feat(v2): T7 网关扩展 + T9 token 计量与账单（/chat v2,/chat/legacy,/runs,/review,/metrics）
This commit is contained in:
tzt
+261
-94
@@ -1,94 +1,261 @@
|
||||
"""FastAPI 网关：对外提供 /chat /health /metrics 接口。
|
||||
|
||||
启动：
|
||||
uvicorn gateway.api:app --host 0.0.0.0 --port 8000
|
||||
或：
|
||||
python -m gateway.api
|
||||
|
||||
依赖：fastapi, uvicorn, pydantic（见 requirements.txt）
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
from typing import List, Optional
|
||||
|
||||
from pydantic import BaseModel, Field
|
||||
|
||||
from router_system.config import load_config
|
||||
from router_system.router import Router, build_router
|
||||
|
||||
# ---- 全局单例 ----
|
||||
_router: Optional[Router] = None
|
||||
|
||||
|
||||
def get_router() -> Router:
|
||||
global _router
|
||||
if _router is None:
|
||||
_router = build_router()
|
||||
return _router
|
||||
|
||||
|
||||
# ---- 请求/响应模型 ----
|
||||
class QueryRequest(BaseModel):
|
||||
query: str = Field(..., min_length=1, max_length=8000, description="用户查询")
|
||||
|
||||
|
||||
class QueryResponse(BaseModel):
|
||||
response: str
|
||||
domain: str
|
||||
difficulty: str
|
||||
confidence: float
|
||||
upgraded: bool
|
||||
quality_score: float
|
||||
model_used: str
|
||||
route: List[str]
|
||||
latency_ms: float
|
||||
cache_hit: bool
|
||||
cache_level: Optional[str]
|
||||
cost_est: float
|
||||
error: Optional[str] = None
|
||||
|
||||
|
||||
class HealthResponse(BaseModel):
|
||||
status: str
|
||||
domains: List[str]
|
||||
classifier: str
|
||||
judge: str
|
||||
fallback: str
|
||||
|
||||
|
||||
# ---- FastAPI 应用 ----
|
||||
try:
|
||||
from fastapi import FastAPI
|
||||
|
||||
app = FastAPI(
|
||||
title="Multi-Expert Router API",
|
||||
description="多专业小模型 + 路由模型系统（MVP）",
|
||||
version="0.1.0",
|
||||
)
|
||||
|
||||
@app.get("/health", response_model=HealthResponse, tags=["system"])
|
||||
async def health():
|
||||
return get_router().health()
|
||||
|
||||
@app.post("/chat", response_model=QueryResponse, tags=["chat"])
|
||||
async def chat(req: QueryRequest):
|
||||
result = await get_router().route(req.query)
|
||||
return QueryResponse(**result.to_dict())
|
||||
|
||||
@app.get("/metrics", tags=["system"])
|
||||
async def metrics():
|
||||
r = get_router()
|
||||
return {
|
||||
"router": r.stats.summary(),
|
||||
"cache": r.cache.stats(),
|
||||
}
|
||||
|
||||
except ImportError:
|
||||
# fastapi 未安装时，提供 CLI 入口提示
|
||||
app = None
|
||||
print("[gateway] 未安装 fastapi，请执行: pip install -r requirements.txt")
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
import uvicorn
|
||||
uvicorn.run("gateway.api:app", host="0.0.0.0", port=8000, reload=False)␍
|
||||
"""FastAPI 网关：对外提供 v2（端云协同）与 v1（legacy 路由）两套接口。
|
||||
|
||||
v2 端点（《实现方案_v2》5.3）：
|
||||
POST /chat 端云协同协作管线
|
||||
POST /chat/legacy 原 v1 L0 行为（离线降级，测试封闭）
|
||||
GET /runs/{request_id}/workspace 交流文本
|
||||
GET /runs/{request_id}/artifacts/{name} 工件下载
|
||||
GET /review/queue、POST /review/{id} 人工检验
|
||||
GET /metrics 含 v2 token/快路径统计
|
||||
|
||||
启动：uvicorn gateway.api:app --host 0.0.0.0 --port 8000
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
from pathlib import Path
|
||||
from typing import List, Optional
|
||||
|
||||
from pydantic import BaseModel, Field
|
||||
|
||||
from router_system.config import load_config
|
||||
from router_system.router import Router, build_router
|
||||
|
||||
# ---- v2 依赖（惰性导入，缺依赖时降级提示） ----
|
||||
try:
|
||||
from router_system.architect import build_architect
|
||||
from router_system.knowledge import KnowledgeBase
|
||||
from router_system.pipeline import CollaborativePipeline, build_pipeline
|
||||
from router_system.review import ReviewQueue
|
||||
from router_system.v2stats import V2Stats
|
||||
from router_system.worker import build_worker
|
||||
_V2_OK = True
|
||||
except Exception: # pragma: no cover
|
||||
_V2_OK = False
|
||||
|
||||
# ---- v1 全局单例 ----
|
||||
_router: Optional[Router] = None
|
||||
|
||||
# ---- v2 全局单例 ----
|
||||
