fix: handle ollama non-stream tool calls (#5865)

* fix: handle ollama non-stream tool calls

* test: cover ollama non-stream tool call paths

---------

Co-authored-by: CaIon <i@caion.me>

relay/channel/ollama/stream.go

@@ -9,6 +9,7 @@ import (
 	"time"
 
 	"github.com/QuantumNous/new-api/common"
+	"github.com/QuantumNous/new-api/constant"
 	"github.com/QuantumNous/new-api/dto"
 	"github.com/QuantumNous/new-api/logger"
 	relaycommon "github.com/QuantumNous/new-api/relay/common"
@@ -27,12 +28,7 @@ type ollamaChatStreamChunk struct {
 		Role      string           `json:"role"`
 		Content   string           `json:"content"`
 		Thinking  json.RawMessage  `json:"thinking"`
-		ToolCalls []struct {
-			Function struct {
-				Name      string      `json:"name"`
-				Arguments interface{} `json:"arguments"`
-			} `json:"function"`
-		} `json:"tool_calls"`
+		ToolCalls []OllamaToolCall `json:"tool_calls"`
 	} `json:"message"`
 	// generate
 	Response           string `json:"response"`
@@ -46,6 +42,39 @@ type ollamaChatStreamChunk struct {
 	EvalDuration       int64  `json:"eval_duration"`
 }
 
