Open-source LLM knowledge platform: turn raw documents into a queryable RAG, an autonomous reasoning agent, and a self-maintaining Wiki.
Go agentagenticaichatbotdsh-pluginembeddingsevaluationgenerative-aigolangknowledge-basellmmulti-tenantollamaopenaiquestion-answeringragrerankingsemantic-searchvector-searchwiki
feat: add LiteLLM as a model provider (#2923)
Prodman Devokadevdb05a64caade
diff --git a/client/model.go b/client/model.goindex b3a1ef7f0..a6d67cc97 100644--- a/client/model.go+++ b/client/model.go@@ -48,6 +48,7 @@ const (ModelSourceSiliconFlow ModelSource = "siliconflow" // SiliconFlow modelModelSourceJina ModelSource = "jina" // Jina AI modelModelSourceOpenRouter ModelSource = "openrouter" // OpenRouter model+ ModelSourceLiteLLM ModelSource = "litellm" // LiteLLM proxy modelModelSourceRequesty ModelSource = "requesty" // Requesty modelModelSourceNvidia ModelSource = "nvidia" // NVIDIA modelModelSourceNovita ModelSource = "novita" // Novita AI model@@ -64,7 +65,8 @@ func AllModelSources() []ModelSource {ModelSourceLocal, ModelSourceRemote, ModelSourceAliyun, ModelSourceZhipu,ModelSourceVolcengine, ModelSourceDeepseek, ModelSourceHunyuan, ModelSourceMinimax,ModelSourceOpenAI, ModelSourceGemini, ModelSourceMimo, ModelSourceSiliconFlow,- ModelSourceJina, ModelSourceOpenRouter, ModelSourceRequesty, ModelSourceNvidia, ModelSourceNovita,+ ModelSourceJina, ModelSourceOpenRouter, ModelSourceLiteLLM, ModelSourceRequesty,+ ModelSourceNvidia, ModelSourceNovita,ModelSourceAzureOpenAI,}}diff --git a/docs/BUILTIN_MODELS.md b/docs/BUILTIN_MODELS.mdindex 6816cd8b9..b4fe9bba2 100644--- a/docs/BUILTIN_MODELS.md+++ b/docs/BUILTIN_MODELS.md@@ -167,7 +167,7 @@ services:### 方式二:直接 SQL 插入-支持的 provider:`generic`(自定义)、`openai`、`aliyun`、`zhipu`、`volcengine`、`hunyuan`、`deepseek`、`minimax`、`mimo`、`siliconflow`、`jina`、`openrouter`、`requesty`、`gemini`、`modelscope`、`moonshot`、`qianfan`、`qiniu`、`longcat`、`gpustack`+支持的 provider:`generic`(自定义)、`openai`、`aliyun`、`zhipu`、`volcengine`、`hunyuan`、`deepseek`、`minimax`、`mimo`、`siliconflow`、`jina`、`openrouter`、`litellm`、`requesty`、`gemini`、`modelscope`、`moonshot`、`qianfan`、`qiniu`、`longcat`、`gpustack````sql-- 示例:LLM 内置模型diff --git a/docs/api/model.md b/docs/api/model.mdindex 9024d161a..61f6a1652 100644--- a/docs/api/model.md+++ b/docs/api/model.md@@ -33,6 +33,7 @@ WeKnora 支持多种主流 AI 模型服务商,在创建模型时可通过 `par| `siliconflow` | 硅基流动 SiliconFlow | Chat, Embedding, Rerank, VLLM || `jina` | Jina | Embedding, Rerank || `openrouter` | OpenRouter | Chat, VLLM |+| `litellm` | LiteLLM (self-hosted proxy) | Chat, Embedding, VLLM || `requesty` | Requesty | Chat, Embedding, VLLM || `gemini` | Google Gemini | Chat || `modelscope` | 魔搭 ModelScope | Chat, Embedding, VLLM |@@ -43,6 +44,8 @@ WeKnora 支持多种主流 AI 模型服务商,在创建模型时可通过 `par| `gpustack` | GPUStack | Chat, Embedding, Rerank, VLLM |> 实际可用的服务商以 `GET /models/providers` 返回为准。+>+> `litellm` 的目录默认地址是占位符 `http://your_litellm_proxy/v1`,保存前请换成实际可解析的主机名。