- Add app/llm/router.py shim — tri-level config lookup: repo config/llm.yaml → ~/.config/circuitforge/llm.yaml → env vars - Add config/llm.cloud.yaml — ollama via cf-orch, llama3.1:8b - Add config/llm.yaml.example — self-hosted reference config - compose.cloud.yml: mount llm.cloud.yaml, set CF_ORCH_URL, add host.docker.internal:host-gateway (required on Linux Docker) - api/main.py: use app.llm.router.LLMRouter (shim) not core directly - .env.example: update LLM section to reference config/llm.yaml.example - .gitignore: exclude config/llm.yaml (keep example + cloud yaml) End-to-end tested: 3.2s for "used RTX 3080 under $400, no mining cards" via cloud container → host.docker.internal:11434 → Ollama llama3.1:8b
57 lines
2.2 KiB
YAML
57 lines
2.2 KiB
YAML
# Snipe — cloud managed instance
|
|
# Project: snipe-cloud (docker compose -f compose.cloud.yml -p snipe-cloud ...)
|
|
# Web: http://127.0.0.1:8514 → menagerie.circuitforge.tech/snipe (via Caddy basicauth)
|
|
# API: internal only on snipe-cloud-net (no host port — only reachable via nginx)
|
|
#
|
|
# Usage: ./manage.sh cloud-start | cloud-stop | cloud-restart | cloud-status | cloud-logs | cloud-build
|
|
|
|
services:
|
|
api:
|
|
build:
|
|
context: ..
|
|
dockerfile: snipe/Dockerfile
|
|
restart: unless-stopped
|
|
env_file: .env
|
|
environment:
|
|
# Cloud mode — enables per-user DB isolation and Heimdall tier resolution.
|
|
# All values may be overridden by setting them in .env (env_file takes precedence
|
|
# over environment: only when the same key appears in both — Docker merges them,
|
|
# env_file wins for duplicates).
|
|
CLOUD_MODE: "true"
|
|
CLOUD_DATA_ROOT: /devl/snipe-cloud-data
|
|
# DIRECTUS_JWT_SECRET, HEIMDALL_URL, HEIMDALL_ADMIN_TOKEN — set in .env (never commit)
|
|
# CF_ORCH_URL routes LLM query builder through cf-orch for VRAM-aware scheduling.
|
|
# Override in .env to use a different coordinator URL.
|
|
CF_ORCH_URL: "http://host.docker.internal:7700"
|
|
extra_hosts:
|
|
- "host.docker.internal:host-gateway"
|
|
# No network_mode: host — isolated on snipe-cloud-net; nginx reaches it via 'api:8510'
|
|
volumes:
|
|
- /devl/snipe-cloud-data:/devl/snipe-cloud-data
|
|
- ./config/llm.cloud.yaml:/app/snipe/config/llm.yaml:ro
|
|
networks:
|
|
- snipe-cloud-net
|
|
|
|
web:
|
|
build:
|
|
context: .
|
|
dockerfile: docker/web/Dockerfile
|
|
args:
|
|
# Vite bakes these at image build time — changing them requires cloud-build.
|
|
# VITE_BASE_URL: app served under /snipe → asset URLs become /snipe/assets/...
|
|
# VITE_API_BASE: prepended to all /api/* fetch calls → /snipe/api/search
|
|
VITE_BASE_URL: /snipe
|
|
VITE_API_BASE: /snipe
|
|
restart: unless-stopped
|
|
ports:
|
|
- "8514:80" # Caddy (caddy-proxy container) reaches via host.docker.internal:8514
|
|
volumes:
|
|
- ./docker/web/nginx.cloud.conf:/etc/nginx/conf.d/default.conf:ro
|
|
networks:
|
|
- snipe-cloud-net
|
|
depends_on:
|
|
- api
|
|
|
|
networks:
|
|
snipe-cloud-net:
|
|
driver: bridge
|