linkong/planet
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

feat: align aiprovider with adapter-based compatibility

Browse Source
This commit is contained in:
linkong
2026-04-07 17:54:44 +08:00
parent bc90e00e25
commit f12719914d
9 changed files with 299 additions and 50 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
TODO.md
View File

@@ -13,6 +13,9 @@
- [x] 接入 `IPtoASN / IPtoCountry` 作为 prefix-centric geography 的主数据源
- [x] 接入 `OpenGeoFeed` 作为 prefix geography 的高质量覆盖/override 数据源
- [x] 把 RIR delegated 设计成 prefix geography 的 fallback，而不是主来源
- [ ] 为 `aiprovider` 建立 `provider -> api adapter -> compat policy` 的配置中心，优先落成 `json` 或 `yaml` 文件，运行时按 `provider/model` 读取兼容设置，而不是把专项兼容继续散落在 Python 分支里
- [ ] 为市面上主流 AI 服务补专项兼容配置并固化到配置文件中，至少覆盖 `OpenAI / Anthropic / MiniMax / Ollama / Moonshot / DeepSeek / Qwen / GLM / Gemini / OpenRouter / vLLM / LM Studio / One API`
- [ ] 在兼容配置中补齐可声明项：`api adapter`、`base_url pattern`、`auth header`、`thinking default`、`reasoning block mapping`、`stream path`、`tool-call capability`、`multimodal capability`、`provider-specific request patch`
- [ ] 接入 `inetnum` / `inet6num` whois 作为比 RIR 更细粒度的后备层
- [x] 在 activity layer 之后继续补 `route leak` 和 `path instability / flap` detector
- [ ] 对 [frontend/public/earth/js/bgp.js](/home/ray/dev/linkong/planet/frontend/public/earth/js/bgp.js) 做按职责拆分的小重构，拆成 data / markers / overlays / animation，降低后续维护复杂度

48
aiprovider/.env.example
View File

@@ -6,29 +6,51 @@ AI_TIMEOUT_SECONDS=60
AI_HTTP_RETRY_ATTEMPTS=2
AI_ANALYSIS_SYSTEM_PROMPT=你是态势感知分析助手。请基于输入的上下文、观测与约束，输出结构化、克制、可执行的分析。
# Select one provider mode:
# - openai_compatible
# - claude_compatible
# Provider identity. Recommended values:
# - minimax
# - openai
# - ollama
AI_PROVIDER=ollama
# Compatibility aliases still accepted:
# - openai_compatible
# - anthropic_compatible
# - claude_compatible
AI_PROVIDER=minimax
# Request adapter style, following OpenClaw's API-seam pattern:
# - auto
# - openai-completions
# - anthropic-messages
# - ollama-generate
AI_PROVIDER_API=anthropic-messages
# Common model selection
AI_MODEL=qwen2.5:7b
AI_MODEL=MiniMax-M2.7
# MiniMax CN Anthropic-compatible example
AI_BASE_URL=https://api.minimaxi.com/anthropic
AI_API_KEY=sk-cp-change-me
AI_MAX_TOKENS=1200
AI_ANTHROPIC_VERSION=2023-06-01
# OpenAI-compatible example (vLLM / LM Studio / One API / local gateway)
# AI_PROVIDER=openai_compatible
# AI_PROVIDER=openai
# AI_PROVIDER_API=openai-completions
# AI_BASE_URL=http://127.0.0.1:8001/v1
# AI_API_KEY=local-key
# AI_MODEL=your-local-model
# Claude-compatible example (Anthropic / MiniMax / Claude-compatible gateway)
# AI_PROVIDER=claude_compatible
# AI_BASE_URL=http://127.0.0.1:8002
# Anthropic-compatible example (Claude-compatible gateway)
# AI_PROVIDER=anthropic
# AI_PROVIDER_API=anthropic-messages
# AI_BASE_URL=http://127.0.0.1:8002/anthropic
# AI_API_KEY=local-key
# AI_MODEL=your-model
# AI_MAX_TOKENS=1200
# AI_ANTHROPIC_VERSION=2023-06-01
# Ollama native example
AI_BASE_URL=http://127.0.0.1:11434
AI_API_KEY=
AI_MAX_TOKENS=1200
AI_ANTHROPIC_VERSION=2023-06-01
# AI_PROVIDER=ollama
# AI_PROVIDER_API=ollama-generate
# AI_BASE_URL=http://127.0.0.1:11434
# AI_API_KEY=
# AI_MODEL=qwen2.5:7b