+func ollamaToolCallsToOpenAI(toolCalls []OllamaToolCall, startIndex int, includeIndex bool) ([]dto.ToolCallResponse, int) {
+	if len(toolCalls) == 0 {
+		return nil, startIndex
+	}
+	result := make([]dto.ToolCallResponse, 0, len(toolCalls))
+	for _, tc := range toolCalls {
+		var argBytes []byte
+		var err error
+		if tc.Function.Arguments == nil {
+			argBytes = []byte("{}")
+		} else {
+			argBytes, err = common.Marshal(tc.Function.Arguments)
+			if err != nil || len(argBytes) == 0 {
+				argBytes = []byte("{}")
+			}
+		}
+		tr := dto.ToolCallResponse{
+			ID:   fmt.Sprintf("call_%d", startIndex),
+			Type: "function",
+			Function: dto.FunctionResponse{
+				Name:      tc.Function.Name,
+				Arguments: string(argBytes),
+			},
+		}
+		if includeIndex {
+			tr.SetIndex(startIndex)
+		}
+		startIndex++
+		result = append(result, tr)
+	}
+	return result, startIndex
+}
+
 func toUnix(ts string) int64 {
 	if ts == "" {
 		return time.Now().Unix()
@@ -87,7 +116,7 @@ func ollamaStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http
 			continue
 		}
 		var chunk ollamaChatStreamChunk
-		if err := json.Unmarshal([]byte(line), &chunk); err != nil {
+		if err := common.Unmarshal([]byte(line), &chunk); err != nil {
 			logger.LogError(c, "ollama stream json decode error: "+err.Error()+" line="+line)
 			return usage, types.NewOpenAIError(err, types.ErrorCodeBadResponseBody, http.StatusInternalServerError)
 		}
@@ -122,7 +151,7 @@ func ollamaStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http
 				if raw != "" && raw != "null" {
 					// Unmarshal the JSON string to get the actual content without quotes
 					var thinkingContent string
-					if err := json.Unmarshal(chunk.Message.Thinking, &thinkingContent); err == nil {
+					if err := common.Unmarshal(chunk.Message.Thinking, &thinkingContent); err == nil {
 						delta.Choices[0].Delta.SetReasoningContent(thinkingContent)
 					} else {
 						// Fallback to raw string if it's not a JSON string
@@ -132,16 +161,7 @@ func ollamaStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http
 			}
 			// tool calls
 			if chunk.Message != nil && len(chunk.Message.ToolCalls) > 0 {
-				delta.Choices[0].Delta.ToolCalls = make([]dto.ToolCallResponse, 0, len(chunk.Message.ToolCalls))
-				for _, tc := range chunk.Message.ToolCalls {
-					// arguments -> string
-					argBytes, _ := json.Marshal(tc.Function.Arguments)
-					toolId := fmt.Sprintf("call_%d", toolCallIndex)
-					tr := dto.ToolCallResponse{ID: toolId, Type: "function", Function: dto.FunctionResponse{Name: tc.Function.Name, Arguments: string(argBytes)}}
-					tr.SetIndex(toolCallIndex)
-					toolCallIndex++
-					delta.Choices[0].Delta.ToolCalls = append(delta.Choices[0].Delta.ToolCalls, tr)
-				}
+				delta.Choices[0].Delta.ToolCalls, toolCallIndex = ollamaToolCallsToOpenAI(chunk.Message.ToolCalls, toolCallIndex, true)
 			}
 			if data, err := common.Marshal(delta); err == nil {
 				_ = helper.StringData(c, string(data))
@@ -157,6 +177,9 @@ func ollamaStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http
 		if finishReason == "" {
 			finishReason = "stop"
 		}
+		if toolCallIndex > 0 {
+			finishReason = constant.FinishReasonToolCalls
+		}
 		// emit stop delta
 		if stop := helper.GenerateStopResponse(responseId, created, model, finishReason); stop != nil {
 			if data, err := common.Marshal(stop); err == nil {
@@ -197,6 +220,8 @@ func ollamaChatHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.R
 		reasoningBuilder strings.Builder
 		lastChunk        ollamaChatStreamChunk
 		parsedAny        bool
+		toolCallIndex    int
+		toolCalls        []dto.ToolCallResponse
 	)
 	for _, ln := range lines {
 		ln = strings.TrimSpace(ln)
@@ -204,7 +229,7 @@ func ollamaChatHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.R
 			continue
 		}
 		var ck ollamaChatStreamChunk
-		if err := json.Unmarshal([]byte(ln), &ck); err != nil {
+		if err := common.Unmarshal([]byte(ln), &ck); err != nil {
 			if len(lines) == 1 {
 				return nil, types.NewOpenAIError(err, types.ErrorCodeBadResponseBody, http.StatusInternalServerError)
 			}
@@ -217,7 +242,7 @@ func ollamaChatHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.R
 			if raw != "" && raw != "null" {
 				// Unmarshal the JSON string to get the actual content without quotes
 				var thinkingContent string
-				if err := json.Unmarshal(ck.Message.Thinking, &thinkingContent); err == nil {
+				if err := common.Unmarshal(ck.Message.Thinking, &thinkingContent); err == nil {
 					reasoningBuilder.WriteString(thinkingContent)
 				} else {
 					// Fallback to raw string if it's not a JSON string
@@ -230,11 +255,16 @@ func ollamaChatHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.R
 		} else if ck.Response != "" {
 			aggContent.WriteString(ck.Response)
 		}
+		if ck.Message != nil && len(ck.Message.ToolCalls) > 0 {
+			var converted []dto.ToolCallResponse
+			converted, toolCallIndex = ollamaToolCallsToOpenAI(ck.Message.ToolCalls, toolCallIndex, false)
+			toolCalls = append(toolCalls, converted...)
+		}
 	}
 
 	if !parsedAny {
 		var single ollamaChatStreamChunk
-		if err := json.Unmarshal(body, &single); err != nil {
+		if err := common.Unmarshal(body, &single); err != nil {
 			return nil, types.NewOpenAIError(err, types.ErrorCodeBadResponseBody, http.StatusInternalServerError)
 		}
 		lastChunk = single
@@ -244,7 +274,7 @@ func ollamaChatHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.R
 				if raw != "" && raw != "null" {
 					// Unmarshal the JSON string to get the actual content without quotes
 					var thinkingContent string
-					if err := json.Unmarshal(single.Message.Thinking, &thinkingContent); err == nil {
+					if err := common.Unmarshal(single.Message.Thinking, &thinkingContent); err == nil {
 						reasoningBuilder.WriteString(thinkingContent)
 					} else {
 						// Fallback to raw string if it's not a JSON string
@@ -253,6 +283,11 @@ func ollamaChatHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.R
 				}
 			}
 			aggContent.WriteString(single.Message.Content)
+			if len(single.Message.ToolCalls) > 0 {
+				var converted []dto.ToolCallResponse
+				converted, toolCallIndex = ollamaToolCallsToOpenAI(single.Message.ToolCalls, toolCallIndex, false)
+				toolCalls = append(toolCalls, converted...)
+			}
 		} else {
 			aggContent.WriteString(single.Response)
 		}
@@ -269,8 +304,16 @@ func ollamaChatHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.R
 	if finishReason == "" {
 		finishReason = "stop"
 	}
+	if len(toolCalls) > 0 {
+		finishReason = constant.FinishReasonToolCalls
+	}
 
