Fix: /v1/models auf geteilten Client (Regression aus 12c387a)
Beim Entfernen des httpx-Imports uebersehen: der /models-Endpoint nutzte httpx noch direkt -> NameError/500. Jetzt zieht auch /models den geteilten app.state.gw_client (request-Param ergaenzt). Live verifiziert. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
@@ -41,10 +41,10 @@ _LANE_LABELS = {"coding": "Coding (Router → coder/heavy/fast)", "chat": "Chat
|
|||||||
|
|
||||||
|
|
||||||
@router.get("/models")
|
@router.get("/models")
|
||||||
async def models():
|
async def models(request: Request):
|
||||||
async with httpx.AsyncClient(timeout=10) as c:
|
client = request.app.state.gw_client # geteilter Keep-Alive-Client (siehe app.py lifespan)
|
||||||
r = await c.get(f"{LLAMA_SWAP_URL}/v1/models")
|
r = await client.get(f"{LLAMA_SWAP_URL}/v1/models", timeout=10.0)
|
||||||
data = r.json()
|
data = r.json()
|
||||||
# Lanes ganz oben einblenden, damit IDEs einfach „coding"/„chat" wählen können.
|
# Lanes ganz oben einblenden, damit IDEs einfach „coding"/„chat" wählen können.
|
||||||
lanes = [{"id": lane, "object": "model", "owned_by": "mc2-router",
|
lanes = [{"id": lane, "object": "model", "owned_by": "mc2-router",
|
||||||
"description": _LANE_LABELS.get(lane, lane)} for lane in LANES]
|
"description": _LANE_LABELS.get(lane, lane)} for lane in LANES]
|
||||||
|
|||||||
Reference in New Issue
Block a user