Files
Enterprise-Knowledge-Base/internal/models/embedding/embedder.go
T
wizardchen fd3e2992b1 feat(models): support custom HTTP headers across all remote model calls
Introduce per-model `custom_headers` config (similar to OpenAI Python SDK's
`extra_headers`) so users can inject gateway auth tokens, trace IDs, etc.
into every outbound API request for chat / embedding / rerank / VLLM / ASR
models. Reserved headers like Authorization / Content-Type are always
preserved to avoid breaking auth or signing flows.

Along the way, unify production and "test connection" paths onto a single
`ConfigFromModel(*types.Model, appID, appSecret)` constructor in each
model package. Both `service.modelService.GetXxxModel` and the four
`handler.initialization.Check*Model` / `TestEmbeddingModel` endpoints now
go through the same field mapping, so new parameters only need to be
added in one place.

Backend:
- Add `CustomHeaders map[string]string` to `types.ModelParameters` with
  JSON/YAML `omitempty` tags for forward compatibility.
- New `internal/utils/extraheaders.go` providing `ApplyCustomHeaders`
  (for hand-rolled HTTP paths) and `WrapHTTPClientWithHeaders` /
  `CustomHeadersRoundTripper` (for SDK paths like go-openai). Reserved
  headers (Authorization, api-key, Content-Type, Accept, Host,
  Content-Length, User-Agent) are filtered out.
- Inject headers in every remote model path: chat (SDK + raw HTTP),
  embedding (openai/aliyun/jina/volcengine/nvidia/azure_openai), rerank
  (remote_api/aliyun/jina/nvidia/zhipu — via shared `customHeaderSetter`
  interface), VLM and ASR (via http.Client wrapping).
- Add `ConfigFromModel` to each of chat/embedding/rerank/vlm/asr,
  consolidating field mapping (ExtraConfig, CustomHeaders, InterfaceType
  defaulting, WeKnoraCloud credentials, etc.) and cover it with
  dedicated unit tests per package.
- Refactor `service/model.go`: replace ~100 lines of hand-written Config
  literals with one-line `ConfigFromModel` calls; drop the now-unused
  `stringMapToAnyMap` helper.
- Refactor `handler/initialization.go` test-connection endpoints: merge
  four ad-hoc request structs into one `ModelTestRequest`, extract
  `buildTestModel` and `resolveTenantWeKnoraCloudCreds` helpers, and
  route all four endpoints through `ConfigFromModel` + `NewXxx` so the
  test path is now behaviorally identical to production.

Frontend:
- Add `custom_headers` to the `ModelConfig` API type.
- `ModelEditorDialog.vue`: new Key-Value header editor (add/remove rows),
  auto-converts map <-> array when loading/saving, and passes the
  current header set to the `Test Connection` button so the preview
  reflects exactly what production will send.
- `ModelSettings.vue`: serialize the header array back into a map for
  the backend (drops empty rows).
- i18n: add custom-header labels / descriptions / placeholders to
  zh-CN, en-US, ru-RU, ko-KR.

Tests:
- `internal/utils/extraheaders_test.go` covers reserved-header filtering
  and round-tripper wrapping (including nil-client fallback).
- New `config_from_model_test.go` in chat / embedding / rerank / vlm /
  asr verifies all fields (CustomHeaders, ExtraConfig, InterfaceType
  defaulting, AppID/AppSecret) are propagated end-to-end.
2026-04-23 23:36:40 +08:00

