zlei6/ragflow
1
0
Fork 0
mirror of https://github.com/infiniflow/ragflow.git synced 2026-07-27 02:42:00 +08:00
Code Issues Packages Projects Releases Wiki Activity

Support stream for multimodal chat (#14537)

Browse Source 
### What problem does this PR solve?

Support stream for multimodal chat

### Type of change

- [x] Refactoring
This commit is contained in:
qinling0210
2026-04-30 19:33:57 +08:00
committed by GitHub
parent 5fd4579a2f
commit 12af73f2ca
15 changed files with 369 additions and 227 deletions
Download Patch File Download Diff File Expand all files Collapse all files

136
internal/entity/models/deepseek.go
View File

@@ -60,7 +60,7 @@ func (z *DeepSeekModel) Name() string {
return "deepseek"
}
func (z *DeepSeekModel) ChatWithMessages(modelName string, apiConfig *APIConfig, messages []Message, chatModelConfig *ChatConfig) (*ChatResponse, error) {
func (z *DeepSeekModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig) (*ChatResponse, error) {
if len(messages) == 0 {
return nil, fmt.Errorf("messages is empty")
}
@@ -227,8 +227,12 @@ func (z *DeepSeekModel) ChatWithMessages(modelName string, apiConfig *APIConfig,
return chatResponse, nil
}
// ChatStreamlyWithSender sends a message and streams response via sender function (best performance, no channel)
func (z *DeepSeekModel) ChatStreamlyWithSender(modelName, message *string, apiConfig *APIConfig, chatModelConfig *ChatConfig, sender func(*string, *string) error) error {
// ChatStreamlyWithSender sends messages and streams response via sender function (best performance, no channel)
func (z *DeepSeekModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig, sender func(*string, *string) error) error {
if len(messages) == 0 {
return fmt.Errorf("messages is empty")
}
var region = "default"
if apiConfig.Region != nil {
region = *apiConfig.Region
@@ -236,77 +240,83 @@ func (z *DeepSeekModel) ChatStreamlyWithSender(modelName, message *string, apiCo
url := fmt.Sprintf("%s/chat/completions", z.BaseURL[region])
// Convert messages to API format
apiMessages := make([]map[string]interface{}, len(messages))
for i, msg := range messages {
apiMessages[i] = map[string]interface{}{
"role": msg.Role,
"content": msg.Content,
}
}
// Build request body with streaming enabled
reqBody := map[string]interface{}{
"model": modelName,
"messages": []map[string]string{
{"role": "user", "content": *message},
},
"stream": false,
"model": modelName,
"messages": apiMessages,
"stream": true,
"temperature": 1,
}
if chatModelConfig.Stream != nil {
reqBody["stream"] = *chatModelConfig.Stream
}
if chatModelConfig != nil {
if chatModelConfig.Stream != nil {
reqBody["stream"] = *chatModelConfig.Stream
}
if chatModelConfig.MaxTokens != nil {
reqBody["max_tokens"] = *chatModelConfig.MaxTokens
}
if chatModelConfig.MaxTokens != nil {
reqBody["max_tokens"] = *chatModelConfig.MaxTokens
}
if chatModelConfig.Temperature != nil {
reqBody["temperature"] = *chatModelConfig.Temperature
}
if chatModelConfig.Temperature != nil {
reqBody["temperature"] = *chatModelConfig.Temperature
}
if chatModelConfig.DoSample != nil {
reqBody["do_sample"] = *chatModelConfig.DoSample
}
if chatModelConfig.DoSample != nil {
reqBody["do_sample"] = *chatModelConfig.DoSample
}
if chatModelConfig.TopP != nil {
reqBody["top_p"] = *chatModelConfig.TopP
}
if chatModelConfig.TopP != nil {
reqBody["top_p"] = *chatModelConfig.TopP
}
if chatModelConfig.Stop != nil {
reqBody["stop"] = *chatModelConfig.Stop
}
if chatModelConfig.Stop != nil {
reqBody["stop"] = *chatModelConfig.Stop
}
if chatModelConfig.Thinking != nil {
if *chatModelConfig.Thinking {
var thinkingFlag string
switch *chatModelConfig.Effort {
case "none":
thinkingFlag = "disabled"
chatModelConfig.Thinking = nil
break
case "low":
thinkingFlag = "disabled"
chatModelConfig.Thinking = nil
break
case "medium":
thinkingFlag = "disabled"
chatModelConfig.Thinking = nil
break
case "high":
thinkingFlag = "enabled"
reqBody["reasoning_effort"] = "high"
break
case "default":
thinkingFlag = "enabled"
reqBody["reasoning_effort"] = "high"
break
case "max":
thinkingFlag = "enabled"
reqBody["reasoning_effort"] = "max"
break
default:
return fmt.Errorf("invalid effort level")
}
reqBody["thinking"] = map[string]interface{}{
"type": thinkingFlag,
}
} else {
reqBody["thinking"] = map[string]interface{}{
"type": "disabled",
if chatModelConfig.Thinking != nil {
if *chatModelConfig.Thinking {
var thinkingFlag string
switch *chatModelConfig.Effort {
case "none":
thinkingFlag = "disabled"
break
case "low":
thinkingFlag = "disabled"
break
case "medium":
thinkingFlag = "disabled"
break
case "high":
thinkingFlag = "enabled"
reqBody["reasoning_effort"] = "high"
break
case "default":
thinkingFlag = "enabled"
reqBody["reasoning_effort"] = "high"
break
case "max":
thinkingFlag = "enabled"
reqBody["reasoning_effort"] = "max"
break
default:
return fmt.Errorf("invalid effort level")
}
reqBody["thinking"] = map[string]interface{}{
"type": thinkingFlag,
}
} else {
reqBody["thinking"] = map[string]interface{}{
"type": "disabled",
}
}
}
}