change name to context_length

Co-authored-by: Michael Yang <mxyng@pm.me>

README.md

@@ -369,6 +369,7 @@ See the [API documentation](./docs/api.md) for all endpoints.
 - [Minima](https://github.com/dmayboroda/minima) (RAG with on-premises or fully local workflow)
 - [aidful-ollama-model-delete](https://github.com/AidfulAI/aidful-ollama-model-delete) (User interface for simplified model cleanup)
 - [Perplexica](https://github.com/ItzCrazyKns/Perplexica) (An AI-powered search engine & an open-source alternative to Perplexity AI)
+- [AI Toolkit for Visual Studio Code](https://aka.ms/ai-tooklit/ollama-docs) (Microsoft-official VSCode extension to chat, test, evaluate models with Ollama support, and use them in your AI applications.)
 
 ### Cloud
 
@@ -481,6 +482,7 @@ See the [API documentation](./docs/api.md) for all endpoints.
 - [GoLamify](https://github.com/prasad89/golamify)
 - [Ollama for Haskell](https://github.com/tusharad/ollama-haskell)
 - [multi-llm-ts](https://github.com/nbonamy/multi-llm-ts) (A Typescript/JavaScript library allowing access to different LLM in unified API)
+- [LlmTornado](https://github.com/lofcz/llmtornado) (C# library providing a unified interface for major FOSS & Commercial inference APIs)
 
 ### Mobile
 

convert/convert.go

@@ -193,8 +193,6 @@ func ConvertModel(fsys fs.FS, ws io.WriteSeeker) error {
 		conv = &bertModel{}
 	case "CohereForCausalLM":
 		conv = &commandrModel{}
-	case "Cohere2ForCausalLM":
-		conv = &cohere2Model{}
 	default:
 		return errors.New("unsupported architecture")
 	}

convert/convert_cohere2.go

@@ -1,85 +0,0 @@
-package convert
-
-import (
-	"cmp"
-
-	"github.com/ollama/ollama/llm"
-)
-
-type cohere2Model struct {
-	ModelParameters
-	MaxPositionEmbeddings uint32  `json:"max_position_embeddings"`
-	HiddenSize            uint32  `json:"hidden_size"`
-	HiddenLayers          uint32  `json:"num_hidden_layers"`
-	IntermediateSize      uint32  `json:"intermediate_size"`
-	NumAttentionHeads     uint32  `json:"num_attention_heads"`
-	NumKeyValueHeads      uint32  `json:"num_key_value_heads"`
-	LayerNormEPS          float32 `json:"layer_norm_eps"`
-	RopeTheta             float32 `json:"rope_theta"`
-	UseQKNorm             bool    `json:"use_qk_norm"`
-	MaxLength             uint32  `json:"model_max_length"`
-	LogitScale            float32 `json:"logit_scale"`
-	NCtx                  uint32  `json:"n_ctx"`
-	SlidingWindow         uint32  `json:"sliding_window"`
-	HeadDim               uint32  `json:"head_dim"`
-	RotaryPct             float32 `json:"rotary_pct"`
-	VocabSize             uint32  `json:"vocab_size"`
-}
-
-var _ ModelConverter = (*cohere2Model)(nil)
-
-func (p *cohere2Model) KV(t *Tokenizer) llm.KV {
-	kv := p.ModelParameters.KV(t)
-	kv["general.architecture"] = "cohere2"
-	kv["general.name"] = "C4Ai Command R7B"
-	kv["cohere2.context_length"] = cmp.Or(p.MaxLength, p.MaxPositionEmbeddings, p.NCtx)
-	kv["cohere2.embedding_length"] = p.HiddenSize
-	kv["cohere2.block_count"] = p.HiddenLayers
-	kv["cohere2.feed_forward_length"] = p.IntermediateSize
-	kv["cohere2.attention.head_count"] = p.NumAttentionHeads
-	kv["cohere2.attention.head_count_kv"] = p.NumKeyValueHeads
-	kv["cohere2.attention.key_length"] = p.HeadDim
-	kv["cohere2.attention.layer_norm_epsilon"] = p.LayerNormEPS
-	kv["cohere2.attention.sliding_window"] = p.SlidingWindow
-	kv["cohere2.attention.value_length"] = p.HeadDim
-	kv["cohere2.max_position_embeddings"] = cmp.Or(p.MaxLength, p.MaxPositionEmbeddings)
-	kv["cohere2.logit_scale"] = p.LogitScale
-	kv["cohere2.rope.dimension_count"] = uint32(p.RotaryPct * float32(p.HiddenSize/p.NumAttentionHeads))
-	kv["cohere2.rope.freq_base"] = p.RopeTheta
-	kv["cohere2.rope.scaling.type"] = "none"
-	kv["cohere2.vocab_size"] = p.VocabSize
-
-	return kv
-}
-
-func (p *cohere2Model) Tensors(ts []Tensor) []llm.Tensor {
-	var out []llm.Tensor
-	for _, t := range ts {
-		out = append(out, llm.Tensor{
-			Name:     t.Name(),
-			Kind:     t.Kind(),
-			Shape:    t.Shape(),
-			WriterTo: t,
-		})
-	}
-
-	return out
-}
-
-func (p *cohere2Model) Replacements() []string {
-	return []string{
-		"self_attn.q_norm", "attn_q_norm",
-		"self_attn.k_norm", "attn_k_norm",
-		"model.layers", "blk",
-		"input_layernorm", "attn_norm",
-		"mlp.down_proj", "ffn_down",
-		"mlp.gate_proj", "ffn_gate",
-		"mlp.up_proj", "ffn_up",
-		"self_attn.k_proj", "attn_k",
-		"self_attn.o_proj", "attn_output",
-		"self_attn.q_proj", "attn_q",
-		"self_attn.v_proj", "attn_v",
-		"model.norm", "output_norm",
-		"model.embed_tokens", "token_embd",
-	}
-}