`localhost` / `127.0.0.1` / `host.docker.internal` 默认会被 SSRF 拦截;自托管时把该主机加入 `SSRF_WHITELIST`。创建模型时 `source` 仍为 `remote`,厂商写在 `parameters.provider`。## GET `/models/providers` - 获取模型服务商列表diff --git a/docs/docs.go b/docs/docs.goindex acc16b183..0e42773c2 100644--- a/docs/docs.go+++ b/docs/docs.go@@ -19908,6 +19908,7 @@ const docTemplate = `{"siliconflow","jina","openrouter",+ "litellm","requesty","nvidia","novita",@@ -19927,6 +19928,7 @@ const docTemplate = `{"ModelSourceNvidia": "NVIDIA model","ModelSourceOpenAI": "OpenAI model","ModelSourceOpenRouter": "OpenRouter model",+ "ModelSourceLiteLLM": "LiteLLM proxy model","ModelSourceRemote": "Remote model","ModelSourceRequesty": "Requesty model","ModelSourceSiliconFlow": "SiliconFlow model",@@ -19948,6 +19950,7 @@ const docTemplate = `{"SiliconFlow model","Jina AI model","OpenRouter model",+ "LiteLLM proxy model","Requesty model","NVIDIA model","Novita AI model",@@ -19968,6 +19971,7 @@ const docTemplate = `{"ModelSourceSiliconFlow","ModelSourceJina","ModelSourceOpenRouter",+ "ModelSourceLiteLLM","ModelSourceRequesty","ModelSourceNvidia","ModelSourceNovita",diff --git a/docs/swagger.json b/docs/swagger.jsonindex 46ee83e64..e7f9e01e6 100644--- a/docs/swagger.json+++ b/docs/swagger.json@@ -19901,6 +19901,7 @@"siliconflow","jina","openrouter",+ "litellm","requesty","nvidia","novita",@@ -19920,6 +19921,7 @@"ModelSourceNvidia": "NVIDIA model","ModelSourceOpenAI": "OpenAI model","ModelSourceOpenRouter": "OpenRouter model",+ "ModelSourceLiteLLM": "LiteLLM proxy model","ModelSourceRemote": "Remote model","ModelSourceRequesty": "Requesty model","ModelSourceSiliconFlow": "SiliconFlow model",@@ -19941,6 +19943,7 @@"SiliconFlow model","Jina AI model","OpenRouter model",+ "LiteLLM proxy model","Requesty model","NVIDIA model","Novita AI model",@@ -19961,6 +19964,7 @@"ModelSourceSiliconFlow","ModelSourceJina","ModelSourceOpenRouter",+ "ModelSourceLiteLLM","ModelSourceRequesty","ModelSourceNvidia","ModelSourceNovita",diff --git a/docs/swagger.yaml b/docs/swagger.yamlindex bee859abf..49ed133c2 100644--- a/docs/swagger.yaml+++ b/docs/swagger.yaml@@ -2468,6 +2468,7 @@ definitions:- siliconflow- jina- openrouter+ - litellm- requesty- nvidia- novita@@ -2487,6 +2488,7 @@ definitions:ModelSourceNvidia: NVIDIA modelModelSourceOpenAI: OpenAI modelModelSourceOpenRouter: OpenRouter model+ ModelSourceLiteLLM: LiteLLM proxy modelModelSourceRemote: Remote modelModelSourceRequesty: Requesty modelModelSourceSiliconFlow: SiliconFlow model@@ -2507,6 +2509,7 @@ definitions:- SiliconFlow model- Jina AI model- OpenRouter model+ - LiteLLM proxy model- Requesty model- NVIDIA model- Novita AI model@@ -2526,6 +2529,7 @@ definitions:- ModelSourceSiliconFlow- ModelSourceJina- ModelSourceOpenRouter+ - ModelSourceLiteLLM- ModelSourceRequesty- ModelSourceNvidia- ModelSourceNovitadiff --git "a/docs/wiki/\346\240\270\345\277\203\345\212\237\350\203\275/\345\206\205\347\275\256\346\250\241\345\236\213\347\256\241\347\220\206.md" "b/docs/wiki/\346\240\270\345\277\203\345\212\237\350\203\275/\345\206\205\347\275\256\346\250\241\345\236\213\347\256\241\347\220\206.md"index 8526c3550..871b96760 100644--- "a/docs/wiki/\346\240\270\345\277\203\345\212\237\350\203\275/\345\206\205\347\275\256\346\250\241\345\236\213\347\256\241\347\220\206.md"+++ "b/docs/wiki/\346\240\270\345\277\203\345\212\237\350\203\275/\345\206\205\347\275\256\346\250\241\345\236\213\347\256\241\347\220\206.md"@@ -30,7 +30,7 @@ source: BUILTIN_MODELS.md- 模型参数(parameters):包括 base_url、api_key、provider 等- 空间ID(tenant_id):建议使用小于10000的空间ID-**支持的服务商(provider)**:`generic`、`openai`、`aliyun`、`zhipu`、`volcengine`、`hunyuan`、`deepseek`、`minimax`、`mimo`、`siliconflow`、`jina`、`openrouter`、`requesty`、`gemini`、`modelscope`、`moonshot`、`qianfan`、`qiniu`、`longcat`、`gpustack`+**支持的服务商(provider)**:`generic`、`openai`、`aliyun`、`zhipu`、`volcengine`、`hunyuan`、`deepseek`、`minimax`、`mimo`、`siliconflow`、`jina`、`openrouter`、`litellm`、`requesty`、`gemini`、`modelscope`、`moonshot`、`qianfan`、`qiniu`、`longcat`、`gpustack`### 2. 执行 SQL 插入语句diff --git a/frontend/src/components/ModelEditorDialog.vue b/frontend/src/components/ModelEditorDialog.vueindex 478346614..94fcb12d6 100644--- a/frontend/src/components/ModelEditorDialog.vue+++ b/frontend/src/components/ModelEditorDialog.vue@@ -550,6 +550,17 @@ const fallbackProviderOptions = computed(() => [description: t('model.editor.providers.openrouter.description'),modelTypes: ['chat', 'embedding']},+ {+ value: 'litellm',+ label: t('model.editor.providers.litellm.label'),+ defaultUrls: {+ chat: 'http://your_litellm_proxy/v1',+ embedding: 'http://your_litellm_proxy/v1',+ vllm: 'http://your_litellm_proxy/v1'+ },+ description: t('model.editor.providers.litellm.description'),+ modelTypes: ['chat', 'embedding', 'vllm']+ },{value: 'requesty',label: t('model.editor.providers.requesty.label'),diff --git a/frontend/src/i18n/locales/en-US.ts b/frontend/src/i18n/locales/en-US.tsindex 8b7bfaa97..63fb71235 100755--- a/frontend/src/i18n/locales/en-US.ts+++ b/frontend/src/i18n/locales/en-US.ts@@ -4272,6 +4272,10 @@ export default {label: 'OpenRouter',description: 'openai/gpt-5.2-chat, google/gemini-3-flash-preview, etc.'},+ litellm: {+ label: 'LiteLLM',+ description: 'Self-hosted proxy to 100+ providers (OpenAI, Anthropic, Gemini, Bedrock, etc.). Replace the placeholder URL; loopback hosts need SSRF_WHITELIST.'+ },requesty: {label: 'Requesty',description: 'openai/gpt-4o-mini, anthropic/claude-sonnet-4-5, etc.'diff --git a/frontend/src/i18n/locales/ko-KR.ts b/frontend/src/i18n/locales/ko-KR.tsindex c9d44dbbe..8deaf14a5 100755--- a/frontend/src/i18n/locales/ko-KR.ts+++ b/frontend/src/i18n/locales/ko-KR.ts@@ -2452,6 +2452,10 @@ export default {label: 'OpenRouter',description: 'openai/gpt-5.2-chat, google/gemini-3-flash-preview 등'},+ litellm: {+ label: 'LiteLLM',+ description: '자체 호스팅 프록시로 OpenAI, Anthropic, Gemini, Bedrock 등 100+ 공급자를 연결합니다. 