44
aiprovider/README.md
View File

@@ -8,17 +8,27 @@
当前支持:
- `AI_PROVIDER=openai`
- `AI_PROVIDER=openai_compatible`
- `AI_PROVIDER=anthropic`
- `AI_PROVIDER=anthropic_compatible`
- `AI_PROVIDER=claude_compatible`
- `AI_PROVIDER=ollama`
- provider identity:
- `AI_PROVIDER=openai`
- `AI_PROVIDER=anthropic`
- `AI_PROVIDER=minimax`
- `AI_PROVIDER=ollama`
- request adapter:
- `AI_PROVIDER_API=openai-completions`
- `AI_PROVIDER_API=anthropic-messages`
- `AI_PROVIDER_API=ollama-generate`
兼容别名仍然保留:
- `openai_compatible`
- `anthropic_compatible`
- `claude_compatible`
典型配置:
```env
AI_PROVIDER=openai_compatible
AI_PROVIDER=openai
AI_PROVIDER_API=openai-completions
AI_BASE_URL=https://api.openai.com/v1
AI_API_KEY=your_api_key
AI_MODEL=gpt-4o-mini
@@ -26,13 +36,14 @@ AI_TIMEOUT_SECONDS=60
AI_PROVIDER_SERVICE_TOKEN=change_me
```
Claude 兼容供应商示例:
MiniMax 中国大陆节点示例:
```env
AI_PROVIDER=claude_compatible
AI_BASE_URL=https://your-claude-compatible-endpoint.example.com
AI_API_KEY=your_api_key
AI_MODEL=your-claude-compatible-model
AI_PROVIDER=minimax
AI_PROVIDER_API=anthropic-messages
AI_BASE_URL=https://api.minimaxi.com/anthropic
AI_API_KEY=sk-cp-xxxxx
AI_MODEL=MiniMax-M2.7
AI_TIMEOUT_SECONDS=60
AI_MAX_TOKENS=1200
AI_ANTHROPIC_VERSION=2023-06-01
@@ -43,12 +54,15 @@ AI_PROVIDER_SERVICE_TOKEN=change_me
- Anthropic 官方 Claude API
- Claude 兼容网关
- MiniMax 等提供 Claude/Anthropic 风格消息接口的服务
- MiniMax 等提供 Anthropic Messages 风格接口的服务
这套命名方式参考了 OpenClaw 的接入模式: provider 负责标识供应商, `AI_PROVIDER_API` 负责标识协议适配层, 避免把“供应商”和“协议”绑死在一起。
Ollama 原生示例:
```env
AI_PROVIDER=ollama
AI_PROVIDER_API=ollama-generate
AI_BASE_URL=http://127.0.0.1:11434
AI_API_KEY=
AI_MODEL=qwen2.5:7b
@@ -58,8 +72,8 @@ AI_PROVIDER_SERVICE_TOKEN=change_me
本地模型接入建议:
- `vLLM`、`LM Studio`、`One API`：优先使用 `openai_compatible`
- `MiniMax`、Claude 兼容网关：使用 `claude_compatible`
- `vLLM`、`LM Studio`、`One API`：`AI_PROVIDER=openai` + `AI_PROVIDER_API=openai-completions`
- `MiniMax`、Claude 兼容网关：`AI_PROVIDER=minimax|anthropic` + `AI_PROVIDER_API=anthropic-messages`
- `Ollama`：可直接使用 `ollama`
启动模板:

1
aiprovider/config.py
View File

@@ -9,6 +9,7 @@ class Settings(BaseSettings):
SERVICE_VERSION: str = "0.1.0"
AI_PROVIDER: str = "disabled"
AI_PROVIDER_API: str = "auto"
AI_BASE_URL: str = "https://api.openai.com/v1"
AI_API_KEY: str = ""
AI_MODEL: str = ""

150
aiprovider/provider_service.py
View File

@@ -8,6 +8,7 @@ from fastapi import HTTPException, status
from aiprovider.config import settings
from aiprovider.schemas import (
AIContentBlock,
AIProviderStatusResponse,
SituationalAnalysisRequest,
SituationalAnalysisResponse,
@@ -18,9 +19,39 @@ def _normalize_provider(value: str) -> str:
return (value or "disabled").strip().lower()
def _normalize_provider_api(value: str) -> str:
return (value or "auto").strip().lower().replace("_", "-")
def _resolve_provider_api(provider: str, configured_api: str) -> str:
if configured_api and configured_api != "auto":
return configured_api
if provider in {"openai", "openai-compatible", "openai_compatible"}:
return "openai-completions"
if provider in {
"anthropic",
"anthropic-compatible",
"anthropic_compatible",
"claude-compatible",
"claude_compatible",
"minimax",
"kimi-coding",
"moonshot-anthropic",
}:
return "anthropic-messages"
if provider == "ollama":
return "ollama-generate"
return "disabled"
class ProviderService:
def __init__(self) -> None:
self.provider = _normalize_provider(settings.AI_PROVIDER)
self.provider_api = _resolve_provider_api(
self.provider,
_normalize_provider_api(settings.AI_PROVIDER_API),
)
self.base_url = settings.AI_BASE_URL.rstrip("/")
self.api_key = settings.AI_API_KEY
self.default_model = settings.AI_MODEL
@@ -32,9 +63,11 @@ class ProviderService:
def get_status(self) -> AIProviderStatusResponse:
enabled = self.provider != "disabled"
configured = enabled and bool(self.base_url and self.api_key and self.default_model)
has_credentials = bool(self.api_key) if self._requires_api_key() else True
configured = enabled and bool(self.base_url and has_credentials and self.default_model)
return AIProviderStatusResponse(
provider=self.provider,
api=self.provider_api if enabled else None,
enabled=enabled,
configured=configured,
model=self.default_model or None,
@@ -49,7 +82,8 @@ class ProviderService:
)
model = payload.preferred_model or self.default_model
if not self.base_url or not self.api_key or not model:
has_credentials = bool(self.api_key) if self._requires_api_key() else True
if not self.base_url or not has_credentials or not model:
raise HTTPException(
status_code=status.HTTP_503_SERVICE_UNAVAILABLE,
detail="AI provider is not fully configured. Check AI_BASE_URL, AI_API_KEY, and AI_MODEL.",
@@ -57,28 +91,40 @@ class ProviderService:
prompt = self._build_prompt(payload)
if self.provider in {"openai", "openai_compatible"}:
if self.provider_api == "openai-completions":
data = await self._request_openai_compatible(model, prompt)
content = self._extract_openai_content(data)
elif self.provider in {"anthropic", "anthropic_compatible", "claude_compatible"}:
data = await self._request_anthropic_compatible(model, prompt)
content_blocks = self._extract_openai_blocks(data)
elif self.provider_api == "anthropic-messages":
data = await self._request_anthropic_messages(model, prompt, payload.thinking)
content = self._extract_anthropic_content(data)
elif self.provider == "ollama":
content_blocks = self._extract_anthropic_blocks(data)
elif self.provider_api == "ollama-generate":
data = await self._request_ollama(model, prompt)
content = self._extract_ollama_content(data)
content_blocks = self._extract_ollama_blocks(data)
else:
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=f"Unsupported AI provider: {self.provider}",
detail=f"Unsupported AI provider API: {self.provider_api}",
)
text_blocks = [block.text for block in content_blocks if block.text]
thinking_blocks = [block.thinking for block in content_blocks if block.thinking]
return SituationalAnalysisResponse(
provider=self.provider,
