完善文档页文本向量与重排序调用支持

This commit is contained in:
wangbo 2026-05-31 21:18:41 +08:00
parent 8ee7a7969e
commit 644a6f9d17
24 changed files with 1945 additions and 71 deletions

View File

@ -4081,6 +4081,99 @@
}
}
},
"/api/v1/embeddings": {
"post": {
"security": [
{
"BearerAuth": []
}
],
"description": "网关任务接口按 model 选择平台模型;除 /api/v1/chat/completions 以外的 /api/v1 任务路径返回任务受理结果OpenAI-compatible 路径同步返回兼容响应或 SSE 流。",
"consumes": [
"application/json"
],
"produces": [
"application/json"
],
"tags": [
"tasks"
],
"summary": "创建或执行 AI 任务",
"parameters": [
{
"type": "boolean",
"description": "true 时异步创建任务并返回 202",
"name": "X-Async",
"in": "header"
},
{
"description": "AI 任务请求,字段随任务类型变化",
"name": "input",
"in": "body",
"required": true,
"schema": {
"$ref": "#/definitions/httpapi.TaskRequest"
}
}
],
"responses": {
"200": {
"description": "OK",
"schema": {
"$ref": "#/definitions/httpapi.CompatibleResponse"
}
},
"202": {
"description": "Accepted",
"schema": {
"$ref": "#/definitions/httpapi.TaskAcceptedResponse"
}
},
"400": {
"description": "Bad Request",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"401": {
"description": "Unauthorized",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"402": {
"description": "Payment Required",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"403": {
"description": "Forbidden",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"404": {
"description": "Not Found",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"429": {
"description": "Too Many Requests",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"502": {
"description": "Bad Gateway",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
}
}
}
},
"/api/v1/files/upload": {
"post": {
"security": [
@ -4641,6 +4734,99 @@
}
}
},
"/api/v1/reranks": {
"post": {
"security": [
{
"BearerAuth": []
}
],
"description": "网关任务接口按 model 选择平台模型;除 /api/v1/chat/completions 以外的 /api/v1 任务路径返回任务受理结果OpenAI-compatible 路径同步返回兼容响应或 SSE 流。",
"consumes": [
"application/json"
],
"produces": [
"application/json"
],
"tags": [
"tasks"
],
"summary": "创建或执行 AI 任务",
"parameters": [
{
"type": "boolean",
"description": "true 时异步创建任务并返回 202",
"name": "X-Async",
"in": "header"
},
{
"description": "AI 任务请求,字段随任务类型变化",
"name": "input",
"in": "body",
"required": true,
"schema": {
"$ref": "#/definitions/httpapi.TaskRequest"
}
}
],
"responses": {
"200": {
"description": "OK",
"schema": {
"$ref": "#/definitions/httpapi.CompatibleResponse"
}
},
"202": {
"description": "Accepted",
"schema": {
"$ref": "#/definitions/httpapi.TaskAcceptedResponse"
}
},
"400": {
"description": "Bad Request",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"401": {
"description": "Unauthorized",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"402": {
"description": "Payment Required",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"403": {
"description": "Forbidden",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"404": {
"description": "Not Found",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"429": {
"description": "Too Many Requests",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"502": {
"description": "Bad Gateway",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
}
}
}
},
"/api/v1/responses": {
"post": {
"security": [
@ -5533,6 +5719,99 @@
}
}
},
"/embeddings": {
"post": {
"security": [
{
"BearerAuth": []
}
],
"description": "网关任务接口按 model 选择平台模型;除 /api/v1/chat/completions 以外的 /api/v1 任务路径返回任务受理结果OpenAI-compatible 路径同步返回兼容响应或 SSE 流。",
"consumes": [
"application/json"
],
"produces": [
"application/json"
],
"tags": [
"tasks"
],
"summary": "创建或执行 AI 任务",
"parameters": [
{
"type": "boolean",
"description": "true 时异步创建任务并返回 202",
"name": "X-Async",
"in": "header"
},
{
"description": "AI 任务请求,字段随任务类型变化",
"name": "input",
"in": "body",
"required": true,
"schema": {
"$ref": "#/definitions/httpapi.TaskRequest"
}
}
],
"responses": {
"200": {
"description": "OK",
"schema": {
"$ref": "#/definitions/httpapi.CompatibleResponse"
}
},
"202": {
"description": "Accepted",
"schema": {
"$ref": "#/definitions/httpapi.TaskAcceptedResponse"
}
},
"400": {
"description": "Bad Request",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"401": {
"description": "Unauthorized",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"402": {
"description": "Payment Required",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"403": {
"description": "Forbidden",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"404": {
"description": "Not Found",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"429": {
"description": "Too Many Requests",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"502": {
"description": "Bad Gateway",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
}
}
}
},
"/healthz": {
"get": {
"description": "返回服务进程、运行环境和身份模式,供负载均衡或人工排障使用。",
@ -5765,6 +6044,99 @@
}
}
},
"/reranks": {
"post": {
"security": [
{
"BearerAuth": []
}
],
"description": "网关任务接口按 model 选择平台模型;除 /api/v1/chat/completions 以外的 /api/v1 任务路径返回任务受理结果OpenAI-compatible 路径同步返回兼容响应或 SSE 流。",
"consumes": [
"application/json"
],
"produces": [
"application/json"
],
"tags": [
"tasks"
],
"summary": "创建或执行 AI 任务",
"parameters": [
{
"type": "boolean",
"description": "true 时异步创建任务并返回 202",
"name": "X-Async",
"in": "header"
},
{
"description": "AI 任务请求,字段随任务类型变化",
"name": "input",
"in": "body",
"required": true,
"schema": {
"$ref": "#/definitions/httpapi.TaskRequest"
}
}
],
"responses": {
"200": {
"description": "OK",
"schema": {
"$ref": "#/definitions/httpapi.CompatibleResponse"
}
},
"202": {
"description": "Accepted",
"schema": {
"$ref": "#/definitions/httpapi.TaskAcceptedResponse"
}
},
"400": {
"description": "Bad Request",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"401": {
"description": "Unauthorized",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"402": {
"description": "Payment Required",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"403": {
"description": "Forbidden",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"404": {
"description": "Not Found",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"429": {
"description": "Too Many Requests",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"502": {
"description": "Bad Gateway",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
}
}
}
},
"/responses": {
"post": {
"security": [
@ -6057,6 +6429,99 @@
}
}
},
"/v1/embeddings": {
"post": {
"security": [
{
"BearerAuth": []
}
],
"description": "网关任务接口按 model 选择平台模型;除 /api/v1/chat/completions 以外的 /api/v1 任务路径返回任务受理结果OpenAI-compatible 路径同步返回兼容响应或 SSE 流。",
"consumes": [
"application/json"
],
"produces": [
"application/json"
],
"tags": [
"tasks"
],
"summary": "创建或执行 AI 任务",
"parameters": [
{
"type": "boolean",
"description": "true 时异步创建任务并返回 202",
"name": "X-Async",
"in": "header"
},
{
"description": "AI 任务请求,字段随任务类型变化",
"name": "input",
"in": "body",
"required": true,
"schema": {
"$ref": "#/definitions/httpapi.TaskRequest"
}
}
],
"responses": {
"200": {
"description": "OK",
"schema": {
"$ref": "#/definitions/httpapi.CompatibleResponse"
}
},
"202": {
"description": "Accepted",
"schema": {
"$ref": "#/definitions/httpapi.TaskAcceptedResponse"
}
},
"400": {
"description": "Bad Request",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"401": {
"description": "Unauthorized",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"402": {
"description": "Payment Required",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"403": {
"description": "Forbidden",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"404": {
"description": "Not Found",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"429": {
"description": "Too Many Requests",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"502": {
"description": "Bad Gateway",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
}
}
}
},
"/v1/files/upload": {
"post": {
"security": [
@ -6311,6 +6776,99 @@
}
}
},
"/v1/reranks": {
"post": {
"security": [
{
"BearerAuth": []
}
],
"description": "网关任务接口按 model 选择平台模型;除 /api/v1/chat/completions 以外的 /api/v1 任务路径返回任务受理结果OpenAI-compatible 路径同步返回兼容响应或 SSE 流。",
"consumes": [
"application/json"
],
"produces": [
"application/json"
],
"tags": [
"tasks"
],
"summary": "创建或执行 AI 任务",
"parameters": [
{
"type": "boolean",
"description": "true 时异步创建任务并返回 202",
"name": "X-Async",
"in": "header"
},
{
"description": "AI 任务请求,字段随任务类型变化",
"name": "input",
"in": "body",
"required": true,
"schema": {
"$ref": "#/definitions/httpapi.TaskRequest"
}
}
],
"responses": {
"200": {
"description": "OK",
"schema": {
"$ref": "#/definitions/httpapi.CompatibleResponse"
}
},
"202": {
"description": "Accepted",
"schema": {
"$ref": "#/definitions/httpapi.TaskAcceptedResponse"
}
},
"400": {
"description": "Bad Request",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"401": {
"description": "Unauthorized",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"402": {
"description": "Payment Required",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"403": {
"description": "Forbidden",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"404": {
"description": "Not Found",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"429": {
"description": "Too Many Requests",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
},
"502": {
"description": "Bad Gateway",
"schema": {
"$ref": "#/definitions/httpapi.ErrorEnvelope"
}
}
}
}
},
"/v1/responses": {
"post": {
"security": [

