.Namespace != strings.ToLower(mp.Namespace) {
+		return fmt.Errorf("namespace must be lowercase, but is %s", mp.Namespace)
+	}
+	if mp.Repository != strings.ToLower(mp.Repository) {
+		return fmt.Errorf("model name must be lowercase, but is %s", mp.Repository)
+	}
 
 	manifest, _, err := GetManifest(mp)
 	if err != nil {

server/images_test.go

@@ -0,0 +1,50 @@
+package server
+
+import (
+	"context"
+	"strings"
+	"testing"
+
+	"github.com/ollama/ollama/api"
+)
+
+func TestPushModel(t *testing.T) {
+	noOpProgress := func(resp api.ProgressResponse) {}
+
+	tests := []struct {
+		modelStr string
+		regOpts  *registryOptions
+		wantErr  string
+	}{
+		{
+			modelStr: "http://example.com/namespace/repo:tag",
+			regOpts:  &registryOptions{Insecure: false},
+			wantErr:  "insecure protocol http",
+		},
+		{
+			modelStr: "docker://Example/repo:tag",
+			regOpts:  &registryOptions{},
+			wantErr:  "namespace must be lowercase, but is Example",
+		},
+		{
+			modelStr: "docker://example/Repo:tag",
+			regOpts:  &registryOptions{},
+			wantErr:  "model name must be lowercase, but is Repo",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.modelStr, func(t *testing.T) {
+			err := PushModel(context.Background(), tt.modelStr, tt.regOpts, noOpProgress)
+
+			if tt.wantErr != "" {
+				if err == nil {
+					t.Errorf("PushModel() error = %v, wantErr %v", err, tt.wantErr)
+				} else if !strings.Contains(err.Error(), tt.wantErr) {
+					t.Errorf("PushModel() error = %v, wantErr %v", err, tt.wantErr)
+				}
+				return
+			}
+		})
+	}
+}

server/model_test.go

@@ -39,6 +39,7 @@ func TestExecuteWithTools(t *testing.T) {
 		{"mistral", `[TOOL_CALLS]  [{"name": "get_current_weather", "arguments": {"format":"fahrenheit","location":"San Francisco, CA"}},{"name": "get_current_weather", "arguments": {"format":"celsius","location":"Toronto, Canada"}}]
 
 The temperature in San Francisco, CA is 70°F and in Toronto, Canada is 20°C.`, true},
+		{"mistral", `[TOOL_CALLS]  [{"name": "get_current_weather", "arguments": {"format":"fahrenheit","location":"San Francisco, CA"}},{"name": "get_current_weather", "arguments": {"format":"celsius","location":"Toronto, Canada"}},{"name": "get_current_weather", "arguments": {"format":"celsius","location":"To }]`, false},
 		{"mistral", `I'm not aware of that information. However, I can suggest searching for the weather using the "get_current_weather" function:
 
 		[{"name": "get_current_weather", "arguments": {"format":"fahrenheit","location":"San Francisco, CA"}},{"name": "get_current_weather", "arguments": {"format":"celsius","location":"Toronto, Canada"}}]`, true},

server/routes.go

@@ -251,6 +251,7 @@ func (s *Server) GenerateHandler(c *gin.Context) {
 
 		var b bytes.Buffer
 		if req.Context != nil {
+			slog.Warn("the context field is deprecated and will be removed in a future version of Ollama")
 			s, err := r.Detokenize(c.Request.Context(), req.Context)
 			if err != nil {
 				c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
@@ -1458,6 +1459,7 @@ func (s *Server) ChatHandler(c *gin.Context) {
 
 	prompt, images, err := chatPrompt(c.Request.Context(), m, r.Tokenize, opts, msgs, req.Tools)
 	if err != nil {
+		slog.Error("chat prompt error", "error", err)
 		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
 		return
 	}
@@ -1467,6 +1469,8 @@ func (s *Server) ChatHandler(c *gin.Context) {
 	ch := make(chan any)
 	go func() {
 		defer close(ch)
+		var sb strings.Builder
+		var toolCallIndex int = 0
 		if err := r.Completion(c.Request.Context(), llm.CompletionRequest{
 			Prompt:  prompt,
 			Images:  images,
@@ -1492,7 +1496,37 @@ func (s *Server) ChatHandler(c *gin.Context) {
 				res.LoadDuration = checkpointLoaded.Sub(checkpointStart)
 			}
 