model=model,
content=content,
content_blocks=content_blocks,
text_blocks=text_blocks,
thinking_blocks=thinking_blocks,
raw_response=data,
)
def _requires_api_key(self) -> bool:
return self.provider_api != "ollama-generate"
def _build_prompt(self, payload: SituationalAnalysisRequest) -> str:
sections = [
f"任务标题:\n{payload.title}",
@@ -113,7 +159,12 @@ class ProviderService:
request_body=request_body,
)
async def _request_anthropic_compatible(self, model: str, prompt: str) -> dict[str, Any]:
async def _request_anthropic_messages(
self,
model: str,
prompt: str,
thinking: dict[str, Any] | None = None,
) -> dict[str, Any]:
request_body = {
"model": model,
"system": self.system_prompt,
@@ -131,8 +182,15 @@ class ProviderService:
"max_tokens": self.max_tokens,
"temperature": 0.2,
}
resolved_thinking = self._resolve_anthropic_thinking(thinking)
if resolved_thinking:
request_body["thinking"] = resolved_thinking
if self.provider == "minimax" and self.base_url.endswith("/anthropic"):
path = "/v1/messages"
else:
path = "/messages"
return await self._post(
path="/messages",
path=path,
headers={
"x-api-key": self.api_key,
"anthropic-version": self.anthropic_version,
@@ -141,6 +199,25 @@ class ProviderService:
request_body=request_body,
)
def _resolve_anthropic_thinking(self, thinking: dict[str, Any] | None) -> dict[str, Any] | None:
if thinking:
return thinking
# OpenClaw treats MiniMax's Anthropic-compatible path specially:
# disable thinking by default unless the caller explicitly opts in.
if self.provider == "minimax":
return {"type": "disabled"}
return None
async def _request_anthropic_compatible(
self,
model: str,
prompt: str,
thinking: dict[str, Any] | None = None,
) -> dict[str, Any]:
return await self._request_anthropic_messages(model, prompt, thinking)
async def _request_ollama(self, model: str, prompt: str) -> dict[str, Any]:
request_body = {
"model": model,
@@ -218,6 +295,31 @@ class ProviderService:
)
return ""
def _extract_openai_blocks(self, payload: dict[str, Any]) -> list[AIContentBlock]:
choices = payload.get("choices") or []
if not choices:
return []
message = choices[0].get("message") or {}
content = message.get("content")
if isinstance(content, str):
return [AIContentBlock(type="text", text=content)]
if not isinstance(content, list):
return []
blocks: list[AIContentBlock] = []
for item in content:
if not isinstance(item, dict):
continue
blocks.append(
AIContentBlock(
type=str(item.get("type", "text")),
text=item.get("text") if isinstance(item.get("text"), str) else None,
metadata={k: v for k, v in item.items() if k not in {"type", "text"}},
)
)
return blocks
def _extract_anthropic_content(self, payload: dict[str, Any]) -> str:
content = payload.get("content")
if isinstance(content, str):
@@ -233,8 +335,38 @@ class ProviderService:
fragments.append(item["text"])
return "".join(fragments)
def _extract_anthropic_blocks(self, payload: dict[str, Any]) -> list[AIContentBlock]:
content = payload.get("content")
if not isinstance(content, list):
return []
blocks: list[AIContentBlock] = []
for item in content:
if not isinstance(item, dict):
continue
blocks.append(
AIContentBlock(
type=str(item.get("type", "unknown")),
text=item.get("text") if isinstance(item.get("text"), str) else None,
thinking=item.get("thinking") if isinstance(item.get("thinking"), str) else None,
signature=item.get("signature") if isinstance(item.get("signature"), str) else None,
metadata={
k: v
for k, v in item.items()
if k not in {"type", "text", "thinking", "signature"}
},
)
)
return blocks
def _extract_ollama_content(self, payload: dict[str, Any]) -> str:
response = payload.get("response")
if isinstance(response, str):
return response
return ""
def _extract_ollama_blocks(self, payload: dict[str, Any]) -> list[AIContentBlock]:
response = payload.get("response")
if isinstance(response, str) and response:
return [AIContentBlock(type="text", text=response)]
return []