236 lines
7.8 KiB
Go

package embedding
import (
"context"
"fmt"
"strings"
"github.com/Tencent/WeKnora/internal/logger"
"github.com/Tencent/WeKnora/internal/models/provider"
"github.com/Tencent/WeKnora/internal/models/utils/ollama"
"github.com/Tencent/WeKnora/internal/types"
)
// Embedder defines the interface for text vectorization
type Embedder interface {
// Embed converts text to vector
Embed(ctx context.Context, text string) ([]float32, error)
// BatchEmbed converts multiple texts to vectors in batch
BatchEmbed(ctx context.Context, texts []string) ([][]float32, error)
// GetModelName returns the model name
GetModelName() string
// GetDimensions returns the vector dimensions
GetDimensions() int
// GetModelID returns the model ID
GetModelID() string
EmbedderPooler
}
type EmbedderPooler interface {
BatchEmbedWithPool(ctx context.Context, model Embedder, texts []string) ([][]float32, error)
}
// EmbedderType represents the embedder type
type EmbedderType string
// Config represents the embedder configuration
type Config struct {
Source types.ModelSource `json:"source"`
BaseURL string `json:"base_url"`
ModelName string `json:"model_name"`
APIKey string `json:"api_key"`
TruncatePromptTokens int `json:"truncate_prompt_tokens"`
Dimensions int `json:"dimensions"`
ModelID string `json:"model_id"`
Provider string `json:"provider"`
ExtraConfig map[string]string `json:"extra_config"`
// CustomHeaders 允许在调用远程 API 时附加自定义 HTTP 请求头(类似 OpenAI Python SDK 的 extra_headers)。
CustomHeaders map[string]string `json:"custom_headers"`
AppID string
AppSecret string // 加密值,工厂函数调用方传入,使用前已解密
}
// ConfigFromModel 根据 types.Model 构造 embedding.Config。
// 生产路径(从 DB 拉起)和测试连接路径(临时表单)共享这份映射。
// appID / appSecret 是已解密的 WeKnoraCloud 凭证,调用方负责传入。
func ConfigFromModel(m *types.Model, appID, appSecret string) Config {
if m == nil {
return Config{}
}
return Config{
Source: m.Source,
BaseURL: m.Parameters.BaseURL,
APIKey: m.Parameters.APIKey,
ModelID: m.ID,
ModelName: m.Name,
Dimensions: m.Parameters.EmbeddingParameters.Dimension,
TruncatePromptTokens: m.Parameters.EmbeddingParameters.TruncatePromptTokens,
Provider: m.Parameters.Provider,
ExtraConfig: m.Parameters.ExtraConfig,
CustomHeaders: m.Parameters.CustomHeaders,
AppID: appID,
AppSecret: appSecret,
}
}
// NewEmbedder creates an embedder based on the configuration
func NewEmbedder(config Config, pooler EmbedderPooler, ollamaService *ollama.OllamaService) (Embedder, error) {
e, err := newEmbedder(config, pooler, ollamaService)
if err != nil || !logger.LLMDebugEnabled() {
return e, err
}
return &debugEmbedder{inner: e}, nil
}
func newEmbedder(config Config, pooler EmbedderPooler, ollamaService *ollama.OllamaService) (Embedder, error) {
var embedder Embedder
var err error
switch strings.ToLower(string(config.Source)) {
case string(types.ModelSourceLocal):
embedder, err = NewOllamaEmbedder(config.BaseURL,
config.ModelName, config.TruncatePromptTokens, config.Dimensions, config.ModelID, pooler, ollamaService)
return embedder, err
case string(types.ModelSourceRemote):
// Detect or use configured provider for routing
providerName := provider.ProviderName(config.Provider)
if providerName == "" {
providerName = provider.DetectProvider(config.BaseURL)
}
// Route to provider-specific embedders
switch providerName {
case provider.ProviderAliyun:
// 检查是否是多模态嵌入模型
// 多模态模型: tongyi-embedding-vision-*, multimodal-embedding-*