-			ch <- res
+			// TODO: tool call checking and filtering should be moved outside of this callback once streaming
+			// however this was a simple change for now without reworking streaming logic of this (and other)
+			// handlers
+			if req.Stream != nil && !*req.Stream || len(req.Tools) == 0 {
+				ch <- res
+				return
+			}
+
+			// Streaming tool calls:
+			// If tools are recognized, use a flag to track the sending of a tool downstream
+			// This ensures that content is cleared from the message on the last chunk sent
+			sb.WriteString(r.Content)
+			if toolCalls, ok := m.parseToolCalls(sb.String()); ok {
+				res.Message.ToolCalls = toolCalls
+				for i := range toolCalls {
+					toolCalls[i].Function.Index = toolCallIndex
+					toolCallIndex++
+				}
+				res.Message.Content = ""
+				sb.Reset()
+				ch <- res
+				return
+			}
+
+			if r.Done {
+				// Send any remaining content if no tool calls were detected
+				if toolCallIndex == 0 {
+					res.Message.Content = sb.String()
+				}
+				ch <- res
+			}
 		}); err != nil {
 			ch <- gin.H{"error": err.Error()}
 		}

server/routes_generate_test.go

@@ -8,6 +8,7 @@ import (
 	"io"
 	"net/http"
 	"strings"
+	"sync"
 	"testing"
 	"time"
 
@@ -25,10 +26,14 @@ type mockRunner struct {
 	// CompletionRequest is only valid until the next call to Completion
 	llm.CompletionRequest
 	llm.CompletionResponse
+	CompletionFn func(context.Context, llm.CompletionRequest, func(llm.CompletionResponse)) error
 }
 
-func (m *mockRunner) Completion(_ context.Context, r llm.CompletionRequest, fn func(r llm.CompletionResponse)) error {
+func (m *mockRunner) Completion(ctx context.Context, r llm.CompletionRequest, fn func(r llm.CompletionResponse)) error {
 	m.CompletionRequest = r
+	if m.CompletionFn != nil {
+		return m.CompletionFn(ctx, r, fn)
+	}
 	fn(m.CompletionResponse)
 	return nil
 }
@@ -88,9 +93,14 @@ func TestGenerateChat(t *testing.T) {
 		Model: "test",
 		Modelfile: fmt.Sprintf(`FROM %s
 		TEMPLATE """
-{{- if .System }}System: {{ .System }} {{ end }}
-{{- if .Prompt }}User: {{ .Prompt }} {{ end }}
-{{- if .Response }}Assistant: {{ .Response }} {{ end }}"""
+{{- if .Tools }}
+{{ .Tools }}
+{{ end }}
+{{- range .Messages }}
+{{- .Role }}: {{ .Content }}
+{{- range .ToolCalls }}{"name": "{{ .Function.Name }}", "arguments": {{ .Function.Arguments }}}
+{{- end }}
+{{ end }}"""
 `, createBinFile(t, llm.KV{
 			"general.architecture":          "llama",
 			"llama.block_count":             uint32(1),
@@ -263,7 +273,7 @@ func TestGenerateChat(t *testing.T) {
 			t.Errorf("expected status 200, got %d", w.Code)
 		}
 
-		if diff := cmp.Diff(mock.CompletionRequest.Prompt, "User: Hello! "); diff != "" {
+		if diff := cmp.Diff(mock.CompletionRequest.Prompt, "user: Hello!\n"); diff != "" {
 			t.Errorf("mismatch (-got +want):\n%s", diff)
 		}
 
@@ -292,7 +302,7 @@ func TestGenerateChat(t *testing.T) {
 			t.Errorf("expected status 200, got %d", w.Code)
 		}
 
-		if diff := cmp.Diff(mock.CompletionRequest.Prompt, "System: You are a helpful assistant. User: Hello! "); diff != "" {
+		if diff := cmp.Diff(mock.CompletionRequest.Prompt, "system: You are a helpful assistant.\nuser: Hello!\n"); diff != "" {
 			t.Errorf("mismatch (-got +want):\n%s", diff)
 		}
 