13
aiprovider/schemas.py
View File

@@ -3,6 +3,14 @@ from typing import Any
from pydantic import BaseModel, Field
class AIContentBlock(BaseModel):
type: str
text: str | None = None
thinking: str | None = None
signature: str | None = None
metadata: dict[str, Any] = Field(default_factory=dict)
class SituationalAnalysisRequest(BaseModel):
title: str = Field(..., min_length=1, max_length=200)
objective: str = Field(..., min_length=1, max_length=1000)
@@ -10,17 +18,22 @@ class SituationalAnalysisRequest(BaseModel):
observations: list[str] = Field(default_factory=list)
constraints: list[str] = Field(default_factory=list)
preferred_model: str | None = Field(default=None, max_length=200)
thinking: dict[str, Any] | None = None
class SituationalAnalysisResponse(BaseModel):
provider: str
model: str
content: str
content_blocks: list[AIContentBlock] = Field(default_factory=list)
text_blocks: list[str] = Field(default_factory=list)
thinking_blocks: list[str] = Field(default_factory=list)
raw_response: dict[str, Any] = Field(default_factory=dict)
class AIProviderStatusResponse(BaseModel):
provider: str
api: str | None = None
enabled: bool
configured: bool
model: str | None = None

13
backend/app/schemas/ai.py
View File

@@ -3,6 +3,14 @@ from typing import Any
from pydantic import BaseModel, Field
class AIContentBlock(BaseModel):
type: str
text: str | None = None
thinking: str | None = None
signature: str | None = None
metadata: dict[str, Any] = Field(default_factory=dict)
class SituationalAnalysisRequest(BaseModel):
title: str = Field(..., min_length=1, max_length=200)
objective: str = Field(..., min_length=1, max_length=1000)
@@ -10,17 +18,22 @@ class SituationalAnalysisRequest(BaseModel):
observations: list[str] = Field(default_factory=list)
constraints: list[str] = Field(default_factory=list)
preferred_model: str | None = Field(default=None, max_length=200)
thinking: dict[str, Any] | None = None
class SituationalAnalysisResponse(BaseModel):
provider: str
model: str
content: str
content_blocks: list[AIContentBlock] = Field(default_factory=list)
text_blocks: list[str] = Field(default_factory=list)
thinking_blocks: list[str] = Field(default_factory=list)
raw_response: dict[str, Any] = Field(default_factory=dict)
class AIProviderStatusResponse(BaseModel):
provider: str
api: str | None = None
enabled: bool
configured: bool
model: str | None = None

10
backend/tests/test_api.py
View File

@@ -165,7 +165,8 @@ async def test_ai_provider_status_with_auth(auth_headers):
class _FakeAIProviderClient:
async def get_status(self, request_id=None):
return AIProviderStatusResponse(
provider="openai_compatible",
provider="minimax",
api="anthropic-messages",
enabled=True,
configured=True,
model="test-model",
@@ -193,6 +194,7 @@ async def test_ai_provider_status_with_auth(auth_headers):
assert response.status_code == 200
data = response.json()
assert "provider" in data
assert "api" in data
assert "configured" in data
finally:
app.dependency_overrides.clear()
@@ -207,6 +209,9 @@ async def test_ai_situational_analysis_returns_503_when_disabled(auth_headers):
provider="openai_compatible",
model="test-model",
content="1) 态势摘要: 测试返回",
content_blocks=[],
text_blocks=["1) 态势摘要: 测试返回"],
thinking_blocks=[],
raw_response={"id": "mock-response"},
)
@@ -241,5 +246,8 @@ async def test_ai_situational_analysis_returns_503_when_disabled(auth_headers):
data = response.json()
assert data["provider"] == "openai_compatible"
assert data["content"]
assert "content_blocks" in data
assert "text_blocks" in data
assert "thinking_blocks" in data
finally:
app.dependency_overrides.clear()