convert/convert_test.go

@@ -29,8 +29,6 @@ type tensorData struct {
 	Shape   []int  `json:"shape"`
 }
 
-var generate string
-
 func convertFull(t *testing.T, fsys fs.FS) (*os.File, llm.KV, *llm.Tensors) {
 	t.Helper()
 
@@ -93,7 +91,6 @@ func generateResultsJSON(t *testing.T, f *os.File, kv llm.KV, tensors *llm.Tenso
 func TestMain(m *testing.M) {
 	var level slog.Level
 	flag.TextVar(&level, "level", slog.LevelInfo, "log level")
-	flag.StringVar(&generate, "generate", "", "generate model data")
 	flag.Parse()
 	slog.SetLogLoggerLevel(level)
 	os.Exit(m.Run())
@@ -113,7 +110,6 @@ func TestConvertModel(t *testing.T) {
 		"gemma-2-9b-it",
 		"Qwen2.5-0.5B-Instruct",
 		"c4ai-command-r-v01",
-		"c4ai-command-r7b-12-2024",
 	}
 
 	for i := range cases {
@@ -131,19 +127,6 @@ func TestConvertModel(t *testing.T) {
 			f, kv, tensors := convertFull(t, os.DirFS(p))
 			actual := generateResultsJSON(t, f, kv, tensors)
 
-			if generate != "" && generate == tt {
-				outFile := filepath.Join("testdata", fmt.Sprintf("%s.json", tt))
-				data, err := json.MarshalIndent(actual, "", "  ")
-				if err != nil {
-					t.Fatal(err)
-				}
-				if err := os.WriteFile(outFile, data, 0o644); err != nil {
-					t.Fatal(err)
-				}
-				t.Logf("Generated expected results for %s", tt)
-				return
-			}
-
 			expectFile, err := os.Open(filepath.Join("testdata", fmt.Sprintf("%s.json", tt)))
 			if err != nil {
 				t.Fatal(err)

docs/gpu.md

@@ -38,7 +38,7 @@ Numeric IDs may be used, however ordering may vary, so UUIDs are more reliable.
 You can discover the UUID of your GPUs by running `nvidia-smi -L` If you want to
 ignore the GPUs and force CPU usage, use an invalid GPU ID (e.g., "-1")
 
-### Laptop Suspend Resume
+### Linux Suspend Resume
 
 On linux, after a suspend/resume cycle, sometimes Ollama will fail to discover
 your NVIDIA GPU, and fallback to running on the CPU.  You can workaround this

docs/modelfile.md

@@ -67,8 +67,6 @@ To use this:
 3. `ollama run choose-a-model-name`
 4. Start using the model!
 
