feat: 完善模型请求适配与输出限制

This commit is contained in:
2026-07-17 13:52:00 +08:00
parent a24eb1aeb0
commit 5ee267ecbd
31 changed files with 3287 additions and 232 deletions
+217 -73
View File
@@ -133,6 +133,118 @@ definitions:
usage:
$ref: '#/definitions/httpapi.ChatCompletionUsage'
type: object
httpapi.ChatCompletionRequest:
properties:
audio:
additionalProperties: true
type: object
frequency_penalty:
example: 0
type: number
function_call: {}
functions:
items:
additionalProperties: true
type: object
type: array
logit_bias:
additionalProperties: true
type: object
logprobs:
type: boolean
max_completion_tokens:
example: 512
type: integer
max_tokens:
example: 512
type: integer
messages:
items:
$ref: '#/definitions/httpapi.ChatMessage'
type: array
metadata:
additionalProperties: true
type: object
modalities:
items:
type: string
type: array
model:
example: gpt-4o-mini
type: string
moderation: {}
"n":
example: 1
type: integer
parallel_tool_calls:
type: boolean
prediction: {}
presence_penalty:
example: 0
type: number
prompt_cache_key:
type: string
prompt_cache_options:
additionalProperties: true
type: object
prompt_cache_retention:
enum:
- in_memory
- 24h
type: string
reasoning_effort:
description: ReasoningEffort 推理强度,OpenAI-compatible 请求字段;支持 none、minimal、low、medium、high、xhigh、max。供应商自定义取值由网关按平台适配。
enum:
- none
- minimal
- low
- medium
- high
- xhigh
- max
example: medium
type: string
response_format: {}
runMode:
example: simulation
type: string
safety_identifier:
type: string
seed:
type: integer
service_tier:
type: string
stop: {}
store:
type: boolean
stream:
example: false
type: boolean
stream_options:
additionalProperties: true
type: object
temperature:
example: 0.7
type: number
tool_choice: {}
tools:
items:
additionalProperties: true
type: object
type: array
top_logprobs:
type: integer
top_p:
example: 1
type: number
user:
type: string
verbosity:
type: string
web_search_options:
additionalProperties: true
type: object
type: object
httpapi.ChatCompletionUsage:
properties:
completion_tokens:
@@ -157,12 +269,16 @@ definitions:
type: object
httpapi.ChatMessage:
properties:
content:
example: Hello
content: {}
function_call: {}
name:
type: string
role:
example: user
type: string
tool_call_id:
type: string
tool_calls: {}
type: object
httpapi.ChatPromptTokensDetails:
properties:
@@ -203,9 +319,13 @@ definitions:
message:
example: invalid json body
type: string
param: {}
status:
example: 400
type: integer
type:
example: invalid_request_error
type: string
type: object
httpapi.FileStorageChannelListResponse:
properties:
@@ -567,6 +687,18 @@ definitions:
type: object
httpapi.ResponsesRequest:
properties:
background:
type: boolean
context_management:
items:
additionalProperties: true
type: object
type: array
conversation: {}
include:
items:
type: string
type: array
input: {}
instructions:
example: Answer concisely
@@ -574,23 +706,47 @@ definitions:
max_output_tokens:
example: 512
type: integer
max_tool_calls:
type: integer
metadata:
additionalProperties: true
type: object
model:
example: Doubao Seed 2.0 Pro
type: string
moderation: {}
parallel_tool_calls:
example: true
type: boolean
previous_response_id:
example: resp_0123456789abcdef0123456789abcdef
type: string
prompt: {}
prompt_cache_key:
type: string
prompt_cache_options:
additionalProperties: true
type: object
prompt_cache_retention:
enum:
- in_memory
- 24h
type: string
reasoning:
additionalProperties: true
type: object
safety_identifier:
type: string
service_tier:
type: string
store:
type: boolean
stream:
example: false
type: boolean
stream_options:
additionalProperties: true
type: object
temperature:
example: 0.7
type: number
@@ -603,9 +759,15 @@ definitions:
additionalProperties: true
type: object
type: array
top_logprobs:
type: integer
top_p:
example: 1
type: number
truncation:
type: string
user:
type: string
type: object
httpapi.RuntimePolicySetListResponse:
properties:
@@ -718,12 +880,17 @@ definitions:
emotion:
example: happy
type: string
input:
example: Tell me a short story
type: string
input: {}
makeInstrumental:
example: false
type: boolean
max_completion_tokens:
description: MaxCompletionTokens includes visible output and reasoning tokens.