67
docs/aiprovider.md
View File

@@ -31,22 +31,40 @@ The recommended default is:
- timeout and lightweight retry
- request tracing via `X-Request-ID`
This now follows an OpenClaw-like seam:
- `AI_PROVIDER` identifies the vendor or logical provider
- `AI_PROVIDER_API` identifies the wire adapter
That split makes MiniMax, Claude-compatible gateways, and self-hosted OpenAI-compatible services easier to model without overloading one config field.
## Supported Providers
`aiprovider` currently supports:
`aiprovider` currently supports these provider identities:
- `openai`
- `openai_compatible`
- `anthropic`
- `minimax`
- `ollama`
Supported request adapters:
- `openai-completions`
- `anthropic-messages`
- `ollama-generate`
Backward-compatible aliases still accepted:
- `openai_compatible`
- `anthropic_compatible`
- `claude_compatible`
- `ollama`
Provider mapping:
- `vLLM`, `LM Studio`, `One API`: `openai_compatible`
- `MiniMax`, Claude-compatible gateways: `claude_compatible`
- `Ollama`: `ollama`
- `vLLM`, `LM Studio`, `One API`: `AI_PROVIDER=openai`, `AI_PROVIDER_API=openai-completions`
- `MiniMax`: `AI_PROVIDER=minimax`, `AI_PROVIDER_API=anthropic-messages`
- Claude-compatible gateways: `AI_PROVIDER=anthropic`, `AI_PROVIDER_API=anthropic-messages`
- `Ollama`: `AI_PROVIDER=ollama`, `AI_PROVIDER_API=ollama-generate`
## API Surfaces
@@ -137,9 +155,13 @@ Both backend and `aiprovider` return the same payload shape:
```json
{
"provider": "openai_compatible",
"model": "gpt-4o-mini",
"provider": "minimax",
"api": "anthropic-messages",
"model": "MiniMax-M2.7",
"content": "1) 态势摘要 ...",
"content_blocks": [],
"text_blocks": [],
"thinking_blocks": [],
"raw_response": {}
}
```
@@ -189,17 +211,37 @@ AI_ANALYSIS_SYSTEM_PROMPT=你是态势感知分析助手。请基于输入的上
### OpenAI-compatible example
```env
AI_PROVIDER=openai_compatible
AI_PROVIDER=openai
AI_PROVIDER_API=openai-completions
AI_BASE_URL=http://127.0.0.1:8001/v1
AI_API_KEY=local-key
AI_MODEL=your-local-model
```
### Claude-compatible example
### MiniMax CN example
```env
AI_PROVIDER=claude_compatible
AI_BASE_URL=https://your-claude-compatible-endpoint.example.com
AI_PROVIDER=minimax
AI_PROVIDER_API=anthropic-messages
AI_BASE_URL=https://api.minimaxi.com/anthropic
AI_API_KEY=sk-cp-xxxxx
AI_MODEL=MiniMax-M2.7
AI_MAX_TOKENS=1200
AI_ANTHROPIC_VERSION=2023-06-01
```
MiniMax note:
- This follows the same Anthropic Messages request shape as the official MiniMax examples.
- For MiniMax, `aiprovider` now disables `thinking` by default unless the caller explicitly passes a `thinking` object.
- This mirrors OpenClaw's caution around MiniMax Anthropic-compatible behavior.
### Anthropic-compatible example
```env
AI_PROVIDER=anthropic
AI_PROVIDER_API=anthropic-messages
AI_BASE_URL=https://your-claude-compatible-endpoint.example.com/anthropic
AI_API_KEY=your_api_key
AI_MODEL=your-model
AI_MAX_TOKENS=1200
@@ -210,6 +252,7 @@ AI_ANTHROPIC_VERSION=2023-06-01
```env
AI_PROVIDER=ollama
AI_PROVIDER_API=ollama-generate
AI_BASE_URL=http://127.0.0.1:11434
AI_API_KEY=
AI_MODEL=qwen2.5:7b