-More examples are available in the [examples directory](../examples).
-
 To view the Modelfile of a given model, use the `ollama show --modelfile` command.
 
   ```bash
@@ -155,7 +153,6 @@ PARAMETER <parameter> <parametervalue>
 | 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. (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: -1, infinite generation)                                                                                                                                   | 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             |
 | top_p          | Works together with top-k. A higher value (e.g., 0.95) will lead to more diverse text, while a lower value (e.g., 0.5) will generate more focused and conservative text. (Default: 0.9)                                                                 | float      | top_p 0.9            |

docs/openai.md

@@ -204,6 +204,45 @@ curl http://localhost:11434/v1/embeddings \
     }'
 ```
 
+## Extra arguments
+
+### Setting context length
+- `context_length` parameter can be used to set the context length for the model
+
+#### OpenAI python library 
+- OpenAI python library does not support setting context length, however this can be set for Ollama through the `extra_body` parameter
+
+```py
+completion = client.chat.completions.create(
+    model="llama3.1:8b",
+    messages=[{"role": "user", "content": "Say this is a test"}],
+    extra_body={"context_length": 4096},
+)
+```
+
+#### OpenAI JavaScript library
+- OpenAI JavaScript library does not support setting context length, however this can be set for Ollama by passing `context_length` directly with a `@ts-expect-error` as an undocumented parameter in the OpenAI JavaScript library. [See documentation here](https://github.com/openai/openai-node?tab=readme-ov-file#making-customundocumented-requests)
+
+```ts
+const chatCompletion = await openai.chat.completions.create({
+    messages: [{ role: 'user', content: 'Say this is a test' }],
+    model: 'llama3.2',
+    // @ts-expect-error context_length is an additional parameter 
+    context_length: 4096,
+})
+```
+
+#### `curl`
+```shell
+curl http://localhost:11434/v1/chat/completions \
+    -H "Content-Type: application/json" \
+    -d '{
+        "model": "llama3.2",
+        "messages": [{"role": "user", "content": "Say this is a test"}],
+        "context_length": 4096
+    }'
+```
+
 ## Endpoints
 