플레이스홀더 URL을 실제 주소로 바꾸세요. localhost는 SSRF_WHITELIST에 추가해야 합니다.'+ },zhipu: {label: 'Zhipu BigModel',description: 'glm-4.7, embedding-3, rerank, etc.'diff --git a/frontend/src/i18n/locales/ru-RU.ts b/frontend/src/i18n/locales/ru-RU.tsindex a31edb83d..c972d3dc3 100755--- a/frontend/src/i18n/locales/ru-RU.ts+++ b/frontend/src/i18n/locales/ru-RU.ts@@ -2452,6 +2452,10 @@ export default {label: 'OpenRouter',description: 'openai/gpt-5.2-chat, google/gemini-3-flash-preview, etc.'},+ litellm: {+ label: 'LiteLLM',+ description: 'Self-hosted прокси к 100+ провайдерам (OpenAI, Anthropic, Gemini, Bedrock и др.). Замените URL-заглушку; localhost нужно добавить в SSRF_WHITELIST.'+ },zhipu: {label: 'Zhipu BigModel',description: 'glm-4.7, embedding-3, rerank, etc.'diff --git a/frontend/src/i18n/locales/zh-CN.ts b/frontend/src/i18n/locales/zh-CN.tsindex c849ecd27..30cf19226 100755--- a/frontend/src/i18n/locales/zh-CN.ts+++ b/frontend/src/i18n/locales/zh-CN.ts@@ -2454,6 +2454,10 @@ export default {label: 'OpenRouter',description: 'openai/gpt-5.2-chat, google/gemini-3-flash-preview, etc.'},+ litellm: {+ label: 'LiteLLM',+ description: '自托管代理,统一接入 OpenAI、Anthropic、Gemini、Bedrock 等 100+ 厂商。请将占位 URL 换成可访问地址;localhost 需加入 SSRF_WHITELIST。'+ },zhipu: {label: '智谱 BigModel',description: 'glm-4.7, embedding-3, rerank, etc.'diff --git a/frontend/src/utils/thinkingControl.test.ts b/frontend/src/utils/thinkingControl.test.tsindex 6502bc0d4..1840e34e6 100644--- a/frontend/src/utils/thinkingControl.test.ts+++ b/frontend/src/utils/thinkingControl.test.ts@@ -8,6 +8,7 @@ test('defaultThinkingControl matches backend provider adapters', () => {const cases: Array<[string, string, ReturnType<typeof defaultThinkingControl>]> = [['generic', 'anything', 'chat_template_kwargs'],['nvidia', 'anything', 'chat_template_kwargs'],+ ['litellm', 'anything', 'chat_template_kwargs'],['volcengine', 'doubao', 'thinking_type'],['aliyun', 'qwen3-32b', 'enable_thinking'],['aliyun', 'qwen-plus', 'enable_thinking'],diff --git a/frontend/src/utils/thinkingControl.ts b/frontend/src/utils/thinkingControl.tsindex 122bced7a..9ec34cbb7 100644--- a/frontend/src/utils/thinkingControl.ts+++ b/frontend/src/utils/thinkingControl.ts@@ -47,6 +47,7 @@ export function defaultThinkingControl(return 'thinking_type'case 'generic':case 'nvidia':+ case 'litellm':return 'chat_template_kwargs'case 'volcengine':return 'thinking_type'diff --git a/internal/models/chat/provider.go b/internal/models/chat/provider.goindex c596a3547..be1a4f799 100644--- a/internal/models/chat/provider.go+++ b/internal/models/chat/provider.go@@ -156,7 +156,7 @@ func (deepseekProvider) ShapeRequest(req *openai.ChatCompletionRequest, opts *Ch}}-// --- Generic (vLLM) / NVIDIA: thinking via chat_template_kwargs ---+// --- Generic (vLLM) / NVIDIA / LiteLLM: thinking via chat_template_kwargs ---type genericProvider struct{ baseProvider }@@ -168,6 +168,11 @@ type nvidiaProvider struct{ baseProvider }func (nvidiaProvider) Name() provider.ProviderName { return provider.ProviderNvidia }func (nvidiaProvider) Thinking() ThinkingStrategy { return chatTemplateKwargs{} }+type liteLLMProvider struct{ baseProvider }++func (liteLLMProvider) Name() provider.ProviderName { return provider.ProviderLiteLLM }+func (liteLLMProvider) Thinking() ThinkingStrategy { return chatTemplateKwargs{} }+// --- Gemini OpenAI compatibility: tool thought signatures live in extra_content ---type geminiProvider struct{ baseProvider }@@ -277,6 +282,7 @@ var providerRegistry = []providerAdapter{lkeapProvider{},deepseekProvider{},genericProvider{},+ liteLLMProvider{},geminiProvider{},volcengineProvider{},nvidiaProvider{},diff --git a/internal/models/chat/provider_test.go b/internal/models/chat/provider_test.goindex bc1049fbc..688485724 100644--- a/internal/models/chat/provider_test.go+++ b/internal/models/chat/provider_test.go@@ -26,6 +26,7 @@ func TestResolveProvider(t *testing.T) {{"lkeap r1 falls back", provider.ProviderLKEAP, "deepseek-r1", baseProvider{}},{"qwen thinking", provider.ProviderAliyun, "qwen3-32b", qwenThinkingProvider{}},{"generic", provider.ProviderGeneric, "anything", genericProvider{}},+ {"litellm", provider.ProviderLiteLLM, "anything", liteLLMProvider{}},{"gemini", provider.ProviderGemini, "gemini-3-flash-preview", geminiProvider{}},{"nvidia", provider.ProviderNvidia, "anything", nvidiaProvider{}},{"volcengine", provider.ProviderVolcengine, "doubao", volcengineProvider{}},diff --git a/internal/models/provider/litellm.go b/internal/models/provider/litellm.gonew file mode 100644index 000000000..c06ca6abb--- /dev/null+++ b/internal/models/provider/litellm.go@@ -0,0 +1,55 @@+package provider++import (+ "fmt"++ "github.com/Tencent/WeKnora/internal/types"+)++const (+ // LiteLLMBaseURL is a placeholder the user must replace with a reachable+ // LiteLLM proxy. Loopback defaults (localhost:4000) are rejected by SSRF+ // unless the host is added to SSRF_WHITELIST; Docker also cannot reach a+ // proxy on the host via localhost. The hostname contains "litellm" so+ // DetectProvider recognises the catalog default.+ LiteLLMBaseURL = "http://your_litellm_proxy/v1"+)++// LiteLLMProvider implements the Provider interface for LiteLLM+// (https://github.com/BerriAI/litellm). LiteLLM exposes a single+// OpenAI-compatible endpoint that routes to 100+ providers (OpenAI, Anthropic,+// Gemini, Bedrock, Vertex, Azure, ...) behind one base URL and key, so it plugs+// into WeKnora's OpenAI-compatible transport like the other gateway providers.+type LiteLLMProvider struct{}++func init() {+ Register(&LiteLLMProvider{})+}++// Info returns the metadata for the LiteLLM provider.