ben/coze-studio
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

fix: replace workflow hardcoding chat model for knowledge recall (#233)

Browse Source
This commit is contained in:
N3ko
2025-07-30 14:27:01 +08:00
committed by GitHub
parent 5013a9ed53
commit b9d03b148c
12 changed files with 218 additions and 123 deletions
Download Patch File Download Diff File Expand all files Collapse all files

91
backend/application/knowledge/init.go
View File

@@ -28,25 +28,18 @@ import (
"github.com/cloudwego/eino-ext/components/embedding/ark"
ollamaEmb "github.com/cloudwego/eino-ext/components/embedding/ollama"
"github.com/cloudwego/eino-ext/components/embedding/openai"
ao "github.com/cloudwego/eino-ext/components/model/ark"
"github.com/cloudwego/eino-ext/components/model/deepseek"
"github.com/cloudwego/eino-ext/components/model/gemini"
"github.com/cloudwego/eino-ext/components/model/ollama"
mo "github.com/cloudwego/eino-ext/components/model/openai"
"github.com/cloudwego/eino-ext/components/model/qwen"
"github.com/cloudwego/eino/components/prompt"
"github.com/cloudwego/eino/schema"
"github.com/coze-dev/coze-studio/backend/application/internal"
"github.com/coze-dev/coze-studio/backend/infra/impl/embedding/http"
"github.com/milvus-io/milvus/client/v2/milvusclient"
"github.com/volcengine/volc-sdk-golang/service/vikingdb"
"github.com/volcengine/volc-sdk-golang/service/visual"
"google.golang.org/genai"
"gorm.io/gorm"
"github.com/coze-dev/coze-studio/backend/application/search"
knowledgeImpl "github.com/coze-dev/coze-studio/backend/domain/knowledge/service"
"github.com/coze-dev/coze-studio/backend/infra/contract/cache"
"github.com/coze-dev/coze-studio/backend/infra/contract/chatmodel"
"github.com/coze-dev/coze-studio/backend/infra/contract/document/nl2sql"
"github.com/coze-dev/coze-studio/backend/infra/contract/document/ocr"
"github.com/coze-dev/coze-studio/backend/infra/contract/document/searchstore"
@@ -131,7 +124,7 @@ func InitService(c *ServiceComponents) (*KnowledgeApplicationService, error) {
}
var rewriter messages2query.MessagesToQuery
if rewriterChatModel, _, err := getBuiltinChatModel(ctx, "M2Q_"); err != nil {
if rewriterChatModel, _, err := internal.GetBuiltinChatModel(ctx, "M2Q_"); err != nil {
return nil, err
} else {
filePath := filepath.Join(root, "resources/conf/prompt/messages_to_query_template_jinja2.json")
@@ -146,7 +139,7 @@ func InitService(c *ServiceComponents) (*KnowledgeApplicationService, error) {
}
var n2s nl2sql.NL2SQL
if n2sChatModel, _, err := getBuiltinChatModel(ctx, "NL2SQL_"); err != nil {
if n2sChatModel, _, err := internal.GetBuiltinChatModel(ctx, "NL2SQL_"); err != nil {
return nil, err
} else {
filePath := filepath.Join(root, "resources/conf/prompt/nl2sql_template_jinja2.json")
@@ -160,7 +153,7 @@ func InitService(c *ServiceComponents) (*KnowledgeApplicationService, error) {
}
}
imageAnnoChatModel, configured, err := getBuiltinChatModel(ctx, "IA_")
imageAnnoChatModel, configured, err := internal.GetBuiltinChatModel(ctx, "IA_")
if err != nil {
return nil, err
}
@@ -372,7 +365,6 @@ func getEmbedding(ctx context.Context) (embedding.Embedder, error) {
return nil, fmt.Errorf("init ollama embedding failed, err=%w", err)
}
case "http":
var (
httpEmbeddingBaseURL = os.Getenv("HTTP_EMBEDDING_ADDR")
@@ -394,81 +386,6 @@ func getEmbedding(ctx context.Context) (embedding.Embedder, error) {
return emb, nil
}
func getBuiltinChatModel(ctx context.Context, envPrefix string) (bcm chatmodel.BaseChatModel, configured bool, err error) {
getEnv := func(key string) string {
if val := os.Getenv(envPrefix + key); val != "" {
return val
}
return os.Getenv(key)
}
switch getEnv("BUILTIN_CM_TYPE") {
case "openai":
byAzure, _ := strconv.ParseBool(getEnv("BUILTIN_CM_OPENAI_BY_AZURE"))
bcm, err = mo.NewChatModel(ctx, &mo.ChatModelConfig{
APIKey: getEnv("BUILTIN_CM_OPENAI_API_KEY"),
ByAzure: byAzure,
BaseURL: getEnv("BUILTIN_CM_OPENAI_BASE_URL"),
Model: getEnv("BUILTIN_CM_OPENAI_MODEL"),
})
case "ark":
bcm, err = ao.NewChatModel(ctx, &ao.ChatModelConfig{
APIKey: getEnv("BUILTIN_CM_ARK_API_KEY"),
Model: getEnv("BUILTIN_CM_ARK_MODEL"),
BaseURL: getEnv("BUILTIN_CM_ARK_BASE_URL"),
})
case "deepseek":
bcm, err = deepseek.NewChatModel(ctx, &deepseek.ChatModelConfig{
APIKey: getEnv("BUILTIN_CM_DEEPSEEK_API_KEY"),
BaseURL: getEnv("BUILTIN_CM_DEEPSEEK_BASE_URL"),
Model: getEnv("BUILTIN_CM_DEEPSEEK_MODEL"),
})
case "ollama":
bcm, err = ollama.NewChatModel(ctx, &ollama.ChatModelConfig{
BaseURL: getEnv("BUILTIN_CM_OLLAMA_BASE_URL"),
Model: getEnv("BUILTIN_CM_OLLAMA_MODEL"),
})
case "qwen":
bcm, err = qwen.NewChatModel(ctx, &qwen.ChatModelConfig{
APIKey: getEnv("BUILTIN_CM_QWEN_API_KEY"),
BaseURL: getEnv("BUILTIN_CM_QWEN_BASE_URL"),
Model: getEnv("BUILTIN_CM_QWEN_MODEL"),
})
case "gemini":
backend, convErr := strconv.ParseInt(getEnv("BUILTIN_CM_GEMINI_BACKEND"), 10, 64)
if convErr != nil {
return nil, false, convErr
}
c, clientErr := genai.NewClient(ctx, &genai.ClientConfig{
APIKey: getEnv("BUILTIN_CM_GEMINI_API_KEY"),
Backend: genai.Backend(backend),
Project: getEnv("BUILTIN_CM_GEMINI_PROJECT"),
Location: getEnv("BUILTIN_CM_GEMINI_LOCATION"),
HTTPOptions: genai.HTTPOptions{
BaseURL: getEnv("BUILTIN_CM_GEMINI_BASE_URL"),
},
})
if clientErr != nil {
return nil, false, clientErr
}
bcm, err = gemini.NewChatModel(ctx, &gemini.Config{
Client: c,
Model: getEnv("BUILTIN_CM_GEMINI_MODEL"),
})
default:
// accept builtin chat model not configured
}
if err != nil {
return nil, false, fmt.Errorf("knowledge init openai chat mode failed, %w", err)
}
if bcm != nil {
configured = true
}
return
}
func readJinja2PromptTemplate(jsonFilePath string) (prompt.ChatTemplate, error) {
b, err := os.ReadFile(jsonFilePath)
if err != nil {