View File

@ -4912,6 +4912,67 @@ paths:
summary: 创建 Chat Completions
tags:
- tasks
/api/v1/embeddings:
post:
consumes:
- application/json
description: 网关任务接口按 model 选择平台模型;除 /api/v1/chat/completions 以外的 /api/v1 任务路径返回任务受理结果OpenAI-compatible
路径同步返回兼容响应或 SSE 流。
parameters:
- description: true 时异步创建任务并返回 202
in: header
name: X-Async
type: boolean
- description: AI 任务请求,字段随任务类型变化
in: body
name: input
required: true
schema:
$ref: '#/definitions/httpapi.TaskRequest'
produces:
- application/json
responses:
"200":
description: OK
schema:
$ref: '#/definitions/httpapi.CompatibleResponse'
"202":
description: Accepted
schema:
$ref: '#/definitions/httpapi.TaskAcceptedResponse'
"400":
description: Bad Request
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"401":
description: Unauthorized
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"402":
description: Payment Required
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"403":
description: Forbidden
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"404":
description: Not Found
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"429":
description: Too Many Requests
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"502":
description: Bad Gateway
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
security:
- BearerAuth: []
summary: 创建或执行 AI 任务
tags:
- tasks
/api/v1/files/upload:
post:
consumes:
@ -5271,6 +5332,67 @@ paths:
summary: 列出目录供应商
tags:
- catalog
/api/v1/reranks:
post:
consumes:
- application/json
description: 网关任务接口按 model 选择平台模型;除 /api/v1/chat/completions 以外的 /api/v1 任务路径返回任务受理结果OpenAI-compatible
路径同步返回兼容响应或 SSE 流。
parameters:
- description: true 时异步创建任务并返回 202
in: header
name: X-Async
type: boolean
- description: AI 任务请求,字段随任务类型变化
in: body
name: input
required: true
schema:
$ref: '#/definitions/httpapi.TaskRequest'
produces:
- application/json
responses:
"200":
description: OK
schema:
$ref: '#/definitions/httpapi.CompatibleResponse'
"202":
description: Accepted
schema:
$ref: '#/definitions/httpapi.TaskAcceptedResponse'
"400":
description: Bad Request
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"401":
description: Unauthorized
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"402":
description: Payment Required
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"403":
description: Forbidden
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"404":
description: Not Found
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"429":
description: Too Many Requests
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"502":
description: Bad Gateway
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
security:
- BearerAuth: []
summary: 创建或执行 AI 任务
tags:
- tasks
/api/v1/responses:
post:
consumes:
@ -5847,6 +5969,67 @@ paths:
summary: 创建或执行 AI 任务
tags:
- tasks
/embeddings:
post:
consumes:
- application/json
description: 网关任务接口按 model 选择平台模型;除 /api/v1/chat/completions 以外的 /api/v1 任务路径返回任务受理结果OpenAI-compatible
路径同步返回兼容响应或 SSE 流。
parameters:
- description: true 时异步创建任务并返回 202
in: header
name: X-Async
type: boolean
- description: AI 任务请求,字段随任务类型变化
in: body
name: input
required: true
schema:
$ref: '#/definitions/httpapi.TaskRequest'
produces:
- application/json
responses:
"200":
description: OK
schema:
$ref: '#/definitions/httpapi.CompatibleResponse'
"202":
description: Accepted
schema:
$ref: '#/definitions/httpapi.TaskAcceptedResponse'
"400":
description: Bad Request
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"401":
description: Unauthorized
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"402":
description: Payment Required
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"403":
description: Forbidden
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"404":
description: Not Found
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"429":
description: Too Many Requests
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"502":
description: Bad Gateway
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
security:
- BearerAuth: []
summary: 创建或执行 AI 任务
tags:
- tasks
/healthz:
get:
description: 返回服务进程、运行环境和身份模式,供负载均衡或人工排障使用。
@ -5999,6 +6182,67 @@ paths:
summary: 就绪检查
tags:
- system
/reranks:
post:
consumes:
- application/json
description: 网关任务接口按 model 选择平台模型;除 /api/v1/chat/completions 以外的 /api/v1 任务路径返回任务受理结果OpenAI-compatible
路径同步返回兼容响应或 SSE 流。
parameters:
- description: true 时异步创建任务并返回 202
in: header
name: X-Async
type: boolean
- description: AI 任务请求,字段随任务类型变化
in: body
name: input
required: true
schema:
$ref: '#/definitions/httpapi.TaskRequest'
produces:
- application/json
responses:
"200":
description: OK
schema:
$ref: '#/definitions/httpapi.CompatibleResponse'
"202":
description: Accepted
schema:
$ref: '#/definitions/httpapi.TaskAcceptedResponse'
"400":
description: Bad Request
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"401":
description: Unauthorized
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"402":
description: Payment Required
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"403":
description: Forbidden
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"404":
description: Not Found
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"429":
description: Too Many Requests
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"502":
description: Bad Gateway
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
security:
- BearerAuth: []
summary: 创建或执行 AI 任务
tags:
- tasks
/responses:
post:
consumes:
@ -6191,6 +6435,67 @@ paths:
summary: 创建或执行 AI 任务
tags:
- tasks
/v1/embeddings:
post:
consumes:
- application/json
description: 网关任务接口按 model 选择平台模型;除 /api/v1/chat/completions 以外的 /api/v1 任务路径返回任务受理结果OpenAI-compatible
路径同步返回兼容响应或 SSE 流。
parameters:
- description: true 时异步创建任务并返回 202
in: header
name: X-Async
type: boolean
- description: AI 任务请求,字段随任务类型变化
in: body
name: input
required: true
schema:
$ref: '#/definitions/httpapi.TaskRequest'
produces:
- application/json
responses:
"200":
description: OK
schema:
$ref: '#/definitions/httpapi.CompatibleResponse'
"202":
description: Accepted
schema:
$ref: '#/definitions/httpapi.TaskAcceptedResponse'
"400":
description: Bad Request
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"401":
description: Unauthorized
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"402":
description: Payment Required
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"403":
description: Forbidden
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"404":
description: Not Found
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"429":
description: Too Many Requests
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"502":
description: Bad Gateway
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
security:
- BearerAuth: []
summary: 创建或执行 AI 任务
tags:
- tasks
/v1/files/upload:
post:
consumes:
@ -6357,6 +6662,67 @@ paths:
summary: 创建或执行 AI 任务
tags:
- tasks
/v1/reranks:
post:
consumes:
- application/json
description: 网关任务接口按 model 选择平台模型;除 /api/v1/chat/completions 以外的 /api/v1 任务路径返回任务受理结果OpenAI-compatible
路径同步返回兼容响应或 SSE 流。
parameters:
- description: true 时异步创建任务并返回 202
in: header
name: X-Async
type: boolean
- description: AI 任务请求,字段随任务类型变化
in: body
name: input
required: true
schema:
$ref: '#/definitions/httpapi.TaskRequest'
produces:
- application/json
responses:
"200":
description: OK
schema:
$ref: '#/definitions/httpapi.CompatibleResponse'
"202":
description: Accepted
schema:
$ref: '#/definitions/httpapi.TaskAcceptedResponse'
"400":
description: Bad Request
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"401":
description: Unauthorized
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"402":
description: Payment Required
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"403":
description: Forbidden
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"404":
description: Not Found
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"429":
description: Too Many Requests
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"502":
description: Bad Gateway
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
security:
- BearerAuth: []
summary: 创建或执行 AI 任务
tags:
- tasks
/v1/responses:
post:
consumes:

View File

@ -151,6 +151,107 @@ func TestOpenAIClientChatContract(t *testing.T) {
}
}
func TestOpenAIClientEmbeddingsContract(t *testing.T) {
var gotPath string
var gotModel string
var gotDimensions float64
server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
gotPath = r.URL.Path
var body map[string]any
if err := json.NewDecoder(r.Body).Decode(&body); err != nil {
t.Fatalf("decode request: %v", err)
}
gotModel, _ = body["model"].(string)
gotDimensions, _ = body["dimensions"].(float64)
_ = json.NewEncoder(w).Encode(map[string]any{
"id": "embd-test",
"object": "list",
"model": gotModel,
"data": []any{map[string]any{
"object": "embedding",
"index": 0,
"embedding": []any{0.1, 0.2, 0.3},
}},
"usage": map[string]any{"prompt_tokens": 3, "total_tokens": 3},
})
}))
defer server.Close()
response, err := (OpenAIClient{HTTPClient: server.Client()}).Run(context.Background(), Request{
Kind: "embeddings",
Model: "aliyun-bailian-openai:text-embedding-v4",
Body: map[string]any{
"model": "aliyun-bailian-openai:text-embedding-v4",
"input": []any{"hello"},
"dimensions": 3,
},
Candidate: store.RuntimeModelCandidate{
BaseURL: server.URL,
ProviderModelName: "text-embedding-v4",
Credentials: map[string]any{"apiKey": "test-key"},
},
})
if err != nil {
t.Fatalf("run embeddings client: %v", err)
}
if gotPath != "/embeddings" || gotModel != "text-embedding-v4" || gotDimensions != 3 {
t.Fatalf("unexpected embeddings request path=%s model=%s dimensions=%v", gotPath, gotModel, gotDimensions)
}
if response.Usage.InputTokens != 3 || response.Usage.TotalTokens != 3 || response.Result["id"] != "embd-test" {
t.Fatalf("unexpected embeddings response: %+v", response)
}
}
func TestOpenAIClientAliyunRerankUsesCompatibleAPIBase(t *testing.T) {
var gotPath string
var gotModel string
server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
gotPath = r.URL.Path
var body map[string]any
if err := json.NewDecoder(r.Body).Decode(&body); err != nil {
t.Fatalf("decode request: %v", err)
}
gotModel, _ = body["model"].(string)
_ = json.NewEncoder(w).Encode(map[string]any{
"id": "rerank-test",
"object": "list",
"model": gotModel,
"results": []any{
map[string]any{"index": 0, "relevance_score": 0.93},
map[string]any{"index": 2, "relevance_score": 0.34},
},
"usage": map[string]any{"total_tokens": 9},
})
}))
defer server.Close()
response, err := (OpenAIClient{HTTPClient: server.Client()}).Run(context.Background(), Request{
Kind: "reranks",
Model: "aliyun-bailian-openai:qwen3-rerank",
Body: map[string]any{
"model": "aliyun-bailian-openai:qwen3-rerank",
"query": "what is rerank",
"documents": []any{"rerank sorts documents", "unrelated"},
"top_n": 2,
},
Candidate: store.RuntimeModelCandidate{
Provider: "aliyun-bailian-openai",
BaseURL: server.URL + "/compatible-mode/v1",
ProviderModelName: "qwen3-rerank",
Credentials: map[string]any{"apiKey": "test-key"},
},
})
if err != nil {
t.Fatalf("run rerank client: %v", err)
}
if gotPath != "/compatible-api/v1/reranks" || gotModel != "qwen3-rerank" {
t.Fatalf("unexpected rerank request path=%s model=%s", gotPath, gotModel)
}
if response.Usage.TotalTokens != 9 || response.Result["id"] != "rerank-test" {
t.Fatalf("unexpected rerank response: %+v", response)
}
}
func TestOpenAIClientChatRequestNormalizesToolContext(t *testing.T) {
var captured map[string]any
server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {

View File

@ -7,6 +7,8 @@ import (
"net/http"
"strings"
"time"
"github.com/easyai/easyai-ai-gateway/apps/api/internal/store"
)
type OpenAIClient struct {
@ -27,10 +29,10 @@ func (c OpenAIClient) Run(ctx context.Context, request Request) (Response, error
body = NormalizeChatCompletionRequestBody(body)
}
body["model"] = upstreamModelName(request.Candidate)
stream := request.Stream || boolValue(body, "stream")
stream := openAIEndpointSupportsStream(request.Kind) && (request.Stream || boolValue(body, "stream"))
ensureOpenAIStreamUsage(body, request.Kind, stream)
raw, _ := json.Marshal(body)
req, err := http.NewRequestWithContext(ctx, http.MethodPost, joinURL(request.Candidate.BaseURL, endpoint), bytes.NewReader(raw))
req, err := http.NewRequestWithContext(ctx, http.MethodPost, joinURL(openAIBaseURL(request.Kind, request.Candidate), endpoint), bytes.NewReader(raw))
if err != nil {
return Response{}, err
}
@ -81,6 +83,10 @@ func openAIEndpoint(kind string) string {
return "/chat/completions"
case "responses":
return "/responses"
case "embeddings":
return "/embeddings"
case "reranks":
return "/reranks"
case "images.generations":
return "/images/generations"
case "images.edits":
@ -90,6 +96,24 @@ func openAIEndpoint(kind string) string {
}
}
func openAIEndpointSupportsStream(kind string) bool {
return kind == "chat.completions" || kind == "responses"
}
func openAIBaseURL(kind string, candidate store.RuntimeModelCandidate) string {
base := strings.TrimSpace(candidate.BaseURL)
if kind != "reranks" {
return base
}
if strings.Contains(base, "/compatible-mode/") && (strings.EqualFold(candidate.Provider, "aliyun-bailian-openai") || strings.Contains(base, "dashscope")) {
return strings.Replace(base, "/compatible-mode/", "/compatible-api/", 1)
}
if base == "" && strings.EqualFold(candidate.Provider, "aliyun-bailian-openai") {
return "https://dashscope.aliyuncs.com/compatible-api/v1"
}
return base
}
func cloneBody(body map[string]any) map[string]any {
out := map[string]any{}
for key, value := range body {

View File

@ -131,6 +131,10 @@ func simulatedResult(request Request) map[string]any {
"output_text": fmt.Sprintf("simulation response from %s", request.Candidate.Provider),
"usage": map[string]any{"input_tokens": 12, "output_tokens": 8, "total_tokens": 20},
}
case "embeddings":
return simulatedEmbeddingResult(request)
case "reranks":
return simulatedRerankResult(request)
case "images.edits":
return map[string]any{
"id": "img-edit-simulated",
@ -172,6 +176,106 @@ func simulatedResult(request Request) map[string]any {
}
}
func simulatedEmbeddingResult(request Request) map[string]any {
inputCount := simulatedEmbeddingInputCount(request.Body["input"])
dimensions := intValue(request.Body, "dimensions", 3)
if dimensions <= 0 {
dimensions = 3
}
if dimensions > 2048 {
dimensions = 2048
}
data := make([]any, 0, inputCount)
for index := 0; index < inputCount; index += 1 {
embedding := make([]any, 0, dimensions)
for dimension := 0; dimension < dimensions; dimension += 1 {
embedding = append(embedding, float64(index+1)/10+float64(dimension)/100)
}
data = append(data, map[string]any{
"object": "embedding",
"index": index,
"embedding": embedding,
})
}
usage := simulatedUsage(request)
return map[string]any{
"id": "embd-simulated",
"object": "list",
"model": request.Model,
"data": data,
"usage": map[string]any{"prompt_tokens": usage.InputTokens, "total_tokens": usage.TotalTokens},
}
}
func simulatedEmbeddingInputCount(value any) int {
switch typed := value.(type) {
case []any:
if len(typed) > 0 {
return len(typed)
}
case []string:
if len(typed) > 0 {
return len(typed)
}
}
return 1
}
func simulatedRerankResult(request Request) map[string]any {
documents := simulatedRerankDocuments(request.Body["documents"])
topN := intValue(request.Body, "top_n", len(documents))
if topN <= 0 || topN > len(documents) {
topN = len(documents)
}
results := make([]any, 0, topN)
for index := 0; index < topN; index += 1 {
score := 0.95 - float64(index)*0.1
if score < 0 {
score = 0
}
result := map[string]any{
"index": index,
"relevance_score": score,
}
if boolValue(request.Body, "return_documents") {
result["document"] = map[string]any{"text": documents[index]}
}
results = append(results, result)
}
usage := simulatedUsage(request)
return map[string]any{
"id": "rerank-simulated",
"object": "list",
"model": request.Model,
"results": results,
"usage": map[string]any{"total_tokens": usage.TotalTokens},
}
}
func simulatedRerankDocuments(value any) []string {
switch typed := value.(type) {
case []any:
out := make([]string, 0, len(typed))
for _, item := range typed {
text := stringValue(map[string]any{"value": item}, "value")
if text == "" {
if record, ok := item.(map[string]any); ok {
text = firstNonEmptyString(stringValue(record, "text"), stringValue(record, "content"))
}
}
out = append(out, text)
}
if len(out) > 0 {
return out
}
case []string:
if len(typed) > 0 {
return typed
}
}
return []string{"simulated document"}
}
func simulatedImageData(request Request, url string, fallbackPrompt string) []any {
count := simulatedOutputCount(request.Body)
items := make([]any, 0, count)
@ -207,6 +311,9 @@ func simulatedUsage(request Request) Usage {
if request.ModelType == "chat" || request.ModelType == "text_generate" || request.Kind == "responses" {
return Usage{InputTokens: 12, OutputTokens: 8, TotalTokens: 20}
}
if request.ModelType == "text_embedding" || request.ModelType == "text_rerank" || request.Kind == "embeddings" || request.Kind == "reranks" {
return Usage{InputTokens: 16, TotalTokens: 16}
}
return Usage{}
}

View File

@ -31,6 +31,33 @@ func TestPlanTaskResponseTreatsAPIV1ChatCompletionsAsSynchronousCompatibleRespon
}
}
func TestPlanTaskResponseTreatsAPIV1EmbeddingAndRerankAsSynchronousCompatibleResponse(t *testing.T) {
for _, item := range []struct {
kind string
path string
}{
{kind: "embeddings", path: "/api/v1/embeddings"},
{kind: "reranks", path: "/api/v1/reranks"},
} {
t.Run(item.kind, func(t *testing.T) {
req := httptest.NewRequest(http.MethodPost, item.path, nil)
req.Header.Set("X-Async", "true")
plan := planTaskResponse(item.kind, false, map[string]any{"stream": true}, req)
if plan.asyncMode {
t.Fatalf("%s must not enter async task mode", item.path)
}
if !plan.compatibleMode {
t.Fatalf("%s should return compatible response payloads", item.path)
}
if plan.streamMode {
t.Fatal("embedding and rerank endpoints should stay JSON-only even when stream=true is present")
}
})
}
}
func TestPlanTaskResponseKeepsAsyncTaskModeForOtherAPIV1Tasks(t *testing.T) {
req := httptest.NewRequest(http.MethodPost, "/api/v1/images/generations", nil)
req.Header.Set("X-Async", "true")

View File

@ -876,6 +876,8 @@ func (s *Server) listModelRateLimitStatuses(w http.ResponseWriter, r *http.Reque
// @Failure 429 {object} ErrorEnvelope
// @Failure 502 {object} ErrorEnvelope
// @Router /api/v1/responses [post]
// @Router /api/v1/embeddings [post]
// @Router /api/v1/reranks [post]
// @Router /api/v1/images/generations [post]
// @Router /api/v1/images/edits [post]
// @Router /api/v1/videos/generations [post]
@ -883,6 +885,10 @@ func (s *Server) listModelRateLimitStatuses(w http.ResponseWriter, r *http.Reque
// @Router /v1/chat/completions [post]
// @Router /responses [post]
// @Router /v1/responses [post]
// @Router /embeddings [post]
// @Router /v1/embeddings [post]
// @Router /reranks [post]
// @Router /v1/reranks [post]
// @Router /images/generations [post]
// @Router /v1/images/generations [post]
// @Router /images/edits [post]
@ -1085,17 +1091,25 @@ type taskResponsePlan struct {
func planTaskResponse(kind string, compatible bool, body map[string]any, r *http.Request) taskResponsePlan {
asyncMode := asyncRequest(r)
compatibleMode := compatible
if kind == "chat.completions" && !compatible {
if synchronousCompatibleKind(kind) && !compatible {
asyncMode = false
compatibleMode = true
}
return taskResponsePlan{
asyncMode: asyncMode,
compatibleMode: compatibleMode,
streamMode: boolValue(body, "stream"),
streamMode: streamCompatibleKind(kind) && boolValue(body, "stream"),
}
}
func synchronousCompatibleKind(kind string) bool {
return kind == "chat.completions" || kind == "embeddings" || kind == "reranks"
}
func streamCompatibleKind(kind string) bool {
return kind == "chat.completions" || kind == "responses"
}
func writeTaskAccepted(w http.ResponseWriter, task store.GatewayTask) {
writeJSON(w, http.StatusAccepted, map[string]any{
"taskId": task.ID,
@ -1120,6 +1134,12 @@ func apiKeyScopeAllowed(user *auth.User, kind string) bool {
if required == "chat" && (scope == "text" || scope == "text_generate") {
return true
}
if required == "embedding" && scope == "text_embedding" {
return true
}
if required == "rerank" && scope == "text_rerank" {
return true
}
}
return false
}
@ -1128,6 +1148,10 @@ func scopeForTaskKind(kind string) string {
switch kind {
case "chat.completions", "responses":
return "chat"
case "embeddings":
return "embedding"
case "reranks":
return "rerank"
case "images.generations", "images.edits":
return "image"
case "videos.generations":

View File

@ -1025,6 +1025,7 @@ func modelCatalogCapabilityDefinitions() []ModelCatalogFilterOption {
{Value: "text_to_speech", Label: "语音合成"},
{Value: "audio_understanding", Label: "音频理解"},
{Value: "text_embedding", Label: "Embedding"},
{Value: "text_rerank", Label: "重排序"},
{Value: "omni", Label: "全模态"},
{Value: "omni_video", Label: "全模态视频"},
{Value: "multimodal", Label: "多模态"},
@ -1120,6 +1121,8 @@ func canonicalCapabilityFilterValue(value string) string {
switch normalized {
case "embedding":
return "text_embedding"
case "rerank", "reranks":
return "text_rerank"
case "model":
return "model_3d"
default:
@ -1143,6 +1146,8 @@ func capabilityFilterValueForTag(tag string) string {
return "structured_output"
case "数字人":
return "digital_human"
case "重排序":
return "text_rerank"
case "3D 模型":
return "model_3d"
case "文生 3D":
@ -1165,6 +1170,9 @@ func capabilityLabel(value string) string {
"responses": "Responses",
"text_embedding": "Embedding",
"embedding": "Embedding",
"text_rerank": "重排序",
"rerank": "重排序",
"reranks": "重排序",
"image_generate": "图像生成",
"image_edit": "图像编辑",
"image_analysis": "图像分析",

View File

@ -128,6 +128,8 @@ func NewServerWithContext(ctx context.Context, cfg config.Config, db *store.Stor
mux.Handle("GET /api/admin/runtime/model-rate-limits", server.requireAdmin(auth.PermissionPower, http.HandlerFunc(server.listModelRateLimitStatuses)))
mux.Handle("POST /api/v1/chat/completions", server.auth.Require(auth.PermissionBasic, server.createAPIV1ChatCompletions()))
mux.Handle("POST /api/v1/responses", server.auth.Require(auth.PermissionBasic, server.createTask("responses", false)))
mux.Handle("POST /api/v1/embeddings", server.auth.Require(auth.PermissionBasic, server.createTask("embeddings", false)))
mux.Handle("POST /api/v1/reranks", server.auth.Require(auth.PermissionBasic, server.createTask("reranks", false)))
mux.Handle("POST /api/v1/images/generations", server.auth.Require(auth.PermissionBasic, server.createTask("images.generations", false)))
mux.Handle("POST /api/v1/images/edits", server.auth.Require(auth.PermissionBasic, server.createTask("images.edits", false)))
mux.Handle("POST /api/v1/videos/generations", server.auth.Require(auth.PermissionBasic, server.createTask("videos.generations", false)))
@ -140,6 +142,10 @@ func NewServerWithContext(ctx context.Context, cfg config.Config, db *store.Stor
mux.Handle("POST /v1/chat/completions", server.auth.Require(auth.PermissionBasic, server.createTask("chat.completions", true)))
mux.Handle("POST /responses", server.auth.Require(auth.PermissionBasic, server.createTask("responses", true)))
mux.Handle("POST /v1/responses", server.auth.Require(auth.PermissionBasic, server.createTask("responses", true)))
mux.Handle("POST /embeddings", server.auth.Require(auth.PermissionBasic, server.createTask("embeddings", true)))
mux.Handle("POST /v1/embeddings", server.auth.Require(auth.PermissionBasic, server.createTask("embeddings", true)))
mux.Handle("POST /reranks", server.auth.Require(auth.PermissionBasic, server.createTask("reranks", true)))
mux.Handle("POST /v1/reranks", server.auth.Require(auth.PermissionBasic, server.createTask("reranks", true)))
mux.Handle("POST /images/generations", server.auth.Require(auth.PermissionBasic, server.createTask("images.generations", true)))
mux.Handle("POST /v1/images/generations", server.auth.Require(auth.PermissionBasic, server.createTask("images.generations", true)))
mux.Handle("POST /images/edits", server.auth.Require(auth.PermissionBasic, server.createTask("images.edits", true)))

View File

@ -159,16 +159,14 @@ func hasRules(policy map[string]any) bool {
}
func estimateRequestTokens(body map[string]any) int {
text := ""
if prompt := stringFromMap(body, "prompt"); prompt != "" {
text += prompt
}
if input := stringFromMap(body, "input"); input != "" {
text += input
}
var text strings.Builder
appendTokenEstimateText(&text, body["prompt"])
appendTokenEstimateText(&text, body["input"])
appendTokenEstimateText(&text, body["query"])
appendTokenEstimateText(&text, body["documents"])
for _, item := range contentItems(body["content"]) {
if stringFromAny(item["type"]) == "text" {
text += stringFromAny(item["text"])
appendTokenEstimateText(&text, item["text"])
}
}
if messages, ok := body["messages"].([]any); ok {
@ -176,19 +174,41 @@ func estimateRequestTokens(body map[string]any) int {
message, _ := raw.(map[string]any)
switch content := message["content"].(type) {
case string:
text += content
appendTokenEstimateText(&text, content)
case []any:
for _, rawPart := range content {
part, _ := rawPart.(map[string]any)
text += stringFromMap(part, "text")
appendTokenEstimateText(&text, part["text"])
}
}
}
}
if text == "" {
estimatedText := text.String()
if estimatedText == "" {
return 1
}
return len([]rune(text))/4 + 1
return len([]rune(estimatedText))/4 + 1
}
func appendTokenEstimateText(out *strings.Builder, value any) {
switch typed := value.(type) {
case string:
out.WriteString(typed)
case []string:
for _, item := range typed {
out.WriteString(item)
}
case []any:
for _, item := range typed {
appendTokenEstimateText(out, item)
}
case map[string]any:
for _, key := range []string{"text", "content", "query", "document"} {
if text := stringFromAny(typed[key]); text != "" {
out.WriteString(text)
}
}
}
}
func tokenUsageAmounts(usage clients.Usage) map[string]float64 {

View File

@ -40,8 +40,11 @@ func (s *Service) Estimate(ctx context.Context, kind string, model string, body
}
func (s *Service) estimatedBillings(ctx context.Context, user *auth.User, kind string, body map[string]any, candidate store.RuntimeModelCandidate) []any {
usage := clients.Usage{InputTokens: estimateRequestTokens(body), OutputTokens: int(floatFromAny(body["max_tokens"]))}
if usage.OutputTokens == 0 {
usage := clients.Usage{InputTokens: estimateRequestTokens(body)}
if isTextGenerationKind(kind) {
usage.OutputTokens = int(floatFromAny(body["max_tokens"]))
}
if isTextGenerationKind(kind) && usage.OutputTokens == 0 {
usage.OutputTokens = 64
}
usage.TotalTokens = usage.InputTokens + usage.OutputTokens
@ -56,19 +59,25 @@ func (s *Service) estimatedBillings(ctx context.Context, user *auth.User, kind s
func (s *Service) billings(ctx context.Context, user *auth.User, kind string, body map[string]any, candidate store.RuntimeModelCandidate, response clients.Response, simulated bool) []any {
config := s.effectiveBillingConfig(ctx, candidate)
discount := effectiveDiscount(ctx, s.store, user, candidate)
if isTextGenerationKind(kind) {
if isTextBillingKind(kind) {
inputTokens := response.Usage.InputTokens
outputTokens := response.Usage.OutputTokens
if isTextInputOnlyKind(kind) && inputTokens == 0 && response.Usage.TotalTokens > 0 {
inputTokens = response.Usage.TotalTokens
}
if inputTokens == 0 && outputTokens == 0 {
inputTokens = estimateRequestTokens(body)
outputTokens = 1
if isTextGenerationKind(kind) {
outputTokens = 1
}
}
inputAmount := roundPrice(float64(inputTokens) / 1000 * resourcePrice(config, "text", "textInputPer1k", "inputTokenPrice", "basePrice") * discount)
outputAmount := roundPrice(float64(outputTokens) / 1000 * resourcePrice(config, "text", "textOutputPer1k", "outputTokenPrice", "basePrice") * discount)
return []any{
billingLine(candidate, "text_input", "1k_tokens", inputTokens, inputAmount, discount, simulated),
billingLine(candidate, "text_output", "1k_tokens", outputTokens, outputAmount, discount, simulated),
lines := []any{billingLine(candidate, "text_input", "1k_tokens", inputTokens, inputAmount, discount, simulated)}
if isTextGenerationKind(kind) {
outputAmount := roundPrice(float64(outputTokens) / 1000 * resourcePrice(config, "text", "textOutputPer1k", "outputTokenPrice", "basePrice") * discount)
lines = append(lines, billingLine(candidate, "text_output", "1k_tokens", outputTokens, outputAmount, discount, simulated))
}
return lines
}
count := requestOutputCount(body)
resource := "image"