_pipeline: Optional["CollaborativePipeline"] = None
|
||||
_v2stats = V2Stats() if _V2_OK else None
|
||||
_review = None
|
||||
|
||||
|
||||
def get_router() -> Router:
|
||||
global _router
|
||||
if _router is None:
|
||||
_router = build_router()
|
||||
return _router
|
||||
|
||||
|
||||
def get_review() -> "ReviewQueue":
|
||||
global _review
|
||||
if _review is None and _V2_OK:
|
||||
cfg = load_config().get("review", {})
|
||||
_review = ReviewQueue(db_path=cfg.get("queue_db", "data/review.sqlite3"))
|
||||
return _review
|
||||
|
||||
|
||||
def build_v2_pipeline(worker_cfg_override: Optional[dict] = None):
|
||||
"""从配置构建 v2 协作管线（architect + worker + pipeline）。
|
||||
|
||||
worker_cfg_override 可注入（测试/演示用 mock）。无 API key 时 /chat 会走
|
||||
本地降级路径（不崩溃）。
|
||||
"""
|
||||
global _pipeline
|
||||
if _pipeline is None:
|
||||
cfg = load_config()
|
||||
kb = KnowledgeBase()
|
||||
architect = build_architect(cfg.get("architect", {}))
|
||||
wcfg = dict(cfg.get("worker", {}))
|
||||
if worker_cfg_override:
|
||||
wcfg.update(worker_cfg_override)
|
||||
worker = build_worker(wcfg, kb=kb)
|
||||
_pipeline = build_pipeline(cfg, architect, worker)
|
||||
return _pipeline
|
||||
|
||||
|
||||
def set_pipeline(pipe) -> None:
|
||||
"""测试注入替身管线。"""
|
||||
global _pipeline
|
||||
_pipeline = pipe
|
||||
|
||||
|
||||
def get_pipeline():
|
||||
global _pipeline
|
||||
if _pipeline is None:
|
||||
return build_v2_pipeline()
|
||||
return _pipeline
|
||||
|
||||
|
||||
# ---- 请求/响应模型 ----
|
||||
class QueryRequest(BaseModel):
|
||||
query: str = Field(..., min_length=1, max_length=8000, description="用户查询")
|
||||
domain_group: Optional[str] = Field(
|
||||
None, description="大领域组（两级路由第一级）：tech | professional | lifestyle | general；不指定则自动检测"
|
||||
)
|
||||
|
||||
|
||||
class HealthResponse(BaseModel):
|
||||
status: str
|
||||
domains: List[str]
|
||||
classifier: str
|
||||
judge: str
|
||||
fallback: str
|
||||
|
||||
|
||||
# ---- FastAPI 应用 ----
|
||||
try:
|
||||
from fastapi import FastAPI, HTTPException
|
||||
|
||||
app = FastAPI(
|
||||
title="端云协同 LLM 协作系统",
|
||||
description="大模型(Architect) + 本地小模型(Worker) 通过交流文本协作；v1 保留为 legacy 路由",
|
||||
version="2.0.0",
|
||||
)
|
||||
|
||||
@app.get("/health", response_model=HealthResponse, tags=["system"])
|
||||
async def health():
|
||||
return get_router().health()
|
||||
|
||||
# ---------------- v2：/chat ----------------
|
||||
@app.post("/chat", tags=["chat"])
|
||||
async def chat(req: QueryRequest):
|
||||
try:
|
||||
result = await get_pipeline().run(req.query)
|
||||
except ValueError as e:
|
||||
raise HTTPException(status_code=400, detail=str(e))
|
||||
# 记账 + 入队
|
||||
if _v2stats is not None:
|
||||
_v2stats.record(result)
|
||||
_maybe_enqueue(result)
|
||||
return {
|
||||
"response": result.response,
|
||||
"request_id": result.request_id,
|
||||
"status": result.status,
|
||||
"fast_path": result.fast_path,
|
||||
"rounds_used": result.rounds_used,
|
||||
"api_input_tokens": result.api_input_tokens,
|
||||
"api_output_tokens": result.api_output_tokens,
|
||||
"cost_est": result.cost_est,
|
||||
"model_used": result.model_used,
|
||||
"latency_ms": round(result.latency_ms, 2),
|
||||
"route": result.route,
|
||||
"workspace_path": result.workspace_path,
|
||||
"error": result.error,
|
||||
}
|
||||
|
||||
# ---------------- v1：/chat/legacy ----------------
|
||||
@app.post("/chat/legacy", tags=["chat"])
|
||||
async def chat_legacy(req: QueryRequest):
|
||||
"""原 v1 L0 专家系统行为（离线可跑，测试封闭）。"""
|
||||
try:
|
||||
result = await get_router().route(req.query, domain_group=req.domain_group)
|
||||
except ValueError as e:
|
||||
raise HTTPException(status_code=400, detail=str(e))
|
||||
return {
|
||||
"response": result.response,
|
||||
"domain": result.domain,
|
||||
"difficulty": result.difficulty,
|
||||
"confidence": round(result.confidence, 4),
|
||||
"upgraded": result.upgraded,
|
||||
"quality_score": round(result.quality_score, 4),
|
||||
"model_used": result.model_used,