// tex-only模型: text-embedding-v1/v2/v3/v4 应该使用 OpenAI 兼容接口,否则响应格式不匹配、embedding 返回空数组
isMultimodalModel := strings.Contains(strings.ToLower(config.ModelName), "vision") ||
strings.Contains(strings.ToLower(config.ModelName), "multimodal")
if isMultimodalModel {
// 多模态模型需要使用DashScope专用 API 端点
// 如果用户填写了 OpenAI 兼容模式的 URL,自动修正为多模态 API 的baseURL
baseURL := config.BaseURL
if baseURL == "" {
baseURL = "https://dashscope.aliyuncs.com"
} else if strings.Contains(baseURL, "/compatible-mode/") {
// 移除 compatible-mode 路径,AliyunEmbedder 会自动添加多模态端点
baseURL = strings.Replace(baseURL, "/compatible-mode/v1", "", 1)
baseURL = strings.Replace(baseURL, "/compatible-mode", "", 1)
}
aliyunEmb, aErr := NewAliyunEmbedder(config.APIKey,
baseURL,
config.ModelName,
config.TruncatePromptTokens,
config.Dimensions,
config.ModelID,
pooler)
if aliyunEmb != nil {
aliyunEmb.SetCustomHeaders(config.CustomHeaders)
}
embedder, err = aliyunEmb, aErr
} else {
baseURL := config.BaseURL
if baseURL == "" || !strings.Contains(baseURL, "/compatible-mode/") {
baseURL = "https://dashscope.aliyuncs.com/compatible-mode/v1"
}
openaiEmb, oErr := NewOpenAIEmbedder(config.APIKey,
baseURL,
config.ModelName,
config.TruncatePromptTokens,
config.Dimensions,
config.ModelID,
pooler)
if openaiEmb != nil {
openaiEmb.SetCustomHeaders(config.CustomHeaders)
}
embedder, err = openaiEmb, oErr
}
return embedder, err
case provider.ProviderVolcengine:
// Volcengine Ark uses multimodal embedding API
volcEmb, vErr := NewVolcengineEmbedder(config.APIKey,
config.BaseURL,
config.ModelName,
config.TruncatePromptTokens,
config.Dimensions,
config.ModelID,
pooler)
if volcEmb != nil {
volcEmb.SetCustomHeaders(config.CustomHeaders)
}
embedder, err = volcEmb, vErr
return embedder, err
case provider.ProviderJina:
// Jina AI uses different API format (truncate instead of truncate_prompt_tokens)
jinaEmb, jErr := NewJinaEmbedder(config.APIKey,
config.BaseURL,
config.ModelName,
config.TruncatePromptTokens,
config.Dimensions,
config.ModelID,
pooler)
if jinaEmb != nil {
jinaEmb.SetCustomHeaders(config.CustomHeaders)
}
embedder, err = jinaEmb, jErr
return embedder, err
case provider.ProviderAzureOpenAI:
apiVersion := "2024-10-21"
if config.ExtraConfig != nil {
if v, ok := config.ExtraConfig["api_version"]; ok {
apiVersion = v
}
}
azureEmb, azErr := NewAzureOpenAIEmbedder(config.APIKey,
config.BaseURL,
config.ModelName,
config.TruncatePromptTokens,
config.Dimensions,
config.ModelID,
apiVersion,
pooler)
if azureEmb != nil {
azureEmb.SetCustomHeaders(config.CustomHeaders)
}
embedder, err = azureEmb, azErr
return embedder, err
case provider.ProviderNvidia:
nvEmb, nErr := NewNvidiaEmbedder(config.APIKey,
config.BaseURL,
config.ModelName,
config.Dimensions,
config.ModelID,
pooler)
if nvEmb != nil {
nvEmb.SetCustomHeaders(config.CustomHeaders)
}
embedder, err = nvEmb, nErr
return embedder, err
case provider.ProviderWeKnoraCloud:
embedder, err = NewWeKnoraCloudEmbedder(config)
return embedder, err
default:
// Use OpenAI-compatible embedder for other providers
openaiEmb, oErr := NewOpenAIEmbedder(config.APIKey,
config.BaseURL,
config.ModelName,
config.TruncatePromptTokens,
config.Dimensions,
config.ModelID,
pooler)
if openaiEmb != nil {
openaiEmb.SetCustomHeaders(config.CustomHeaders)
}
embedder, err = openaiEmb, oErr
return embedder, err
}
default:
return nil, fmt.Errorf("unsupported embedder source: %s", config.Source)
}
}