@@ -314,7 +324,7 @@ func TestGenerateChat(t *testing.T) {
 			t.Errorf("expected status 200, got %d", w.Code)
 		}
 
-		if diff := cmp.Diff(mock.CompletionRequest.Prompt, "System: You can perform magic tricks. User: Hello! "); diff != "" {
+		if diff := cmp.Diff(mock.CompletionRequest.Prompt, "system: You can perform magic tricks.\nuser: Hello!\n"); diff != "" {
 			t.Errorf("mismatch (-got +want):\n%s", diff)
 		}
 
@@ -337,12 +347,242 @@ func TestGenerateChat(t *testing.T) {
 			t.Errorf("expected status 200, got %d", w.Code)
 		}
 
-		if diff := cmp.Diff(mock.CompletionRequest.Prompt, "System: You are a helpful assistant. User: Hello! Assistant: I can help you with that. System: You can perform magic tricks. User: Help me write tests. "); diff != "" {
+		if diff := cmp.Diff(mock.CompletionRequest.Prompt, "system: You are a helpful assistant.\nuser: Hello!\nassistant: I can help you with that.\nsystem: You can perform magic tricks.\nuser: Help me write tests.\n"); diff != "" {
 			t.Errorf("mismatch (-got +want):\n%s", diff)
 		}
 