View File

@ -698,7 +698,7 @@ func (s *Service) recordTaskParameterPreprocessing(ctx context.Context, taskID s
func skipTaskParameterPreprocessingLog(modelType string) bool {
switch strings.TrimSpace(modelType) {
case "text_generate", "chat", "responses", "text":
case "text_generate", "text_embedding", "text_rerank", "chat", "responses", "text":
return true
default:
return false
@ -923,6 +923,10 @@ func modelTypeFromKind(kind string, body map[string]any) string {
switch kind {
case "chat.completions", "responses":
return "text_generate"
case "embeddings":
return "text_embedding"
case "reranks":
return "text_rerank"
case "images.generations", "images.edits":
if kind == "images.edits" {
return "image_edit"
@ -943,7 +947,7 @@ func modelTypeFromKind(kind string, body map[string]any) string {
func requestedModelTypeFromBody(body map[string]any) string {
for _, key := range []string{"modelType", "model_type", "capability", "capabilityType"} {
value := strings.TrimSpace(stringFromMap(body, key))
value := canonicalModelType(strings.TrimSpace(stringFromMap(body, key)))
if isKnownModelType(value) {
return value
}
@ -951,9 +955,21 @@ func requestedModelTypeFromBody(body map[string]any) string {
return ""
}
func canonicalModelType(value string) string {
normalized := strings.ReplaceAll(strings.ToLower(strings.TrimSpace(value)), "-", "_")
switch normalized {
case "embedding":
return "text_embedding"
case "rerank", "reranks":
return "text_rerank"
default:
return normalized
}
}
func isKnownModelType(value string) bool {
switch value {
case "text_generate", "image_generate", "image_edit", "video_generate", "image_to_video", "text_to_video", "video_edit", "video_reference", "video_first_last_frame", "omni_video", "omni":
case "text_generate", "text_embedding", "text_rerank", "image_generate", "image_edit", "video_generate", "image_to_video", "text_to_video", "video_edit", "video_reference", "video_first_last_frame", "omni_video", "omni":
return true
default:
return false
@ -1001,6 +1017,14 @@ func isTextGenerationKind(kind string) bool {
return kind == "chat.completions" || kind == "responses"
}
func isTextInputOnlyKind(kind string) bool {
return kind == "embeddings" || kind == "reranks"
}
func isTextBillingKind(kind string) bool {
return isTextGenerationKind(kind) || isTextInputOnlyKind(kind)
}
func isSimulation(task store.GatewayTask, candidate store.RuntimeModelCandidate) bool {
if task.RunMode == "simulation" {
return true
@ -1115,6 +1139,17 @@ func validateRequest(kind string, body map[string]any) error {
if body["input"] == nil && body["messages"] == nil {
return errors.New("input or messages is required")
}
case "embeddings":
if body["input"] == nil {
return errors.New("input is required")
}
case "reranks":
if body["query"] == nil {
return errors.New("query is required")
}
if !hasRerankDocuments(body["documents"]) {
return errors.New("documents is required")
}
case "images.generations", "images.edits":
if strings.TrimSpace(stringFromMap(body, "prompt")) == "" {
return errors.New("prompt is required")
@ -1123,6 +1158,17 @@ func validateRequest(kind string, body map[string]any) error {
return nil
}
func hasRerankDocuments(value any) bool {
switch typed := value.(type) {
case []any:
return len(typed) > 0
case []string:
return len(typed) > 0
default:
return false
}
}
func parameterPreprocessClientError(err error) *clients.ClientError {
if err == nil {
return nil

View File

@ -47,7 +47,7 @@ func billingResourcesForModelTypes(modelTypes []string) map[string]bool {
resources := map[string]bool{}
for _, modelType := range modelTypes {
switch normalizeBillingType(modelType) {
case "chat", "text", "responses", "text_generate", "text_embedding", "embedding",
case "chat", "text", "responses", "text_generate", "text_embedding", "embedding", "text_rerank", "rerank",
"image_analysis", "video_understanding", "audio_understanding", "omni", "tools_call":
resources["text"] = true
case "image", "images.generations", "image_generate":

View File

@ -0,0 +1,253 @@
WITH aliyun_model_defs AS (
SELECT
'aliyun-bailian-openai' AS provider_key,
'aliyun-bailian-openai:text-embedding-v4' AS canonical_model_key,
'text-embedding-v4' AS provider_model_name,
'Qwen3-Embedding-v4' AS display_name,
'Qwen3-Embedding-v4' AS model_alias,
'Qwen3-Embedding 系列,默认维度 1024最长输入 8192 tokens支持 100+ 语种与多种编程语言。' AS description,
'https://static.51easyai.com/qwen-color.webp' AS icon_path,
jsonb_build_array('text_embedding') AS model_type,
jsonb_build_object(
'text_embedding', jsonb_build_object(
'dimensions', jsonb_build_array(2048, 1536, 1024, 768, 512, 256, 128, 64),
'defaultDimension', 1024,
'maxRows', 10,
'maxTokensPerRow', 8192
),
'originalTypes', jsonb_build_array('text_embedding')
) AS capabilities
UNION ALL
SELECT
'aliyun-bailian-openai',
'aliyun-bailian-openai:text-embedding-v3',
'text-embedding-v3',
'Qwen3-Embedding-v3',
'Qwen3-Embedding-v3',
'Qwen3-Embedding 系列,默认维度 1024支持中文、英文及 50+ 主流语种。',
'https://static.51easyai.com/qwen-color.webp',
jsonb_build_array('text_embedding'),
jsonb_build_object(
'text_embedding', jsonb_build_object(
'dimensions', jsonb_build_array(1024, 768, 512, 256, 128, 64),
'defaultDimension', 1024
),
'originalTypes', jsonb_build_array('text_embedding')
)
UNION ALL
SELECT
'aliyun-bailian-openai',
'aliyun-bailian-openai:text-embedding-v2',
'text-embedding-v2',
'Text-Embedding-v2',
'Text-Embedding-v2',
'固定维度 1536最长输入 2048 tokens支持中英西法葡印尼日韩德俄等语种。',
'https://static.51easyai.com/qwen-color.webp',
jsonb_build_array('text_embedding'),
jsonb_build_object(
'text_embedding', jsonb_build_object(
'dimensions', jsonb_build_array(1536),
'defaultDimension', 1536,
'maxRows', 25,
'maxTokensPerRow', 2048
),
'originalTypes', jsonb_build_array('text_embedding')
)
UNION ALL
SELECT
'aliyun-bailian-openai',
'aliyun-bailian-openai:text-embedding-v1',
'text-embedding-v1',
'Text-Embedding-v1',
'Text-Embedding-v1',
'百炼 OpenAI 兼容渠道 embedding 基础模型。',
'https://static.51easyai.com/qwen-color.webp',
jsonb_build_array('text_embedding'),
jsonb_build_object('originalTypes', jsonb_build_array('text_embedding'))
UNION ALL
SELECT
'aliyun-bailian-openai',
'aliyun-bailian-openai:qwen3-rerank',
'qwen3-rerank',
'Qwen3-Rerank',
'Qwen3-Rerank',
'阿里云百炼 OpenAI 兼容重排序模型,支持 100+ 语种,适用于语义文本搜索和 RAG。',
'https://static.51easyai.com/qwen-color.webp',
jsonb_build_array('text_rerank'),
jsonb_build_object(
'text_rerank', jsonb_build_object(
'maxDocuments', 500,
'maxTokensPerDocument', 4000,
'maxRequestTokens', 120000,
'supportTopN', true,
'supportInstruct', true
),
'originalTypes', jsonb_build_array('text_rerank')
)
),
source_rows AS (
SELECT
providers.id AS provider_id,
defs.provider_key,
defs.canonical_model_key,
defs.provider_model_name,
defs.model_type,
defs.display_name,
defs.model_alias,
defs.description,
defs.icon_path,
defs.capabilities,
COALESCE(template.base_billing_config, '{"text":{"basePrice":0.01,"baseWeight":1}}'::jsonb) AS base_billing_config,
COALESCE(template.default_rate_limit_policy, '{}'::jsonb) AS default_rate_limit_policy,
COALESCE(
template.pricing_rule_set_id,
(SELECT id FROM model_pricing_rule_sets WHERE rule_set_key = 'default-multimodal-v1' LIMIT 1)
) AS pricing_rule_set_id,
COALESCE(
template.runtime_policy_set_id,
(SELECT id FROM model_runtime_policy_sets WHERE policy_key = 'default-runtime-v1' LIMIT 1)
) AS runtime_policy_set_id,
COALESCE(template.runtime_policy_override, '{}'::jsonb) AS runtime_policy_override,
COALESCE(template.pricing_version, 1) AS pricing_version
FROM aliyun_model_defs defs
LEFT JOIN model_catalog_providers providers
ON providers.provider_key = defs.provider_key
OR providers.provider_code = defs.provider_key
LEFT JOIN base_model_catalog template
ON template.canonical_model_key = 'aliyun-bailian-openai:text-embedding-v4'
),
payload AS (
SELECT
source_rows.*,
jsonb_build_object(
'source', 'aliyun.model-studio.docs',
'sourceProviderCode', provider_key,
'sourceProviderName', '阿里云百炼(OpenAI兼容',
'sourceSpecType', 'openai',
'originalTypes', model_type,
'alias', model_alias,
'description', description,
'iconPath', icon_path,
'billingType', 'external-api',
'billingMode', '',
'referenceModel', '',
'modelWeight', NULL,
'selectable', true,
'rawModel', jsonb_build_object(
'name', provider_model_name,
'types', model_type,
'icon_path', icon_path,
'alias', model_alias,
'description', description,
'capabilities', capabilities - 'originalTypes'
)
) AS metadata
FROM source_rows
),
snapshot AS (
SELECT
payload.*,
jsonb_build_object(
'providerKey', provider_key,
'canonicalModelKey', canonical_model_key,
'providerModelName', provider_model_name,
'modelType', model_type,
'modelAlias', model_alias,
'capabilities', capabilities,
'baseBillingConfig', base_billing_config,
'defaultRateLimitPolicy', default_rate_limit_policy,
'pricingRuleSetId', COALESCE(pricing_rule_set_id::text, ''),
'runtimePolicySetId', COALESCE(runtime_policy_set_id::text, ''),
'runtimePolicyOverride', runtime_policy_override,
'metadata', metadata,
'pricingVersion', pricing_version,
'status', 'active'
) AS default_snapshot
FROM payload
)
INSERT INTO base_model_catalog (
provider_id,
provider_key,
canonical_model_key,
provider_model_name,
model_type,
display_name,
capabilities,
base_billing_config,
default_rate_limit_policy,
pricing_rule_set_id,
runtime_policy_set_id,
runtime_policy_override,
metadata,
catalog_type,
default_snapshot,
pricing_version,
status
)
SELECT
provider_id,
provider_key,
canonical_model_key,
provider_model_name,
model_type,
display_name,
capabilities,
base_billing_config,
default_rate_limit_policy,
pricing_rule_set_id,
runtime_policy_set_id,
runtime_policy_override,
metadata,
'system',
default_snapshot,
pricing_version,
'active'
FROM snapshot
ON CONFLICT (canonical_model_key) DO UPDATE
SET provider_id = EXCLUDED.provider_id,
provider_key = EXCLUDED.provider_key,
provider_model_name = CASE
WHEN base_model_catalog.customized_at IS NULL THEN EXCLUDED.provider_model_name
ELSE base_model_catalog.provider_model_name
END,
model_type = CASE
WHEN base_model_catalog.customized_at IS NULL THEN EXCLUDED.model_type
ELSE base_model_catalog.model_type
END,
display_name = CASE
WHEN base_model_catalog.customized_at IS NULL THEN EXCLUDED.display_name
ELSE base_model_catalog.display_name
END,
capabilities = CASE
WHEN base_model_catalog.customized_at IS NULL THEN EXCLUDED.capabilities
ELSE base_model_catalog.capabilities
END,
base_billing_config = CASE
WHEN base_model_catalog.customized_at IS NULL THEN EXCLUDED.base_billing_config
ELSE base_model_catalog.base_billing_config
END,
default_rate_limit_policy = CASE
WHEN base_model_catalog.customized_at IS NULL THEN EXCLUDED.default_rate_limit_policy
ELSE base_model_catalog.default_rate_limit_policy
END,
pricing_rule_set_id = COALESCE(base_model_catalog.pricing_rule_set_id, EXCLUDED.pricing_rule_set_id),
runtime_policy_set_id = COALESCE(base_model_catalog.runtime_policy_set_id, EXCLUDED.runtime_policy_set_id),
runtime_policy_override = CASE
WHEN base_model_catalog.customized_at IS NULL THEN EXCLUDED.runtime_policy_override
ELSE base_model_catalog.runtime_policy_override
END,
metadata = CASE
WHEN base_model_catalog.customized_at IS NULL THEN EXCLUDED.metadata
ELSE base_model_catalog.metadata
END,
catalog_type = 'system',
default_snapshot = EXCLUDED.default_snapshot,
pricing_version = CASE
WHEN base_model_catalog.customized_at IS NULL THEN EXCLUDED.pricing_version
ELSE base_model_catalog.pricing_version
END,
status = CASE
WHEN base_model_catalog.customized_at IS NULL THEN 'active'
ELSE base_model_catalog.status
END,
updated_at = now();

