c4b94a327d
Node Worker (services/node-worker/):
- NATS subscriber for node.{NODE_ID}.llm.request / vision.request
- Canonical JobRequest/JobResponse envelope (Pydantic)
- Idempotency cache (TTL 10min) with inflight dedup
- Deadline enforcement (DEADLINE_EXCEEDED on expired jobs)
- Concurrency limiter (semaphore, returns busy)
- Ollama + Swapper vision providers
Router offload (services/router/offload_client.py):
- NATS req/reply with configurable retries
- Circuit breaker per node+type (3 fails/60s → open 120s)
- Concurrency semaphore for remote requests
Model selection (services/router/model_select.py):
- exclude_nodes parameter for circuit-broken nodes
- force_local flag for fallback re-selection
- Integrated circuit breaker state awareness
Router /infer pipeline:
- Remote offload path when NCS selects remote node
- Automatic fallback: exclude failed node → force_local re-select
- Deadline propagation from router to node-worker
Tests: 17 unit tests (idempotency, deadline, circuit breaker)
Docs: ops/offload_routing.md (subjects, envelope, verification)
Made-with: Cursor
82 lines
2.2 KiB
Python
82 lines
2.2 KiB
Python
"""Ollama LLM provider for node-worker."""
|
|
import logging
|
|
from typing import Any, Dict, List, Optional
|
|
|
|
import httpx
|
|
|
|
from config import OLLAMA_BASE_URL, DEFAULT_LLM
|
|
|
|
logger = logging.getLogger("provider.ollama")
|
|
|
|
|
|
async def infer(
|
|
messages: Optional[List[Dict[str, str]]] = None,
|
|
prompt: str = "",
|
|
model: str = "",
|
|
system: str = "",
|
|
max_tokens: int = 2048,
|
|
temperature: float = 0.2,
|
|
timeout_s: float = 25.0,
|
|
) -> Dict[str, Any]:
|
|
model = model or DEFAULT_LLM
|
|
|
|
if messages:
|
|
return await _chat(messages, model, max_tokens, temperature, timeout_s)
|
|
return await _generate(prompt, system, model, max_tokens, temperature, timeout_s)
|
|
|
|
|
|
async def _chat(
|
|
messages: List[Dict[str, str]],
|
|
model: str,
|
|
max_tokens: int,
|
|
temperature: float,
|
|
timeout_s: float,
|
|
) -> Dict[str, Any]:
|
|
async with httpx.AsyncClient(timeout=timeout_s) as c:
|
|
resp = await c.post(
|
|
f"{OLLAMA_BASE_URL}/api/chat",
|
|
json={
|
|
"model": model,
|
|
"messages": messages,
|
|
"stream": False,
|
|
"options": {"num_predict": max_tokens, "temperature": temperature},
|
|
},
|
|
)
|
|
resp.raise_for_status()
|
|
data = resp.json()
|
|
return {
|
|
"text": data.get("message", {}).get("content", ""),
|
|
"model": model,
|
|
"provider": "ollama",
|
|
"eval_count": data.get("eval_count", 0),
|
|
}
|
|
|
|
|
|
async def _generate(
|
|
prompt: str,
|
|
system: str,
|
|
model: str,
|
|
max_tokens: int,
|
|
temperature: float,
|
|
timeout_s: float,
|
|
) -> Dict[str, Any]:
|
|
async with httpx.AsyncClient(timeout=timeout_s) as c:
|
|
resp = await c.post(
|
|
f"{OLLAMA_BASE_URL}/api/generate",
|
|
json={
|
|
"model": model,
|
|
"prompt": prompt,
|
|
"system": system,
|
|
"stream": False,
|
|
"options": {"num_predict": max_tokens, "temperature": temperature},
|
|
},
|
|
)
|
|
resp.raise_for_status()
|
|
data = resp.json()
|
|
return {
|
|
"text": data.get("response", ""),
|
|
"model": model,
|
|
"provider": "ollama",
|
|
"eval_count": data.get("eval_count", 0),
|
|
}
|