 		checkChatResponse(t, w.Body, "test-system", "Abra kadabra!")
 	})
+
+	t.Run("messages with tools (non-streaming)", func(t *testing.T) {
+		if w.Code != http.StatusOK {
+			t.Fatalf("failed to create test-system model: %d", w.Code)
+		}
+
+		tools := []api.Tool{
+			{
+				Type: "function",
+				Function: api.ToolFunction{
+					Name:        "get_weather",
+					Description: "Get the current weather",
+					Parameters: struct {
+						Type       string   `json:"type"`
+						Required   []string `json:"required"`
+						Properties map[string]struct {
+							Type        string   `json:"type"`
+							Description string   `json:"description"`
+							Enum        []string `json:"enum,omitempty"`
+						} `json:"properties"`
+					}{
+						Type:     "object",
+						Required: []string{"location"},
+						Properties: map[string]struct {
+							Type        string   `json:"type"`
+							Description string   `json:"description"`
+							Enum        []string `json:"enum,omitempty"`
+						}{
+							"location": {
+								Type:        "string",
+								Description: "The city and state",
+							},
+							"unit": {
+								Type: "string",
+								Enum: []string{"celsius", "fahrenheit"},
+							},
+						},
+					},
+				},
+			},
+		}
+
+		mock.CompletionResponse = llm.CompletionResponse{
+			Content:            `{"name":"get_weather","arguments":{"location":"Seattle, WA","unit":"celsius"}}`,
+			Done:               true,
+			DoneReason:         "done",
+			PromptEvalCount:    1,
+			PromptEvalDuration: 1,
+			EvalCount:          1,
+			EvalDuration:       1,
+		}
+
+		streamRequest := true
+
+		w := createRequest(t, s.ChatHandler, api.ChatRequest{
+			Model: "test-system",
+			Messages: []api.Message{
+				{Role: "user", Content: "What's the weather in Seattle?"},
+			},
+			Tools:  tools,
+			Stream: &streamRequest,
+		})
+
+		if w.Code != http.StatusOK {
+			var errResp struct {
+				Error string `json:"error"`
+			}
+			if err := json.NewDecoder(w.Body).Decode(&errResp); err != nil {
+				t.Logf("Failed to decode error response: %v", err)
+			} else {
+				t.Logf("Error response: %s", errResp.Error)
+			}
+		}
+
+		if w.Code != http.StatusOK {
+			t.Errorf("expected status 200, got %d", w.Code)
+		}
+
+		var resp api.ChatResponse
+		if err := json.NewDecoder(w.Body).Decode(&resp); err != nil {
+			t.Fatal(err)
+		}
+
+		if resp.Message.ToolCalls == nil {
+			t.Error("expected tool calls, got nil")
+		}
+
+		expectedToolCall := api.ToolCall{
+			Function: api.ToolCallFunction{
+				Name: "get_weather",
+				Arguments: api.ToolCallFunctionArguments{
+					"location": "Seattle, WA",
+					"unit":     "celsius",
+				},
+			},
+		}
+
+		if diff := cmp.Diff(resp.Message.ToolCalls[0], expectedToolCall); diff != "" {
+			t.Errorf("tool call mismatch (-got +want):\n%s", diff)
+		}
+	})
+
+	t.Run("messages with tools (streaming)", func(t *testing.T) {
+		tools := []api.Tool{
+			{
+				Type: "function",
+				Function: api.ToolFunction{
+					Name:        "get_weather",
+					Description: "Get the current weather",
+					Parameters: struct {
+						Type       string   `json:"type"`
+						Required   []string `json:"required"`
+						Properties map[string]struct {
+							Type        string   `json:"type"`
+							Description string   `json:"description"`
+							Enum        []string `json:"enum,omitempty"`
+						} `json:"properties"`
+					}{
+						Type:     "object",
+						Required: []string{"location"},
+						Properties: map[string]struct {
+							Type        string   `json:"type"`
+							Description string   `json:"description"`
+							Enum        []string `json:"enum,omitempty"`
+						}{
+							"location": {
+								Type:        "string",
+								Description: "The city and state",
+							},
+							"unit": {
+								Type: "string",
+								Enum: []string{"celsius", "fahrenheit"},
+							},
+						},
+					},
+				},
+			},
+		}
+
+		// Simulate streaming response with multiple chunks
+		var wg sync.WaitGroup
+		wg.Add(1)
+
+		mock.CompletionFn = func(ctx context.Context, r llm.CompletionRequest, fn func(r llm.CompletionResponse)) error {
+			defer wg.Done()
+
+			// Send chunks with small delays to simulate streaming
+			responses := []llm.CompletionResponse{
+				{
+					Content:            `{"name":"get_`,
+					Done:               false,
+					PromptEvalCount:    1,
+					PromptEvalDuration: 1,
+				},
+				{
+					Content:            `weather","arguments":{"location":"Seattle`,
+					Done:               false,
+					PromptEvalCount:    2,
+					PromptEvalDuration: 1,
+				},
+				{
+					Content:            `, WA","unit":"celsius"}}`,
+					Done:               true,
+					DoneReason:         "tool_call",
+					PromptEvalCount:    3,
+					PromptEvalDuration: 1,
+				},
+			}
+
+			for _, resp := range responses {
+				select {
+				case <-ctx.Done():
+					return ctx.Err()
+				default:
+					fn(resp)
+					time.Sleep(10 * time.Millisecond) // Small delay between chunks
+				}
+			}
+			return nil
+		}
+
+		w := createRequest(t, s.ChatHandler, api.ChatRequest{
+			Model: "test-system",
+			Messages: []api.Message{
+				{Role: "user", Content: "What's the weather in Seattle?"},
+			},
+			Tools:  tools,
+			Stream: &stream,
+		})
+
+		wg.Wait()
+
+		if w.Code != http.StatusOK {
+			t.Errorf("expected status 200, got %d", w.Code)
+		}
+
+		// Read and validate the streamed responses
+		decoder := json.NewDecoder(w.Body)
+		var finalToolCall api.ToolCall
+
+		for {
+			var resp api.ChatResponse
+			if err := decoder.Decode(&resp); err == io.EOF {
+				break
+			} else if err != nil {
+				t.Fatal(err)
+			}
+
+			if resp.Done {
+				if len(resp.Message.ToolCalls) != 1 {
+					t.Errorf("expected 1 tool call in final response, got %d", len(resp.Message.ToolCalls))
+				}
+				finalToolCall = resp.Message.ToolCalls[0]
+			}
+		}
+
+		expectedToolCall := api.ToolCall{
+			Function: api.ToolCallFunction{
+				Name: "get_weather",
+				Arguments: api.ToolCallFunctionArguments{
+					"location": "Seattle, WA",
+					"unit":     "celsius",
+				},
+			},
+		}
+
+		if diff := cmp.Diff(finalToolCall, expectedToolCall); diff != "" {
+			t.Errorf("final tool call mismatch (-got +want):\n%s", diff)
+		}
+	})
 }
 
 func TestGenerate(t *testing.T) {