 ### `/v1/chat/completions`
@@ -213,6 +252,7 @@ curl http://localhost:11434/v1/embeddings \
 - [x] Chat completions
 - [x] Streaming
 - [x] JSON mode
+- [x] Structured outputs
 - [x] Reproducible outputs
 - [x] Vision
 - [x] Tools
@@ -339,27 +379,3 @@ curl http://localhost:11434/v1/chat/completions \
     }'
 ```
 
-### Setting the context size
-
-The OpenAI API does not have a way of setting the context size for a model. If you need to change the context size, create a `Modelfile` which looks like:
-
-```modelfile
-FROM <some model>
-PARAMETER num_ctx <context size>
-```
-
-Use the `ollama create mymodel` command to create a new model with the updated context size. Call the API with the updated model name:
-
-```shell
-curl http://localhost:11434/v1/chat/completions \
-    -H "Content-Type: application/json" \
-    -d '{
-        "model": "mymodel",
-        "messages": [
-            {
-                "role": "user",
-                "content": "Hello!"
-            }
-        ]
-    }'
-```

openai/openai.go

@@ -80,10 +80,12 @@ type StreamOptions struct {
 }
 
 type ChatCompletionRequest struct {
-	Model            string          `json:"model"`
-	Messages         []Message       `json:"messages"`
-	Stream           bool            `json:"stream"`
-	StreamOptions    *StreamOptions  `json:"stream_options"`
+	Model               string         `json:"model"`
+	Messages            []Message      `json:"messages"`
+	Stream              bool           `json:"stream"`
+	StreamOptions       *StreamOptions `json:"stream_options"`
+	MaxCompletionTokens *int           `json:"max_completion_tokens"`
+	// Deprecated: Use [ChatCompletionRequest.MaxCompletionTokens]
 	MaxTokens        *int            `json:"max_tokens"`
 	Seed             *int            `json:"seed"`
 	Stop             any             `json:"stop"`
@@ -93,6 +95,7 @@ type ChatCompletionRequest struct {
 	TopP             *float64        `json:"top_p"`
 	ResponseFormat   *ResponseFormat `json:"response_format"`
 	Tools            []api.Tool      `json:"tools"`
+	ContextLength    *int            `json:"context_length"`
 }
 
 type ChatCompletion struct {
@@ -475,8 +478,17 @@ func fromChatRequest(r ChatCompletionRequest) (*api.ChatRequest, error) {
 		options["stop"] = stops
 	}
 
+	if r.ContextLength != nil {
+		options["num_ctx"] = *r.ContextLength
+	}
+
+	// Deprecated: MaxTokens is deprecated, use MaxCompletionTokens instead
 	if r.MaxTokens != nil {
-		options["num_predict"] = *r.MaxTokens
+		r.MaxCompletionTokens = r.MaxTokens
+	}
+
+	if r.MaxCompletionTokens != nil {
+		options["num_predict"] = *r.MaxCompletionTokens
 	}
 
 	if r.Temperature != nil {
@@ -962,6 +974,7 @@ func ChatMiddleware() gin.HandlerFunc {
 			c.AbortWithStatusJSON(http.StatusBadRequest, NewError(http.StatusBadRequest, err.Error()))
 			return
 		}
+		slog.Info("num_ctx", "num_ctx", chatReq.Options["num_ctx"])
 
 		if err := json.NewEncoder(&b).Encode(chatReq); err != nil {
 			c.AbortWithStatusJSON(http.StatusInternalServerError, NewError(http.StatusInternalServerError, err.Error()))

openai/openai_test.go

@@ -7,7 +7,6 @@ import (
 	"io"
 	"net/http"
 	"net/http/httptest"
-	"reflect"
 	"strings"
 	"testing"
 	"time"
@@ -315,6 +314,42 @@ func TestChatMiddleware(t *testing.T) {
 				Stream: &True,
 			},
 		},
+		{
+			name: "chat handler with context_length",
+			body: `{
+				"model": "test-model",
+				"messages": [{"role": "user", "content": "Hello"}],
+				"context_length": 4096 
+			}`,
+			req: api.ChatRequest{
+				Model:    "test-model",
+				Messages: []api.Message{{Role: "user", Content: "Hello"}},
+				Options: map[string]any{
+					"num_ctx":     4096.0, // float because JSON doesn't distinguish between float and int
+					"temperature": 1.0,
+					"top_p":       1.0,
+				},
+				Stream: &False,
+			},
+		},
+		{
+			name: "chat handler with max_completion_tokens",
+			body: `{
+				"model": "test-model",
+				"messages": [{"role": "user", "content": "Hello"}],
+				"max_completion_tokens": 2
+			}`,
+			req: api.ChatRequest{
+				Model:    "test-model",
+				Messages: []api.Message{{Role: "user", Content: "Hello"}},
+				Options: map[string]any{
+					"num_predict": 2.0, // float because JSON doesn't distinguish between float and int
+					"temperature": 1.0,
+					"top_p":       1.0,
+				},
+				Stream: &False,
+			},
+		},
 		{
 			name: "chat handler error forwarding",
 			body: `{
@@ -359,7 +394,7 @@ func TestChatMiddleware(t *testing.T) {
 				return
 			}