View File

@ -540,7 +540,7 @@ export function App() {
try {
const response = await createApiKey(token, {
name: apiKeyForm.name,
scopes: ['chat', 'image', 'video'],
scopes: ['chat', 'embedding', 'rerank', 'image', 'video'],
expiresAt: apiKeyForm.expiresAt ? new Date(apiKeyForm.expiresAt).toISOString() : undefined,
});
setApiKeySecret(response.secret);
@ -907,11 +907,20 @@ export function App() {
async function submitTask(event: FormEvent<HTMLFormElement>) {
event.preventDefault();
const credential = apiKeySecret || token;
const selectedApiKeySecret = selectedPlaygroundApiKeyId ? apiKeySecretsById[selectedPlaygroundApiKeyId] ?? '' : '';
const fallbackApiKeySecret = apiKeys.find((item) => Boolean(apiKeySecretsById[item.id]))?.id;
const credential = selectedApiKeySecret || (fallbackApiKeySecret ? apiKeySecretsById[fallbackApiKeySecret] : '') || apiKeySecret || token;
const credentialLabel = selectedApiKeySecret || fallbackApiKeySecret || apiKeySecret ? '本地 API Key' : '当前 Access Token';
setCoreState('loading');
setCoreMessage('');
try {
const response = await runTask(credential, taskForm);
if (response.localOnly) {
setTaskResult(response.task);
setCoreState('ready');
setCoreMessage(`${taskForm.kind} 已通过 ${credentialLabel} 完成 simulation。`);
return;
}
const syncTask = (detail: GatewayTask) => {
setTaskResult(detail);
setTasks((current) => [detail, ...current.filter((item) => item.id !== detail.id)]);
@ -921,7 +930,7 @@ export function App() {
setTasks((current) => [detail, ...current.filter((item) => item.id !== detail.id)]);
invalidateDataKeys('tasks', 'wallet', 'walletTransactions');
setCoreState('ready');
setCoreMessage(`${taskForm.kind} 已通过 ${apiKeySecret ? '本地 API Key' : '当前 Access Token'} 完成 simulation。`);
setCoreMessage(`${taskForm.kind} 已通过 ${credentialLabel} 完成 simulation。`);
} catch (err) {
setCoreState('error');
setCoreMessage(err instanceof Error ? err.message : '测试任务失败');
@ -1170,12 +1179,16 @@ export function App() {
{activePage === 'docs' && (
<ApiDocsPage
activeDocSection={apiDocSection}
apiKeySecret={apiKeySecret}
apiKeySecretsById={apiKeySecretsById}
apiKeys={apiKeys}
canRun={isAuthenticated}
coreMessage={coreMessage}
coreState={coreState}
selectedApiKeyId={selectedPlaygroundApiKeyId}
taskForm={taskForm}
taskResult={taskResult}
onApiKeyChange={setSelectedPlaygroundApiKeyId}
onCreateApiKey={openApiKeyCreation}
onDocSectionChange={navigateApiDocSection}
onLogin={showLogin}
onSubmitTask={submitTask}
@ -1318,7 +1331,8 @@ function dataKeysForRoute(
? ['modelCatalog']
: ['publicCatalog'];
}
if (activePage === 'home' || activePage === 'docs') return [];
if (activePage === 'docs') return isAuthenticated ? ['playgroundApiKeys'] : [];
if (activePage === 'home') return [];
if (!isAuthenticated) return [];
if (activePage === 'workspace') {

View File

@ -547,6 +547,39 @@ export async function createChatTask(
});
}
export async function createCompatibleChatCompletion(
token: string,
input: { model: string; messages: Array<Record<string, unknown>>; runMode?: string; simulation?: boolean; stream?: boolean },
): Promise<Record<string, unknown>> {
return request<Record<string, unknown>>('/v1/chat/completions', {
body: input,
method: 'POST',
token,
});
}
export async function createEmbedding(
token: string,
input: { model: string; input: string | string[]; dimensions?: number; runMode?: string; simulation?: boolean },
): Promise<Record<string, unknown>> {
return request<Record<string, unknown>>('/v1/embeddings', {
body: input,
method: 'POST',
token,
});
}
export async function createRerank(
token: string,
input: { model: string; query: string; documents: string[]; top_n?: number; runMode?: string; simulation?: boolean },
): Promise<Record<string, unknown>> {
return request<Record<string, unknown>>('/v1/reranks', {
body: input,
method: 'POST',
token,
});
}
export async function streamChatCompletions(
token: string,
input: { model: string; messages: Array<Record<string, unknown>>; simulation?: boolean },

View File

@ -1,8 +1,45 @@
import type { GatewayTask } from '@easyai-ai-gateway/contracts';
import { createChatTask, createImageEditTask, createImageGenerationTask } from '../api';
import { createCompatibleChatCompletion, createEmbedding, createImageEditTask, createImageGenerationTask, createRerank } from '../api';
import type { TaskForm } from '../types';
export function runTask(token: string, task: TaskForm): Promise<{ task: GatewayTask; next: Record<string, string> }> {
export interface RunTaskResponse {
localOnly?: boolean;
next?: Record<string, string>;
task: GatewayTask;
}
export async function runTask(token: string, task: TaskForm): Promise<RunTaskResponse> {
if (task.kind === 'chat.completions') {
const result = await createCompatibleChatCompletion(token, {
model: task.model,
runMode: 'simulation',
simulation: true,
stream: false,
messages: [{ role: 'user', content: task.prompt }],
});
return { localOnly: true, task: compatibleTask(task, result) };
}
if (task.kind === 'embeddings') {
const result = await createEmbedding(token, {
model: task.model,
input: embeddingInput(task.prompt),
dimensions: task.dimensions,
runMode: 'simulation',
simulation: true,
});
return { localOnly: true, task: compatibleTask(task, result) };
}
if (task.kind === 'reranks') {
const result = await createRerank(token, {
model: task.model,
query: task.prompt,
documents: rerankDocuments(task.documents),
top_n: task.topN,
runMode: 'simulation',
simulation: true,
});
return { localOnly: true, task: compatibleTask(task, result) };
}
if (task.kind === 'images.generations') {
return createImageGenerationTask(token, {
model: task.model,
@ -23,10 +60,78 @@ export function runTask(token: string, task: TaskForm): Promise<{ task: GatewayT
simulation: true,
});
}
return createChatTask(token, {
throw new Error(`Unsupported task kind: ${task.kind}`);
}
function compatibleTask(task: TaskForm, result: Record<string, unknown>): GatewayTask {
const now = new Date().toISOString();
return {
id: `docs-${task.kind}-${Date.now()}`,
asyncMode: false,
createdAt: now,
finishedAt: now,
kind: task.kind,
model: task.model,
modelType: modelTypeForKind(task.kind),
request: requestSnapshot(task),
result,
runMode: 'simulation',
status: 'succeeded',
updatedAt: now,
userId: 'docs-runner',
};
}
function requestSnapshot(task: TaskForm): Record<string, unknown> {
if (task.kind === 'embeddings') {
return {
model: task.model,
input: embeddingInput(task.prompt),
dimensions: task.dimensions,
runMode: 'simulation',
simulation: true,
};
}
if (task.kind === 'reranks') {
return {
model: task.model,
query: task.prompt,
documents: rerankDocuments(task.documents),
top_n: task.topN,
runMode: 'simulation',
simulation: true,
};
}
return {
model: task.model,
messages: [{ role: 'user', content: task.prompt }],
runMode: 'simulation',
simulation: true,
messages: [{ role: 'user', content: task.prompt }],
});
stream: false,
};
}
function embeddingInput(prompt: string) {
const lines = splitLines(prompt);
return lines.length > 1 ? lines : (lines[0] ?? prompt);
}
function rerankDocuments(value?: string) {
const documents = splitLines(value ?? '');
return documents.length ? documents : ['AI Gateway 提供 OpenAI 兼容接口。', '图片生成任务支持异步队列。'];
}
function splitLines(value: string) {
return value
.split(/\n+/)
.map((item) => item.trim())
.filter(Boolean);
}
function modelTypeForKind(kind: TaskForm['kind']) {
if (kind === 'embeddings') return 'text_embedding';
if (kind === 'reranks') return 'text_rerank';
if (kind === 'images.generations') return 'image_generate';
if (kind === 'images.edits') return 'image_edit';
return 'text_generate';
}