|
||||
"route": result.route,
|
||||
"latency_ms": round(result.latency_ms, 2),
|
||||
"cache_hit": result.cache_hit,
|
||||
"cache_level": result.cache_level,
|
||||
"cost_est": round(result.cost_est, 6),
|
||||
"error": result.error,
|
||||
"subdomain": result.subdomain,
|
||||
"subdomain2": result.subdomain2,
|
||||
"domain_group": result.domain_group,
|
||||
"request_id": result.request_id,
|
||||
}
|
||||
|
||||
@app.get("/traces/{request_id}", tags=["system"])
|
||||
async def get_trace(request_id: str):
|
||||
trace = get_router().trace_store.get(request_id)
|
||||
if trace is None:
|
||||
raise HTTPException(status_code=404, detail=f"未找到请求 {request_id} 的推理链")
|
||||
return trace
|
||||
|
||||
# ---------------- v2：workspace / artifacts ----------------
|
||||
@app.get("/runs/{request_id}/workspace", tags=["v2"])
|
||||
async def get_workspace(request_id: str):
|
||||
p = Path("runs") / request_id / "workspace.json"
|
||||
if not p.exists():
|
||||
raise HTTPException(status_code=404, detail=f"未找到运行 {request_id}")
|
||||
import json
|
||||
return json.loads(p.read_text(encoding="utf-8"))
|
||||
|
||||
@app.get("/runs/{request_id}/artifacts/{name}", tags=["v2"])
|
||||
async def get_artifact(request_id: str, name: str):
|
||||
p = Path("runs") / request_id / "artifacts" / name
|
||||
if not p.exists():
|
||||
raise HTTPException(status_code=404, detail=f"未找到工件 {name}")
|
||||
return {"name": name, "content": p.read_text(encoding="utf-8")}
|
||||
|
||||
# ---------------- v2：人工检验 ----------------
|
||||
@app.get("/review/queue", tags=["v2"])
|
||||
async def review_queue(status: Optional[str] = None, limit: int = 50):
|
||||
return get_review().list(status=status, limit=limit)
|
||||
|
||||
@app.post("/review/{review_id}", tags=["v2"])
|
||||
async def review_submit(review_id: int, verdict: str, correction: Optional[str] = None):
|
||||
try:
|
||||
ok = get_review().submit(review_id, verdict, correction=correction)
|
||||
except ValueError as e:
|
||||
raise HTTPException(status_code=400, detail=str(e))
|
||||
if not ok:
|
||||
raise HTTPException(status_code=404, detail=f"审核记录不存在或已审核: {review_id}")
|
||||
return {"ok": True, "review_id": review_id, "verdict": verdict}
|
||||
|
||||
# ---------------- metrics ----------------
|
||||
@app.get("/metrics", tags=["system"])
|
||||
async def metrics():
|
||||
out = {
|
||||
"router": get_router().stats.summary(),
|
||||
"cache": get_router().cache.stats(),
|
||||
}
|
||||
if _V2_OK and _v2stats is not None:
|
||||
out["v2"] = _v2stats.summary()
|
||||
out["review"] = {
|
||||
"pending": get_review().count(status="pending"),
|
||||
"total": get_review().count(),
|
||||
}
|
||||
return out
|
||||
|
||||
except ImportError:
|
||||
app = None
|
||||
print("[gateway] 未安装 fastapi，请执行: pip install -r requirements.txt")
|
||||
|
||||
|
||||
def _maybe_enqueue(result) -> None:
|
||||
"""按 review 配置抽样/强制入队（异步场景用同步快速调用）。"""
|
||||
if not _V2_OK or get_review() is None:
|
||||
return
|
||||
if result.status not in ("done", "escalated", "fast_path"):
|
||||
return
|
||||
cfg = load_config().get("review", {})
|
||||
sample_rate = float(cfg.get("sample_rate", 0.10))
|
||||
force_tags = cfg.get("force_tags", ["safety"])
|
||||
# 从交流文本读取 tags（若可）
|
||||
tags: List[str] = []
|
||||
if result.workspace_path:
|
||||
try:
|
||||
import json
|
||||
ws = json.loads(Path(result.workspace_path).read_text(encoding="utf-8"))
|
||||
tags = (ws.get("brief") or {}).get("tags") or []
|
||||
except Exception:
|
||||
tags = []
|
||||
if ReviewQueue.should_enqueue(tags, sample_rate=sample_rate, force_tags=force_tags):
|
||||
get_review().enqueue(
|
||||
request_id=result.request_id, query=result.query, response=result.response,
|
||||
tags=tags, reason="auto", workspace_path=result.workspace_path,
|
||||
)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
import uvicorn
|
||||
uvicorn.run("gateway.api:app", host="0.0.0.0", port=8000, reload=False)
|
||||
|
||||
Reference in New Issue
Block a user