harmony: move harmony parsing into a package (#12016)

server/harmonyparser.go → harmony/harmonyparser.go

-package server
+package harmony
 
 import (
 	"context"
 	"fmt"
 	"log/slog"
-	"slices"
 	"strings"
 	"unicode"
 
@@ -20,18 +19,6 @@ const (
 	harmonyParserState_ParsingContent
 )
 
-func shouldUseHarmony(model Model) bool {
-	if slices.Contains([]string{"gptoss", "gpt-oss"}, model.Config.ModelFamily) {
-		// heuristic to check whether the template expects to be parsed via harmony:
-		// search for harmony tags that are nearly always used
-		if model.Template.Contains("<|start|>") && model.Template.Contains("<|end|>") {
-			return true
-		}
-	}
-
-	return false
-}
-
 func (s harmonyParserState) String() string {
 	switch s {
 	// we're looking for the message start tag
@@ -277,20 +264,20 @@ const (
 // This is a higher level interface that maps harmony concepts into ollama concepts
 type HarmonyMessageHandler struct {
 	state           harmonyMessageState
-	harmonyParser   *HarmonyParser
-	functionNameMap *FunctionNameMap
+	HarmonyParser   *HarmonyParser
+	FunctionNameMap *FunctionNameMap
 }
 
 // NewHarmonyMessageHandler creates a new message handler
 func NewHarmonyMessageHandler() *HarmonyMessageHandler {
 	return &HarmonyMessageHandler{
 		state: harmonyMessageState_Normal,
-		harmonyParser: &HarmonyParser{
+		HarmonyParser: &HarmonyParser{
 			MessageStartTag: "<|start|>",
 			MessageEndTag:   "<|end|>",
 			HeaderEndTag:    "<|message|>",
 		},
-		functionNameMap: NewFunctionNameMap(),
+		FunctionNameMap: NewFunctionNameMap(),
 	}
 }
 
@@ -301,7 +288,7 @@ func (h *HarmonyMessageHandler) AddContent(content string, toolParser *HarmonyTo
 	thinkingSb := strings.Builder{}
 	toolContentSb := strings.Builder{}
 
-	events := h.harmonyParser.AddContent(content)
+	events := h.HarmonyParser.AddContent(content)
 	for _, event := range events {
 		switch event := event.(type) {
 		case HarmonyEventHeaderComplete:

server/harmonyparser_test.go → harmony/harmonyparser_test.go

-package server
+package harmony
 
 import (
 	"fmt"

server/routes.go

@@ -32,6 +32,7 @@ import (
 	"github.com/ollama/ollama/envconfig"
 	"github.com/ollama/ollama/format"
 	"github.com/ollama/ollama/fs/ggml"
+	"github.com/ollama/ollama/harmony"
 	"github.com/ollama/ollama/llm"
 	"github.com/ollama/ollama/logutil"
 	"github.com/ollama/ollama/openai"
@@ -45,6 +46,18 @@ import (
 	"github.com/ollama/ollama/version"
 )
 
+func shouldUseHarmony(model *Model) bool {
+	if slices.Contains([]string{"gptoss", "gpt-oss"}, model.Config.ModelFamily) {
+		// heuristic to check whether the template expects to be parsed via harmony:
+		// search for harmony tags that are nearly always used
+		if model.Template.Contains("<|start|>") && model.Template.Contains("<|end|>") {
+			return true
+		}
+	}
+
+	return false
+}
+
 func experimentEnabled(name string) bool {
 	return slices.Contains(strings.Split(os.Getenv("OLLAMA_EXPERIMENT"), ","), name)
 }
@@ -194,12 +207,12 @@ func (s *Server) GenerateHandler(c *gin.Context) {
 		return
 	}
 
-	useHarmony := shouldUseHarmony(*m) && !req.Raw
-	var harmonyMessageHandler *HarmonyMessageHandler
-	var harmonyToolParser *HarmonyToolCallAccumulator
+	useHarmony := shouldUseHarmony(m) && !req.Raw
+	var harmonyMessageHandler *harmony.HarmonyMessageHandler
+	var harmonyToolParser *harmony.HarmonyToolCallAccumulator
 	if useHarmony {
-		harmonyMessageHandler = NewHarmonyMessageHandler()
-		harmonyMessageHandler.harmonyParser.AddImplicitStart()
+		harmonyMessageHandler = harmony.NewHarmonyMessageHandler()
+		harmonyMessageHandler.HarmonyParser.AddImplicitStart()
 		harmonyToolParser = harmonyMessageHandler.CreateToolParser()
 	}
 
@@ -1603,19 +1616,19 @@ func (s *Server) ChatHandler(c *gin.Context) {
 	}
 	msgs = filterThinkTags(msgs, m)
 
-	var harmonyMessageHandler *HarmonyMessageHandler
-	var harmonyToolParser *HarmonyToolCallAccumulator
+	var harmonyMessageHandler *harmony.HarmonyMessageHandler
+	var harmonyToolParser *harmony.HarmonyToolCallAccumulator
 
-	useHarmony := shouldUseHarmony(*m)
+	useHarmony := shouldUseHarmony(m)
 
 	processedTools := req.Tools
 	if useHarmony {
-		harmonyMessageHandler = NewHarmonyMessageHandler()
+		harmonyMessageHandler = harmony.NewHarmonyMessageHandler()
 		var lastMessage *api.Message
 		if len(msgs) > 0 {
 			lastMessage = &msgs[len(msgs)-1]
 		}
-		harmonyMessageHandler.harmonyParser.AddImplicitStartOrPrefill(lastMessage)
+		harmonyMessageHandler.HarmonyParser.AddImplicitStartOrPrefill(lastMessage)
 		harmonyToolParser = harmonyMessageHandler.CreateToolParser()
 
 		// make a copy of tools to pass to the chat prompt. Function names may be
@@ -1623,7 +1636,7 @@ func (s *Server) ChatHandler(c *gin.Context) {
 		processedTools = make([]api.Tool, len(req.Tools))
 		copy(processedTools, req.Tools)
 		for i, tool := range processedTools {
-			processedTools[i].Function.Name = harmonyMessageHandler.functionNameMap.ConvertAndAdd(tool.Function.Name)
+			processedTools[i].Function.Name = harmonyMessageHandler.FunctionNameMap.ConvertAndAdd(tool.Function.Name)
 		}
 	}
 
@@ -1705,7 +1718,7 @@ func (s *Server) ChatHandler(c *gin.Context) {
 					toolName, toolContent := harmonyToolParser.Drain()
 					if toolName != nil {
 						*toolName = strings.TrimPrefix(*toolName, "functions.")
-						*toolName = harmonyMessageHandler.functionNameMap.OriginalFromConverted(*toolName)
+						*toolName = harmonyMessageHandler.FunctionNameMap.OriginalFromConverted(*toolName)
 						var args api.ToolCallFunctionArguments
 						if err := json.Unmarshal([]byte(toolContent), &args); err != nil {
 							errStr := fmt.Sprintf("error parsing tool call: raw='%s', err=%s", toolContent, err.Error())