+func (p *LiteLLMProvider) Info() ProviderInfo {+ return ProviderInfo{+ Name: ProviderLiteLLM,+ DisplayName: "LiteLLM",+ Description: "Self-hosted LiteLLM proxy: one OpenAI-compatible endpoint to 100+ providers.",+ DefaultURLs: map[types.ModelType]string{+ types.ModelTypeKnowledgeQA: LiteLLMBaseURL,+ types.ModelTypeEmbedding: LiteLLMBaseURL,+ types.ModelTypeVLLM: LiteLLMBaseURL,+ },+ ModelTypes: []types.ModelType{+ types.ModelTypeKnowledgeQA,+ types.ModelTypeEmbedding,+ types.ModelTypeVLLM,+ },+ RequiresAuth: true,+ }+}++// ValidateConfig validates the LiteLLM provider configuration.+func (p *LiteLLMProvider) ValidateConfig(config *Config) error {+ if config.APIKey == "" {+ return fmt.Errorf("API key is required for LiteLLM provider")+ }+ return nil+}diff --git a/internal/models/provider/provider.go b/internal/models/provider/provider.goindex 0cc13342c..d96036621 100644--- a/internal/models/provider/provider.go+++ b/internal/models/provider/provider.go@@ -23,6 +23,8 @@ const (ProviderZhipu ProviderName = "zhipu"// OpenRouterProviderOpenRouter ProviderName = "openrouter"+ // ProviderLiteLLM is the LiteLLM self-hosted proxy (OpenAI-compatible gateway to 100+ providers).+ ProviderLiteLLM ProviderName = "litellm"// RequestyProviderRequesty ProviderName = "requesty"// 硅基流动@@ -85,6 +87,7 @@ func AllProviders() []ProviderName {ProviderAnthropic,ProviderGemini,ProviderOpenRouter,+ ProviderLiteLLM,ProviderRequesty,ProviderJina,ProviderMimo,@@ -227,6 +230,11 @@ func DetectProvider(baseURL string) ProviderName {return ProviderZhipucase containsAny(baseURL, "openrouter.ai"):return ProviderOpenRouter+ // Hostname/path containing "litellm" (including the catalog placeholder+ // your_litellm_proxy). Loopback URLs such as localhost:4000 stay generic+ // because they are SSRF-blocked unless explicitly whitelisted.+ case containsAny(baseURL, "litellm"):+ return ProviderLiteLLMcase containsAny(baseURL, "router.requesty.ai", "requesty.ai"):return ProviderRequestycase containsAny(baseURL, "siliconflow.cn"):diff --git a/internal/models/provider/provider_test.go b/internal/models/provider/provider_test.goindex bf13b22fe..a0a9320e4 100644--- a/internal/models/provider/provider_test.go+++ b/internal/models/provider/provider_test.go@@ -38,6 +38,9 @@ func TestDetectProvider(t *testing.T) {{"https://api.openai.com/v1", ProviderOpenAI},{"https://api.anthropic.com/v1", ProviderAnthropic},{"https://openrouter.ai/api/v1", ProviderOpenRouter},+ {"https://litellm.example.com/v1", ProviderLiteLLM},+ {LiteLLMBaseURL, ProviderLiteLLM},+ {"http://localhost:4000/v1", ProviderGeneric},{"https://router.requesty.ai/v1", ProviderRequesty},{"https://dashscope.aliyuncs.com/compatible-mode/v1", ProviderAliyun},{"https://open.bigmodel.cn/api/paas/v4", ProviderZhipu},@@ -317,3 +320,62 @@ func TestListByModelType(t *testing.T) {assert.True(t, found, "Gemini should support embedding via the native Gemini API")})}++func TestLiteLLMProviderValidation(t *testing.T) {+ p := &LiteLLMProvider{}++ t.Run("valid config", func(t *testing.T) {+ config := &Config{+ APIKey: "test-key",+ ModelName: "gpt-4.1-mini",+ }+ err := p.ValidateConfig(config)+ assert.NoError(t, err)+ })++ t.Run("missing API key", func(t *testing.T) {+ config := &Config{+ ModelName: "gpt-4.1-mini",+ }+ err := p.ValidateConfig(config)+ assert.Error(t, err)+ assert.Contains(t, err.Error(), "API key")+ })++ t.Run("info", func(t *testing.T) {+ info := p.Info()+ assert.Equal(t, ProviderLiteLLM, info.Name)+ assert.Equal(t, "LiteLLM", info.DisplayName)+ assert.True(t, info.RequiresAuth)+ assert.Equal(t, LiteLLMBaseURL, info.DefaultURLs[types.ModelTypeKnowledgeQA])+ assert.Equal(t, LiteLLMBaseURL, info.DefaultURLs[types.ModelTypeEmbedding])+ assert.Equal(t, LiteLLMBaseURL, info.DefaultURLs[types.ModelTypeVLLM])+ assert.Contains(t, info.ModelTypes, types.ModelTypeKnowledgeQA)+ assert.Contains(t, info.ModelTypes, types.ModelTypeEmbedding)+ assert.Contains(t, info.ModelTypes, types.ModelTypeVLLM)+ })++ t.Run("registered and listed", func(t *testing.T) {+ got, ok := Get(ProviderLiteLLM)+ require.True(t, ok)+ require.NotNil(t, got)++ foundChat := false+ for _, info := range ListByModelType(types.ModelTypeKnowledgeQA) {+ if info.Name == ProviderLiteLLM {+ foundChat = true+ break+ }+ }+ assert.True(t, foundChat, "LiteLLM should appear for chat models")++ foundEmbed := false+ for _, info := range ListByModelType(types.ModelTypeEmbedding) {+ if info.Name == ProviderLiteLLM {+ foundEmbed = true+ break+ }+ }+ assert.True(t, foundEmbed, "LiteLLM should appear for embedding models")+ })+}diff --git a/internal/types/model.go b/internal/types/model.goindex 2af982838..176060439 100644--- a/internal/types/model.go+++ b/internal/types/model.go@@ -49,6 +49,7 @@ const (ModelSourceSiliconFlow ModelSource = "siliconflow" // SiliconFlow modelModelSourceJina ModelSource = "jina" // Jina AI modelModelSourceOpenRouter ModelSource = "openrouter" // OpenRouter model+ ModelSourceLiteLLM ModelSource = "litellm" // LiteLLM proxy modelModelSourceRequesty ModelSource = "requesty" // Requesty modelModelSourceNvidia ModelSource = "nvidia" // NVIDIA modelModelSourceNovita ModelSource = "novita" // Novita AI modeldiff --git a/website-docs/01-getting-started/01-introduction.md b/website-docs/01-getting-started/01-introduction.mdindex d8ee8ae1d..d0fca60cc 100644--- a/website-docs/01-getting-started/01-introduction.md+++ b/website-docs/01-getting-started/01-introduction.md@@ -56,7 +56,7 @@ WeKnora(维娜拉)是腾讯开源的知识库问答系统,做的事情是| --- | --- || 会话 Session | 一次多轮对话。记录 `LastRequestState`(上次提问时选中的 Agent、模型、知识库范围、Web 搜索、MCP 服务),重开会话时恢复;上下文压缩策略(`sliding_window` / `smart` LLM 摘要)来自 `ContextConfig` || 消息 Message | `user` / `assistant` 角色消息,支持图片、附件、@提及(知识库/文档/标签/MCP/Skill),并统计 `TokenUsage`(含 prompt cache 命中情况) |-| 模型 Model | 模型注册项。`Type`:`KnowledgeQA`(对话 LLM)/ `Embedding` / `Rerank` / `VLLM`(视觉)/ `ASR`(语音);`Source`:`local`(Ollama)、`remote` 及 `openai`、`azure_openai`、`gemini`、`deepseek`、`aliyun`、`zhipu`、`volcengine`、`hunyuan`、`siliconflow`、`openrouter`、`jina` 等厂商;`ManagedBy: "yaml"` 表示由 `config/builtin_models.yaml` 声明式管理 |+| 模型 Model | 模型注册项。