 	msg := dto.Message{Role: "assistant", Content: contentPtr(content)}
+	if len(toolCalls) > 0 {
+		if rawToolCalls, err := common.Marshal(toolCalls); err == nil {
+			msg.ToolCalls = rawToolCalls
+		}
+	}
 	if rc := reasoningBuilder.String(); rc != "" {
 		msg.ReasoningContent = &rc
 	}

relay/channel/ollama/stream_test.go

+package ollama
+
+import (
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"testing"
+
+	"github.com/QuantumNous/new-api/common"
+	"github.com/QuantumNous/new-api/constant"
+	"github.com/QuantumNous/new-api/dto"
+	relaycommon "github.com/QuantumNous/new-api/relay/common"
+
+	"github.com/gin-gonic/gin"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestOllamaChatHandlerNonStreamToolCalls(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+
+	tests := []struct {
+		name string
+		raw  string
+	}{
+		{
+			name: "compact json per-line parse path",
+			raw:  `{"model":"llama3.1","created_at":"2026-05-27T12:00:00Z","message":{"role":"assistant","content":"","tool_calls":[{"function":{"name":"get_weather","arguments":{"city":"Paris","days":0}}}]},"done":true,"done_reason":"stop","prompt_eval_count":5,"eval_count":7}`,
+		},
+		{
+			name: "pretty json fallback parse path",
+			raw: `{
+  "model": "llama3.1",
+  "created_at": "2026-05-27T12:00:00Z",
+  "message": {
+    "role": "assistant",
+    "content": "",
+    "tool_calls": [
+      {
+        "function": {
+          "name": "get_weather",
+          "arguments": {
+            "city": "Paris",
+            "days": 0
+          }
+        }
+      }
+    ]
+  },
+  "done": true,
+  "done_reason": "stop",
+  "prompt_eval_count": 5,
+  "eval_count": 7
+}`,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			w := httptest.NewRecorder()
+			c, _ := gin.CreateTestContext(w)
+
+			resp := &http.Response{
+				StatusCode: http.StatusOK,
+				Header:     make(http.Header),
+				Body:       io.NopCloser(strings.NewReader(tt.raw)),
+			}
+
+			usage, apiErr := ollamaChatHandler(c, &relaycommon.RelayInfo{
+				ChannelMeta: &relaycommon.ChannelMeta{UpstreamModelName: "fallback-model"},
+			}, resp)
+			require.Nil(t, apiErr)
+			require.NotNil(t, usage)
+			assert.Equal(t, 12, usage.TotalTokens)
+
+			var out dto.OpenAITextResponse
+			require.NoError(t, common.Unmarshal(w.Body.Bytes(), &out))
+			require.Len(t, out.Choices, 1)
+			assert.Equal(t, constant.FinishReasonToolCalls, out.Choices[0].FinishReason)
+
+			var toolCalls []dto.ToolCallResponse
+			require.NoError(t, common.Unmarshal(out.Choices[0].Message.ToolCalls, &toolCalls))
+			require.Len(t, toolCalls, 1)
+			assert.NotEmpty(t, toolCalls[0].ID)
+			assert.Equal(t, "function", toolCalls[0].Type)
+			assert.Equal(t, "get_weather", toolCalls[0].Function.Name)
+			assert.Nil(t, toolCalls[0].Index)
+
+			var args map[string]any
+			require.NoError(t, common.Unmarshal([]byte(toolCalls[0].Function.Arguments), &args))
+			assert.Equal(t, "Paris", args["city"])
+			assert.Equal(t, float64(0), args["days"])
+		})
+	}
+}