example: 512
type: integer
max_output_tokens:
example: 512
type: integer
max_tokens:
example: 512
type: integer
@@ -744,7 +911,15 @@ definitions:
example: A watercolor robot reading a book
type: string
reasoning_effort:
description: ReasoningEffort 推理强度,OpenAI-compatible 请求字段;支持 none、minimal、low、medium、high、xhigh。供应商自定义取值由网关按平台适配。
description: ReasoningEffort 推理强度,OpenAI-compatible 请求字段;支持 none、minimal、low、medium、high、xhigh、max。供应商自定义取值由网关按平台适配。
enum:
- none
- minimal
- low
- medium
- high
- xhigh
- max
example: medium
type: string
resolution:
@@ -5460,7 +5635,7 @@ paths:
name: input
required: true
schema:
$ref: '#/definitions/httpapi.TaskRequest'
$ref: '#/definitions/httpapi.ChatCompletionRequest'
produces:
- application/json
- text/event-stream
@@ -6107,32 +6282,31 @@ paths:
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。
description: 公开 OpenAI-compatible Responses 入口。模型声明 openai_responses 时原生转发,否则使用
Chat Completions 转换;store 缺省为 true。previous_response_id 严格绑定首次成功的平台模型和上游协议,链路不可用时不跨平台续接。未提供
previous_response_id 时由调用方管理完整状态,Gateway 以本轮 input/messages 为准且不追加本地历史。
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
- description: Responses 请求;Chat 回退只支持自定义 function tools
in: body
name: input
required: true
schema:
$ref: '#/definitions/httpapi.TaskRequest'
$ref: '#/definitions/httpapi.ResponsesRequest'
produces:
- application/json
- text/event-stream
responses:
"200":
description: OK
headers:
X-Gateway-Task-Id:
description: 网关审计任务 ID
type: string
schema:
$ref: '#/definitions/httpapi.CompatibleResponse'
"202":
description: Accepted
schema:
$ref: '#/definitions/httpapi.TaskAcceptedResponse'
$ref: '#/definitions/httpapi.ResponsesCompatibleResponse'
"400":
description: Bad Request
description: invalid_previous_response_id / unsupported_response_tool /
unsupported_response_parameter
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"401":
@@ -6147,23 +6321,15 @@ paths:
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
"503":
description: response_chain_unavailable
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
security:
- BearerAuth: []
summary: 创建或执行 AI 任务
summary: 创建 OpenAI Responses
tags:
- tasks
- responses
/api/v1/song/generations:
post:
consumes:
@@ -7027,32 +7193,25 @@ paths:
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
description: OpenAI-compatible Chat Completions 入口;仅接受官方字段及文档声明的 EasyAI 路由扩展,未知顶层字段返回
400 invalid_parameter
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
- description: Chat Completions 请求
in: body
name: input
required: true
schema:
$ref: '#/definitions/httpapi.TaskRequest'
$ref: '#/definitions/httpapi.ChatCompletionRequest'
produces:
- application/json
- text/event-stream
responses:
"200":
description: OK
schema:
$ref: '#/definitions/httpapi.CompatibleResponse'
"202":
description: Accepted
schema:
$ref: '#/definitions/httpapi.TaskAcceptedResponse'
$ref: '#/definitions/httpapi.ChatCompletionCompatibleResponse'
"400":
description: Bad Request
description: invalid_parameter
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"401":
@@ -7067,10 +7226,6 @@ paths:
description: Forbidden
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"404":
description: Not Found
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"429":
description: Too Many Requests
schema:
@@ -7081,9 +7236,9 @@ paths:
$ref: '#/definitions/httpapi.ErrorEnvelope'
security:
- BearerAuth: []
summary: 创建或执行 AI 任务
summary: 创建 OpenAI Chat Completions
tags:
- tasks
- chat
/embeddings:
post:
consumes:
@@ -7866,32 +8021,25 @@ paths:
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
description: OpenAI-compatible Chat Completions 入口;仅接受官方字段及文档声明的 EasyAI 路由扩展,未知顶层字段返回
400 invalid_parameter
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
- description: Chat Completions 请求
in: body
name: input
required: true
schema:
$ref: '#/definitions/httpapi.TaskRequest'
$ref: '#/definitions/httpapi.ChatCompletionRequest'
produces:
- application/json
- text/event-stream
responses:
"200":
description: OK
schema:
$ref: '#/definitions/httpapi.CompatibleResponse'
"202":
description: Accepted
schema:
$ref: '#/definitions/httpapi.TaskAcceptedResponse'
$ref: '#/definitions/httpapi.ChatCompletionCompatibleResponse'
"400":
description: Bad Request
description: invalid_parameter
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"401":
@@ -7906,10 +8054,6 @@ paths:
description: Forbidden
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"404":
description: Not Found
schema:
$ref: '#/definitions/httpapi.ErrorEnvelope'
"429":
description: Too Many Requests
schema:
@@ -7920,9 +8064,9 @@ paths:
$ref: '#/definitions/httpapi.ErrorEnvelope'
security:
- BearerAuth: []
summary: 创建或执行 AI 任务
summary: 创建 OpenAI Chat Completions
tags:
- tasks
- chat
/v1/embeddings:
post:
consumes: