pynchy-ops

# Pynchy Ops The live Pynchy host and checkout path are deployment-specific. Public repo instructions must not assume a private hostname or home-directory layout. Set `PYNCHY_HOST` and `PYNCHY_REMOTE_ROOT` from local memory, environment, or the operator before running remote commands. ## Auto-deploy: Never Restart Manually Pynchy self-manages. Two mechanisms trigger automatic restarts: 1. **Git changes on `main`** — the polling mechanism detects new commits, pulls, and restarts (with container rebuild if source files changed). 2. **Config file changes** — editing `config.toml`, `litellm_config.yaml`, or other settings files triggers an automatic deploy on the next host git-sync poll. The default interval is 300 seconds; check `[scheduler].git_sync_interval_seconds` before deciding it was missed. **Do not manually restart containers or the service.** This includes `docker restart`, `systemctl restart`, and direct container management (`docker kill/stop/rm`). Manual restarts bypass lifecycle management and can leave things in a bad state. Only use manual commands when the service is unhealthy and needs fixing. See [references/server-debug.md](references/server-debug.md) for diagnostic steps. ## Quick Status Check **Preferred: the `/status` endpoint.** Single command that returns everything: ```bash # On the live host directly: curl -s http://localhost:8484/status | python3 -m json.tool # Remotely (via Tailscale): PYNCHY_HOST="${PYNCHY_HOST:?set the live host}" PYNCHY_...