View File

@ -57,7 +57,7 @@ export const adminPages = [
export const apiDocPages = [
{ title: '鉴权与限流', path: '/docs/auth', description: '本地账号、JWT、OpenAPI Key、TPM/RPM/并发限制和错误码。' },
{ title: 'Chat / Responses', path: '/docs/api/chat', description: '对话、stream、结构化输出、取消请求和示例代码。' },
{ title: 'Chat / Embeddings / Reranks', path: '/docs/api/chat', description: '对话、文本向量、重排序、鉴权和示例代码。' },
{ title: '图片 / 视频', path: '/docs/api/media', description: '生图、图像编辑、生视频、任务进度和结果取回。' },
{ title: '在线调用测试', path: '/docs/playground', description: '选择模型和 API Key编辑参数查看实时响应和 billings。' },
];

View File

@ -1,33 +1,52 @@
import { useMemo, type FormEvent } from 'react';
import type { GatewayTask } from '@easyai-ai-gateway/contracts';
import { BookOpen, Play, Search, Send } from 'lucide-react';
import { useEffect, useMemo, type FormEvent } from 'react';
import type { GatewayApiKey, GatewayTask } from '@easyai-ai-gateway/contracts';
import { BookOpen, KeyRound, Play, Search, Send } from 'lucide-react';
import { Badge, Button, Select, Textarea } from '../components/ui';
import type { ApiDocSection, LoadState, TaskForm } from '../types';
import type { ApiDocSection, LoadState, TaskForm, TaskKind } from '../types';
import { ApiKeySelect, apiKeyNoticeText, resolveSelectedApiKeyId } from './playground-shared';
const docs: Array<{ key: ApiDocSection; group: string; method: string; path: string; title: string }> = [
{ key: 'chat', group: '聊天(Chat)', method: 'POST', path: '/v1/chat/completions', title: 'Chat(聊天)' },
{ key: 'imageGeneration', group: '图片', method: 'POST', path: '/v1/images/generations', title: '创建图片' },
{ key: 'imageEdit', group: '图片', method: 'POST', path: '/v1/images/edits', title: '编辑图片' },
{ key: 'pricing', group: '计费', method: 'POST', path: '/api/v1/pricing/estimate', title: '价格预估' },
{ key: 'files', group: '文件', method: 'POST', path: '/v1/files/upload', title: '上传文件' },
interface ApiDocItem {
group: string;
key: ApiDocSection;
kind?: TaskKind;
lead: string;
method: string;
path: string;
title: string;
}
const docs: ApiDocItem[] = [
{ key: 'chat', group: '文本', kind: 'chat.completions', method: 'POST', path: '/v1/chat/completions', title: 'Chat Completions', lead: 'OpenAI 兼容的对话接口,支持本地 API Key 授权、simulation 测试和非流式/流式响应。' },
{ key: 'embeddings', group: '文本', kind: 'embeddings', method: 'POST', path: '/v1/embeddings', title: '文本向量 Embeddings', lead: 'OpenAI 兼容的文本向量接口,可直接用 input 数组或字符串生成 embeddingAPI Key 需要 embedding 权限。' },
{ key: 'reranks', group: '文本', kind: 'reranks', method: 'POST', path: '/v1/reranks', title: '文本重排序 Reranks', lead: 'OpenAI 风格的重排序接口,传入 query 和 documents 后返回 relevance_scoreAPI Key 需要 rerank 权限。' },
{ key: 'imageGeneration', group: '图片', kind: 'images.generations', method: 'POST', path: '/v1/images/generations', title: '创建图片', lead: 'OpenAI 兼容的图片生成接口,支持 prompt、size、quality 和 simulation 测试。' },
{ key: 'imageEdit', group: '图片', kind: 'images.edits', method: 'POST', path: '/v1/images/edits', title: '编辑图片', lead: 'OpenAI 兼容的图片编辑接口,支持 image、mask、prompt 和 simulation 测试。' },
{ key: 'pricing', group: '计费', method: 'POST', path: '/api/v1/pricing/estimate', title: '价格预估', lead: '按请求体估算输入输出 token、模型倍率和折扣后的预估费用。' },
{ key: 'files', group: '文件', method: 'POST', path: '/v1/files/upload', title: '上传文件', lead: '上传在线测试所需的图片、音频或视频资源,后续请求可复用返回的文件 URL。' },
];
const guideItems = ['获取 Base URL 和 API Key', '通知设置-WebHook 参数介绍', '错误码', '测试模式'];
const taskKindOptions = [
['chat.completions', 'Chat'],
['embeddings', '文本向量'],
['reranks', '重排序'],
['images.generations', '生图'],
['images.edits', '图像编辑'],
] as const;
export function ApiDocsPage(props: {
activeDocSection: ApiDocSection;
apiKeySecret: string;
apiKeySecretsById: Record<string, string>;
apiKeys: GatewayApiKey[];
canRun: boolean;
coreMessage: string;
coreState: LoadState;
selectedApiKeyId: string;
taskForm: TaskForm;
taskResult: GatewayTask | null;
onApiKeyChange: (apiKeyId: string) => void;
onCreateApiKey: () => void;
onLogin: () => void;
onDocSectionChange: (value: ApiDocSection) => void;
onSubmitTask: (event: FormEvent<HTMLFormElement>) => void;
@ -35,8 +54,16 @@ export function ApiDocsPage(props: {
}) {
const current = docs.find((item) => item.key === props.activeDocSection) ?? docs[0];
const isFileDoc = current.key === 'files';
const apiKeyNotice = apiKeyNoticeText(props.apiKeys, props.apiKeySecretsById);
const activeApiKeyId = resolveSelectedApiKeyId(props.apiKeys, props.apiKeySecretsById, props.selectedApiKeyId);
const bodyExample = useMemo(() => requestBodyExample(props.taskForm), [props.taskForm]);
useEffect(() => {
if (current.kind && props.taskForm.kind !== current.kind) {
props.onTaskFormChange(defaultTaskForKind(current.kind, props.taskForm));
}
}, [current.kind, props.taskForm.kind]);
function handleSubmit(event: FormEvent<HTMLFormElement>) {
if (!props.canRun) {
event.preventDefault();
@ -46,6 +73,21 @@ export function ApiDocsPage(props: {
props.onSubmitTask(event);
}
function handleDocClick(item: ApiDocItem) {
if (item.kind) {
props.onTaskFormChange(defaultTaskForKind(item.kind, props.taskForm));
}
props.onDocSectionChange(item.key);
}
function handleKindChange(kind: TaskKind) {
props.onTaskFormChange(defaultTaskForKind(kind, props.taskForm));
const nextSection = docSectionForKind(kind);
if (nextSection !== props.activeDocSection) {
props.onDocSectionChange(nextSection);
}
}
return (
<div className="apiDocsShell">
<aside className="docsSidebar">
@ -66,7 +108,7 @@ export function ApiDocsPage(props: {
active: item.key === props.activeDocSection,
method: item.method,
title: item.title,
onClick: () => props.onDocSectionChange(item.key),
onClick: () => handleDocClick(item),
}))}
/>
))}
@ -79,9 +121,7 @@ export function ApiDocsPage(props: {
<Badge variant="warning">{current.method}</Badge>
<code>{current.path}</code>
</div>
<p className="docsLead">
integration-platform / OpenAI API Key server-main token
</p>
<p className="docsLead">{current.lead}</p>
<section className="paramCard">
<header>
@ -90,7 +130,7 @@ export function ApiDocsPage(props: {
</header>
<ParamRow name="Content-Type" type="string" required value={isFileDoc ? 'multipart/form-data' : 'application/json'} />
<ParamRow name="Accept" type="string" required value="application/json" />
<ParamRow name="Authorization" type="string" value="Bearer {{YOUR_API_KEY}}" />
<ParamRow name="Authorization" type="string" required value="Bearer {{YOUR_API_KEY}},支持本地 API Key管理接口仍使用 JWT。向量需 embedding 权限,重排序需 rerank 权限。" />
</section>
<section className="paramCard">
@ -104,12 +144,9 @@ export function ApiDocsPage(props: {
<ParamRow name="source" type="string" value="上传来源标记" />
</>
) : (
<>
<ParamRow name="model" type="string" required value="模型 ID 或别名" />
<ParamRow name="messages / prompt" type="array|string" required value="对话消息或图片提示词" />
<ParamRow name="simulation" type="boolean" value="测试模式开关" />
<ParamRow name="stream" type="boolean" value="对话进度流式返回" />
</>
bodyParamRows(current.key).map((row) => (
<ParamRow key={row.name} {...row} />
))
)}
</section>
</main>
@ -127,9 +164,27 @@ export function ApiDocsPage(props: {
<Badge variant="warning">POST</Badge>
<span>{current.path}</span>
</div>
<label className="shLabel">
API Key
<ApiKeySelect
apiKeySecretsById={props.apiKeySecretsById}
apiKeys={props.apiKeys}
selectedApiKeyId={activeApiKeyId}
onApiKeyChange={props.onApiKeyChange}
/>
</label>
{apiKeyNotice && (
<div className="docsKeyNotice">
<span>{apiKeyNotice}</span>
<Button type="button" size="sm" variant="secondary" onClick={props.onCreateApiKey}>
<KeyRound size={14} />
Key
</Button>
</div>
)}
<label className="shLabel">
<Select value={props.taskForm.kind} onChange={(event) => props.onTaskFormChange(defaultTaskForKind(event.target.value as TaskForm['kind'], props.taskForm))}>
<Select value={props.taskForm.kind} onChange={(event) => handleKindChange(event.target.value as TaskKind)}>
{taskKindOptions.map(([value, label]) => (
<option value={value} key={value}>{label}</option>
))}
@ -198,16 +253,39 @@ function groupDocs(items: typeof docs) {
function defaultTaskForKind(kind: TaskForm['kind'], current: TaskForm): TaskForm {
if (kind === 'chat.completions') return { ...current, kind, model: 'gpt-4o-mini' };
if (kind === 'embeddings') {
return {
...current,
dimensions: current.dimensions ?? 4,
kind,
model: 'text-embedding-v4',
prompt: current.prompt || 'AI Gateway 提供 OpenAI 兼容接口。',
};
}
if (kind === 'reranks') {
return {
...current,
documents: current.documents ?? 'AI Gateway 提供 OpenAI 兼容接口。\n图片生成任务支持异步队列。\n文本向量接口返回 embedding 数组。',
kind,
model: 'qwen3-rerank',
prompt: current.prompt || 'OpenAI 兼容接口',
topN: current.topN ?? 2,
};
}
if (kind === 'images.edits') return { ...current, kind, image: current.image ?? 'https://example.com/source.png', mask: current.mask ?? 'https://example.com/mask.png', model: 'gpt-image-1' };
return { ...current, kind, model: 'gpt-image-1' };
}
function requestBodyExample(task: TaskForm) {
const body = task.kind === 'chat.completions'
? { model: task.model, messages: [{ role: 'user', content: task.prompt }], simulation: true, stream: true }
: task.kind === 'images.edits'
? { model: task.model, prompt: task.prompt, image: task.image, mask: task.mask, simulation: true }
: { model: task.model, prompt: task.prompt, quality: 'medium', simulation: true, size: '1024x1024' };
? { model: task.model, messages: [{ role: 'user', content: task.prompt }], runMode: 'simulation', simulation: true, stream: false }
: task.kind === 'embeddings'
? { model: task.model, input: embeddingInputExample(task.prompt), dimensions: task.dimensions ?? 4, runMode: 'simulation', simulation: true }
: task.kind === 'reranks'
? { model: task.model, query: task.prompt, documents: rerankDocumentsExample(task.documents), top_n: task.topN ?? 2, runMode: 'simulation', simulation: true }
: task.kind === 'images.edits'
? { model: task.model, prompt: task.prompt, image: task.image, mask: task.mask, runMode: 'simulation', simulation: true }
: { model: task.model, prompt: task.prompt, quality: 'medium', runMode: 'simulation', simulation: true, size: '1024x1024' };
return JSON.stringify(body, null, 2);
}
@ -219,15 +297,84 @@ function parseBody(value: string, current: TaskForm): TaskForm {
messages?: Array<{ content?: string }>;
model?: string;
prompt?: string;
input?: string | string[];
query?: string;
documents?: string[];
top_n?: number;
dimensions?: number;
};
return {
...current,
dimensions: numberOrCurrent(body.dimensions, current.dimensions),
documents: Array.isArray(body.documents) ? body.documents.join('\n') : current.documents,
image: body.image ?? current.image,
mask: body.mask ?? current.mask,
model: body.model ?? current.model,
prompt: body.prompt ?? body.messages?.[0]?.content ?? current.prompt,
prompt: body.prompt ?? body.query ?? inputText(body.input) ?? body.messages?.[0]?.content ?? current.prompt,
topN: numberOrCurrent(body.top_n, current.topN),
};
} catch {
return current;
}
}
function bodyParamRows(section: ApiDocSection) {
if (section === 'embeddings') {
return [
{ name: 'model', type: 'string', required: true, value: '模型 ID 或别名,例如 text-embedding-v4' },
{ name: 'input', type: 'string|array', required: true, value: '需要向量化的文本或文本数组' },
{ name: 'dimensions', type: 'number', value: '可选向量维度,需模型支持' },
{ name: 'runMode / simulation', type: 'string|boolean', value: '在线测试时使用 simulation不消耗真实上游额度' },
];
}
if (section === 'reranks') {
return [
{ name: 'model', type: 'string', required: true, value: '模型 ID 或别名,例如 qwen3-rerank' },
{ name: 'query', type: 'string', required: true, value: '用于相关性排序的查询文本' },
{ name: 'documents', type: 'array', required: true, value: '候选文档文本数组,至少 1 条' },
{ name: 'top_n', type: 'number', value: '返回前 N 条结果' },
{ name: 'runMode / simulation', type: 'string|boolean', value: '在线测试时使用 simulation不消耗真实上游额度' },
];
}
return [
{ name: 'model', type: 'string', required: true, value: '模型 ID 或别名' },
{ name: 'messages / prompt', type: 'array|string', required: true, value: '对话消息或图片提示词' },
{ name: 'simulation', type: 'boolean', value: '测试模式开关' },
{ name: 'stream', type: 'boolean', value: '对话进度流式返回' },
];
}
function docSectionForKind(kind: TaskKind): ApiDocSection {
if (kind === 'embeddings') return 'embeddings';
if (kind === 'reranks') return 'reranks';
if (kind === 'images.generations') return 'imageGeneration';
if (kind === 'images.edits') return 'imageEdit';
return 'chat';
}
function embeddingInputExample(prompt: string) {
const lines = splitLines(prompt);
return lines.length > 1 ? lines : (lines[0] ?? prompt);
}
function rerankDocumentsExample(value?: string) {
return splitLines(value ?? '').length
? splitLines(value ?? '')
: ['AI Gateway 提供 OpenAI 兼容接口。', '图片生成任务支持异步队列。'];
}
function splitLines(value: string) {
return value
.split(/\n+/)
.map((item) => item.trim())
.filter(Boolean);
}
function inputText(value: string | string[] | undefined) {
if (Array.isArray(value)) return value.join('\n');
return value;
}
function numberOrCurrent(value: unknown, current?: number) {
return typeof value === 'number' && Number.isFinite(value) ? value : current;
}

