zlei6/ragflow
1
0
Fork 0
mirror of https://github.com/infiniflow/ragflow.git synced 2026-07-31 21:13:49 +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

30
internal/entity/models/vllm.go
View File

@@ -199,7 +199,7 @@ func (z *VllmModel) Chat(modelName, message *string, apiConfig *APIConfig, chatM
}
// ChatWithMessages sends multiple messages with roles and returns response
func (z *VllmModel) ChatWithMessages(modelName string, apiConfig *APIConfig, messages []Message, chatModelConfig *ChatConfig) (*ChatResponse, error) {
func (z *VllmModel) ChatWithMessages(modelName string, messages []Message, apiConfig *APIConfig, chatModelConfig *ChatConfig) (*ChatResponse, error) {
if len(messages) == 0 {
return nil, fmt.Errorf("messages is empty")
}
@@ -332,27 +332,37 @@ func (z *VllmModel) ChatWithMessages(modelName string, apiConfig *APIConfig, mes
return chatResponse, nil
}
// ChatStreamlyWithSender sends a message and streams response via sender function (best performance, no channel)
func (z *VllmModel) ChatStreamlyWithSender(modelName, message *string, apiConfig *APIConfig, modelConfig *ChatConfig, sender func(*string, *string) error) error {
// ChatStreamlyWithSender sends messages and streams response via sender function (best performance, no channel)
func (z *VllmModel) ChatStreamlyWithSender(modelName string, messages []Message, apiConfig *APIConfig, modelConfig *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
}
url := fmt.Sprintf("%s/%s", z.BaseURL[region], z.URLSuffix.Chat)
// TODO figure out what's the point of these codes
modelType := strings.Split(*modelName, "-")[0]
modelType := strings.Split(modelName, "-")[0]
if modelType == "qwen" || modelType == "glm" {
url = fmt.Sprintf("%s/%s", z.BaseURL[region], z.URLSuffix.AsyncChat)
}
// Convert messages to API format (supporting multimodal content)
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": true,
"model": modelName,
"messages": apiMessages,
"stream": true,
}
if modelConfig.Stream != nil {