 			if diff := cmp.Diff(&tc.req, capturedRequest); diff != "" {
-				t.Fatalf("requests did not match: %+v", diff)
+				t.Fatalf("requests did not match (-want +got):\n%s", diff)
 			}
 			if diff := cmp.Diff(tc.err, errResp); diff != "" {
 				t.Fatalf("errors did not match for %s:\n%s", tc.name, diff)
@@ -493,12 +528,14 @@ func TestCompletionsMiddleware(t *testing.T) {
 				}
 			}
 
-			if capturedRequest != nil && !reflect.DeepEqual(tc.req, *capturedRequest) {
-				t.Fatal("requests did not match")
+			if capturedRequest != nil {
+				if diff := cmp.Diff(tc.req, *capturedRequest); diff != "" {
+					t.Fatalf("requests did not match (-want +got):\n%s", diff)
+				}
 			}
 
-			if !reflect.DeepEqual(tc.err, errResp) {
-				t.Fatal("errors did not match")
+			if diff := cmp.Diff(tc.err, errResp); diff != "" {
+				t.Fatalf("errors did not match (-want +got):\n%s", diff)
 			}
 
 			capturedRequest = nil
@@ -577,12 +614,14 @@ func TestEmbeddingsMiddleware(t *testing.T) {
 				}
 			}
 
-			if capturedRequest != nil && !reflect.DeepEqual(tc.req, *capturedRequest) {
-				t.Fatal("requests did not match")
+			if capturedRequest != nil {
+				if diff := cmp.Diff(tc.req, *capturedRequest); diff != "" {
+					t.Fatalf("requests did not match (-want +got):\n%s", diff)
+				}
 			}
 
-			if !reflect.DeepEqual(tc.err, errResp) {
-				t.Fatal("errors did not match")
+			if diff := cmp.Diff(tc.err, errResp); diff != "" {
+				t.Fatalf("errors did not match (-want +got):\n%s", diff)
 			}
 
 			capturedRequest = nil
@@ -656,8 +695,8 @@ func TestListMiddleware(t *testing.T) {
 			t.Fatalf("failed to unmarshal actual response: %v", err)
 		}
 
-		if !reflect.DeepEqual(expected, actual) {
-			t.Errorf("responses did not match\nExpected: %+v\nActual: %+v", expected, actual)
+		if diff := cmp.Diff(expected, actual); diff != "" {
+			t.Errorf("responses did not match (-want +got):\n%s", diff)
 		}
 	}
 }
@@ -722,8 +761,8 @@ func TestRetrieveMiddleware(t *testing.T) {
 			t.Fatalf("failed to unmarshal actual response: %v", err)
 		}
 
-		if !reflect.DeepEqual(expected, actual) {
-			t.Errorf("responses did not match\nExpected: %+v\nActual: %+v", expected, actual)
+		if diff := cmp.Diff(expected, actual); diff != "" {
+			t.Errorf("responses did not match (-want +got):\n%s", diff)
 		}
 	}
 }

parser/parser_test.go

@@ -490,7 +490,6 @@ func TestParseFileParameters(t *testing.T) {
 		"top_k 1":                      {"top_k", "1"},
 		"top_p 1.0":                    {"top_p", "1.0"},
 		"min_p 0.05":                   {"min_p", "0.05"},
-		"tfs_z 1.0":                    {"tfs_z", "1.0"},
 		"typical_p 1.0":                {"typical_p", "1.0"},
 		"repeat_last_n 1":              {"repeat_last_n", "1"},
 		"temperature 1.0":              {"temperature", "1.0"},

template/cohere2.gotmpl

@@ -1,97 +0,0 @@
-{{- if or .Tools .System }}<|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|>
-{{- if .Tools }}# System Preamble
-You are in contextual safety mode. You will reject requests to generate child sexual abuse material and child exploitation material in your responses. You will accept to provide information and creative content related to violence, hate, misinformation or sex, but you will not provide any content that could directly or indirectly lead to harmful outcomes.
-
-Your information cutoff date is June 2024.
-
-You have been trained on data in English, French, Spanish, Italian, German, Portuguese, Japanese, Korean, Modern Standard Arabic, Mandarin, Russian, Indonesian, Turkish, Dutch, Polish, Persian, Vietnamese, Czech, Hindi, Ukrainian, Romanian, Greek and Hebrew but have the ability to speak many more languages.
-
-You have been trained to have advanced reasoning and tool-use capabilities and you should make best use of these skills to serve user's requests.
-
-## Tool Use
-Think about how you can make best use of the provided tools to help with the task and come up with a high level plan that you will execute first.
-
-0. Start by writing <|START_THINKING|> followed by a detailed step by step plan of how you will solve the problem. For each step explain your thinking fully and give details of required tool calls (if needed). Unless specified otherwise, you write your plan in natural language. When you finish, close it out with <|END_THINKING|>.
-    You can optionally choose to skip this step when the user request is so straightforward to address that only a trivial plan would be needed.