`Type`:`KnowledgeQA`(对话 LLM)/ `Embedding` / `Rerank` / `VLLM`(视觉)/ `ASR`(语音);`Source`:`local`(Ollama)、`remote` 及 `openai`、`azure_openai`、`gemini`、`deepseek`、`aliyun`、`zhipu`、`volcengine`、`hunyuan`、`siliconflow`、`openrouter`、`litellm`、`jina` 等厂商;`ManagedBy: "yaml"` 表示由 `config/builtin_models.yaml` 声明式管理 || Agent(自定义智能体) CustomAgent | 两种模式:`quick-answer`(经典 RAG 管线)与 `smart-reasoning`(ReAct 多步推理 + 工具调用)。smart-reasoning 下有类型预设 `AgentType`:`rag-qa` / `wiki-qa` / `hybrid-rag-wiki` / `data-analysis` / `custom`(定义见 `config/agent_type_presets.yaml`) || 内置 Agent | 开箱可用:`builtin-quick-answer`(快速问答)、`builtin-smart-reasoning`(智能推理)、`builtin-data-analyst`(数据分析)、`builtin-wiki-researcher`(Wiki 研究员)、`builtin-wiki-fixer`(Wiki 修复员)等 || MCP 服务 MCPService | Model Context Protocol 工具接入:`sse` / `http-streamable` / `stdio` 三种传输;认证支持 API Key / Bearer / OAuth2;Agent 可按 `all` / `selected` / `none` 选用其工具 |diff --git a/website-docs/03-features/06-models.md b/website-docs/03-features/06-models.mdindex 5eea10907..cda2ab0ef 100644--- a/website-docs/03-features/06-models.md+++ b/website-docs/03-features/06-models.md@@ -57,7 +57,7 @@ type Provider interface {### 支持的厂商清单-`AllProviders()`(`provider/provider.go`)返回的完整列表(共 26 个,每个厂商在自己的文件里 `init()` 注册)。表格最后一行的 Ollama 不在其中,它走 `source=local` 这条独立路径,列在这里只为方便对照:+`AllProviders()`(`provider/provider.go`)返回的完整列表(共 27 个,每个厂商在自己的文件里 `init()` 注册)。表格最后一行的 Ollama 不在其中,它走 `source=local` 这条独立路径,列在这里只为方便对照:| Provider 标识 | 名称 | 说明 ||---------------|------|------|@@ -78,6 +78,7 @@ type Provider interface {| `anthropic` | Anthropic Claude | 独立 Messages 协议实现 || `gemini` | Google Gemini | Embedding 走专用 API || `openrouter` | OpenRouter | |+| `litellm` | LiteLLM(自托管 OpenAI 兼容代理) | 默认 URL 为占位符,loopback 需加入 `SSRF_WHITELIST` || `requesty` | Requesty | || `jina` | Jina AI | Embedding 与 Rerank || `mimo` | 小米 MiMo | |diff --git a/website-docs/06-development/03-extension-points.md b/website-docs/06-development/03-extension-points.mdindex 597df9919..687dce40d 100644--- a/website-docs/06-development/03-extension-points.md+++ b/website-docs/06-development/03-extension-points.md@@ -342,7 +342,7 @@ type Reranker interface {### 现有实现-`internal/models/provider/provider.go` 中已定义 26 个 `ProviderName` 常量:openai、anthropic、aliyun、zhipu、openrouter、requesty、siliconflow、jina、generic、deepseek、gemini、volcengine、hunyuan、minimax、mimo、gpustack、moonshot、modelscope、qianfan、qiniu、longcat、lkeap、nvidia 等。具体 Provider 实现分布在 `internal/models/provider/` 下的各文件(如 `zhipu.go`、`gemini.go`、`hunyuan.go`、`generic.go`);特殊 embedding 实现如 `internal/models/embedding/jina.go`、`volcengine.go`、`nvidia.go`。+`internal/models/provider/provider.go` 中已定义 27 个 `ProviderName` 常量:openai、anthropic、aliyun、zhipu、openrouter、litellm、requesty、siliconflow、jina、generic、deepseek、gemini、volcengine、hunyuan、minimax、mimo、gpustack、moonshot、modelscope、qianfan、qiniu、longcat、lkeap、nvidia 等。具体 Provider 实现分布在 `internal/models/provider/` 下的各文件(如 `zhipu.go`、`gemini.go`、`hunyuan.go`、`generic.go`);特殊 embedding 实现如 `internal/models/embedding/jina.go`、`volcengine.go`、`nvidia.go`。### 新增步骤