View File

@ -252,6 +252,7 @@ function modelTypeMatchesCapability(value: string, capability: string) {
function canonicalCapabilityValue(value: string) {
const type = value.trim().toLowerCase().replaceAll('-', '_');
if (type === 'embedding') return 'text_embedding';
if (type === 'rerank' || type === 'reranks') return 'text_rerank';
if (type === 'model') return 'model_3d';
return type;
}
@ -265,6 +266,7 @@ function capabilityValueForTag(tag: string) {
: 'reasoning',
: 'structured_output',
: 'digital_human',
: 'text_rerank',
'3D 模型': 'model_3d',
'文生 3D': 'text_to_model',
'图生 3D': 'image_to_model',

View File

@ -49,6 +49,7 @@ export type PlatformModelTypeDefinition = {
export const platformModelTypeDefinitions: PlatformModelTypeDefinition[] = [
{ key: 'text_generate', label: '文本生成', group: '文本', area: 'text' },
{ key: 'text_embedding', label: '文本向量', group: '文本', area: 'embedding' },
{ key: 'text_rerank', label: '重排序', group: '文本', area: 'text' },
{ key: 'image_generate', label: '图像生成', group: '图像', area: 'image' },
{ key: 'image_edit', label: '图像编辑', group: '图像', area: 'image' },
{ key: 'image_analysis', label: '图像理解', group: '理解', area: 'text' },
@ -469,7 +470,7 @@ function boolFrom(value: unknown) {
}
function looksLikeCapabilityType(key: string) {
return modelTypeDefinitionMap.has(key) || key.includes('_') || ['chat', 'image', 'video', 'audio', 'embedding', 'music', 'digital_human', 'model_3d'].includes(key);
return modelTypeDefinitionMap.has(key) || key.includes('_') || ['chat', 'image', 'video', 'audio', 'embedding', 'rerank', 'music', 'digital_human', 'model_3d'].includes(key);
}
function inferArea(type: string): CapabilityConfigArea {
@ -484,7 +485,7 @@ function inferArea(type: string): CapabilityConfigArea {
}
function isTextLike(type: string) {
return ['text_generate', 'image_analysis', 'video_understanding', 'audio_understanding', 'omni', 'tools_call'].includes(type);
return ['text_generate', 'text_rerank', 'image_analysis', 'video_understanding', 'audio_understanding', 'omni', 'tools_call'].includes(type);
}
function isImageLike(type: string) {

View File

@ -44,6 +44,8 @@ const adminPaths: Record<AdminSection, string> = {
const docsPaths: Record<ApiDocSection, string> = {
chat: '/docs/chat',
embeddings: '/docs/embeddings',
reranks: '/docs/reranks',
imageGeneration: '/docs/images/generations',
imageEdit: '/docs/images/edits',
pricing: '/docs/pricing',
@ -134,6 +136,8 @@ function parseAdminSection(path: string): AdminSection {
function parseDocSection(path: string): ApiDocSection {
if (path === '/docs') return 'chat';
if (path === '/docs/api/chat') return 'chat';
if (path === '/docs/api/embeddings') return 'embeddings';
if (path === '/docs/api/reranks') return 'reranks';
if (path === '/docs/api/media') return 'imageGeneration';
if (path === '/docs/playground') return 'chat';
return docsSections[path] ?? 'chat';

View File

@ -98,6 +98,22 @@
background: #fff;
}
.docsKeyNotice {
display: grid;
gap: 10px;
padding: 12px;
border: 1px solid #fde68a;
border-radius: 8px;
background: #fffbeb;
color: #92400e;
font-size: 0.8125rem;
line-height: 1.5;
}
.docsKeyNotice .shButton {
justify-self: start;
}
.docsLead {
margin: 18px 0 24px;
color: var(--text-soft);

View File

@ -1,10 +1,10 @@
export type LoadState = 'idle' | 'loading' | 'ready' | 'error';
export type AuthMode = 'login' | 'register' | 'external';
export type TaskKind = 'chat.completions' | 'images.generations' | 'images.edits';
export type TaskKind = 'chat.completions' | 'embeddings' | 'reranks' | 'images.generations' | 'images.edits';
export type PageKey = 'home' | 'playground' | 'models' | 'workspace' | 'admin' | 'docs';
export type PlaygroundMode = 'chat' | 'image' | 'video';
export type WorkspaceSection = 'overview' | 'billing' | 'apiKeys' | 'tasks' | 'transactions';
export type ApiDocSection = 'chat' | 'imageGeneration' | 'imageEdit' | 'pricing' | 'files';
export type ApiDocSection = 'chat' | 'embeddings' | 'reranks' | 'imageGeneration' | 'imageEdit' | 'pricing' | 'files';
export type AdminSection =
| 'overview'
| 'globalModels'
@ -37,8 +37,11 @@ export interface TaskForm {
kind: TaskKind;
model: string;
prompt: string;
dimensions?: number;
documents?: string;
image?: string;
mask?: string;
topN?: number;
}
export interface ApiKeyForm {