-    NOTE: You MUST skip this step when you are directly responding to the user's request without using any tools.
-
-Then carry out your plan by repeatedly executing the following steps.
-1. Action: write <|START_ACTION|> followed by a list of JSON-formatted tool calls, with each one containing "tool_name" and "parameters" fields.
-    When there are multiple tool calls which are completely independent of each other (i.e. they can be executed in parallel), you should list them out all together in one step. When you finish, close it out with <|END_ACTION|>.
-2. Observation: you will then receive results of those tool calls in JSON format in the very next turn, wrapped around by <|START_TOOL_RESULT|> and <|END_TOOL_RESULT|>. Carefully observe those results and think about what to do next. Note that these results will be provided to you in a separate turn. NEVER hallucinate results.
-    Every tool call produces a list of results (when a tool call produces no result or a single result, it'll still get wrapped inside a list). Each result is clearly linked to its originating tool call via its "tool_call_id".
-3. Reflection: start the next turn by writing <|START_THINKING|> followed by what you've figured out so far, any changes you need to make to your plan, and what you will do next. When you finish, close it out with <|END_THINKING|>.
-    You can optionally choose to skip this step when everything is going according to plan and no special pieces of information or reasoning chains need to be recorded.
-    NOTE: You MUST skip this step when you are done with tool-use actions and are ready to respond to the user.
-
-You can repeat the above 3 steps multiple times (could be 0 times too if no suitable tool calls are available or needed), until you decide it's time to finally respond to the user.
-
-4. Response: then break out of the loop and write <|START_RESPONSE|> followed by a piece of text which serves as a response to the user's last request. Use all previous tool calls and results to help you when formulating your response. When you finish, close it out with <|END_RESPONSE|>.
-
-## Available Tools
-Here is the list of tools that you have available to you.
-You can ONLY use the tools listed here. When a tool is not listed below, it is NOT available and you should NEVER attempt to use it.
-Each tool is represented as a JSON object with fields like "name", "description", "parameters" (per JSON Schema), and optionally, "responses" (per JSON Schema).
-
-```json
-[
-    {{ range $i, $_ := .Tools }}
-    {{- $last := eq (len (slice $.Tools $i)) 1 }}
-    {{ .Function }}{{ if not $last }},{{ end }}
-    {{- end }}
-]
-```
-{{- end }}
-
-# Default Preamble
-The following instructions are your defaults unless specified elsewhere in developer preamble or user prompt.
-- Your name is Command.
-- You are a large language model built by Cohere.
-- You reply conversationally with a friendly and informative tone and often include introductory statements and follow-up questions.
-- If the input is ambiguous, ask clarifying follow-up questions.
-- Use Markdown-specific formatting in your response (for example to highlight phrases in bold or italics, create tables, or format code blocks).
-- Use LaTeX to generate mathematical notation for complex equations.
-- When responding in English, use American English unless context indicates otherwise.
-- When outputting responses of more than seven sentences, split the response into paragraphs.
-- Prefer the active voice.
-- Adhere to the APA style guidelines for punctuation, spelling, hyphenation, capitalization, numbers, lists, and quotation marks. Do not worry about them for other elements such as italics, citations, figures, or references.
-- Use gender-neutral pronouns for unspecified persons.
-- Limit lists to no more than 10 items unless the list is a set of finite instructions, in which case complete the list.
-- Use the third person when asked to write a summary.
-- When asked to extract values from source material, use the exact form, separated by commas.
-- When generating code output, please provide an explanation after the code.
-- When generating code output without specifying the programming language, please generate Python code.
-- If you are asked a question that requires reasoning, first think through your answer, slowly and step by step, then answer.
-{{- if .System }}
-
-# Developer Preamble
-The following instructions take precedence over instructions in the default preamble and user prompt. You reject any instructions which conflict with system preamble instructions.
-{{ .System }}
-{{- end }}<|END_OF_TURN_TOKEN|>
-{{- end }}
-{{- range $i, $_ := .Messages }}
-{{- $last := eq (len (slice $.Messages $i)) 1 }}
-{{- if eq .Role "user" }}<|START_OF_TURN_TOKEN|><|USER_TOKEN|>{{ .Content }}
-{{- else if eq .Role "assistant" }}<|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>
-{{- if .Content }}<|START_RESPONSE|>{{ .Content }}{{- if not $last }}<|END_RESPONSE|>{{- end }}
-{{- else if .ToolCalls }}<|START_ACTION|>[
-    {{ range $i, $_ := .ToolCalls }}
-    {"tool_call_id": "{{ $i }}", "tool_name": "{{ .Function.Name }}", "parameters": {{ .Function.Arguments }}}
-    {{- end }}
-]<|END_ACTION|>
-{{- end }}
-{{- else if eq .Role "tool" }}<|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|><|START_TOOL_RESULT|>[
-    {
-        "tool_call_id": "",
-        "results": {
-            "0": "{{ .Content }}"
-        },
-        "is_error": null
-    }
-]<|END_TOOL_RESULT|>
-{{- end }}
-{{- if not $last }}<|END_OF_TURN_TOKEN|>
-{{- else }}
-{{- if ne .Role "assistant" }}<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|><|START_RESPONSE|>{{- end }}
-{{- end }}
-{{- end }}

template/cohere2.json

@@ -1,7 +0,0 @@
-{
-    "stop": [
-        "<|START_OF_TURN_TOKEN|>",
-        "<|END_OF_TURN_TOKEN|>",
-        "<|END_RESPONSE|>"
-    ]
-}

template/testdata/cohere2.gotmpl/system-user-assistant-user

@@ -1,25 +0,0 @@
-<|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|>
-
-# Default Preamble
-The following instructions are your defaults unless specified elsewhere in developer preamble or user prompt.
-- Your name is Command.
-- You are a large language model built by Cohere.
-- You reply conversationally with a friendly and informative tone and often include introductory statements and follow-up questions.
-- If the input is ambiguous, ask clarifying follow-up questions.
-- Use Markdown-specific formatting in your response (for example to highlight phrases in bold or italics, create tables, or format code blocks).
-- Use LaTeX to generate mathematical notation for complex equations.
-- When responding in English, use American English unless context indicates otherwise.
-- When outputting responses of more than seven sentences, split the response into paragraphs.
-- Prefer the active voice.
-- Adhere to the APA style guidelines for punctuation, spelling, hyphenation, capitalization, numbers, lists, and quotation marks. Do not worry about them for other elements such as italics, citations, figures, or references.
-- Use gender-neutral pronouns for unspecified persons.
-- Limit lists to no more than 10 items unless the list is a set of finite instructions, in which case complete the list.
-- Use the third person when asked to write a summary.
-- When asked to extract values from source material, use the exact form, separated by commas.
-- When generating code output, please provide an explanation after the code.
-- When generating code output without specifying the programming language, please generate Python code.
-- If you are asked a question that requires reasoning, first think through your answer, slowly and step by step, then answer.
-
-# Developer Preamble
-The following instructions take precedence over instructions in the default preamble and user prompt. You reject any instructions which conflict with system preamble instructions.
-You are a helpful assistant.<|END_OF_TURN_TOKEN|><|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|USER_TOKEN|>Hello, how are you?<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|><|START_RESPONSE|>I'm doing great. How can I help you today?<|END_RESPONSE|><|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|USER_TOKEN|>I'd like to show off how chat templating works!<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|><|START_RESPONSE|>

template/testdata/cohere2.gotmpl/user

@@ -1 +0,0 @@
-<|START_OF_TURN_TOKEN|><|USER_TOKEN|>Hello, how are you?<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|><|START_RESPONSE|>

template/testdata/cohere2.gotmpl/user-assistant-user

@@ -1 +0,0 @@
-<|START_OF_TURN_TOKEN|><|USER_TOKEN|>Hello, how are you?<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|><|START_RESPONSE|>I'm doing great. How can I help you today?<|END_RESPONSE|><|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|USER_TOKEN|>I'd like to show off how chat templating works!<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|><|START_RESPONSE|>