tracer-cloud--opensre/docs/deployment.mdx

---
title: "Deployment"
slug: "deployment"
description: "Deploy OpenSRE as a standard Python/FastAPI runtime"
---

OpenSRE deploys as a standard Python/FastAPI runtime. Use the repo `Dockerfile`,
Railway, EC2, ECS, Vercel, or another ASGI-capable host.

## Runtime environment

<Steps>
  <Step title="Deploy the app">
    Deploy this repository using your hosting provider's normal app workflow.
  </Step>

<Step title="Configure your model provider">
    Add `LLM_PROVIDER` as an environment variable (for example `anthropic`,
    `openai`, `openrouter`, or `gemini`).
</Step>

  <Step title="Add the matching provider API key">
    Use the API key that matches your provider:

    - `ANTHROPIC_API_KEY` for `LLM_PROVIDER=anthropic`
    - `OPENAI_API_KEY` for `LLM_PROVIDER=openai`
    - `OPENROUTER_API_KEY` for `LLM_PROVIDER=openrouter`
    - `DEEPSEEK_API_KEY` for `LLM_PROVIDER=deepseek`
    - `GEMINI_API_KEY` for `LLM_PROVIDER=gemini`

  </Step>

  <Step title="Add integration env vars and deploy">
    Add any storage and integration environment variables required by your runtime, then deploy and verify health.
  </Step>
</Steps>

Minimum LLM environment example:

```bash
LLM_PROVIDER=anthropic
ANTHROPIC_API_KEY=...
```

The complete list of provider keys and optional model overrides is documented in `.env.example`.

## Prebuilt Docker image

Public images are published to GitHub Container Registry for every release
(`linux/amd64` and `linux/arm64`), so you can skip building from source.

| Tag | What it is |
| --- | --- |
| `ghcr.io/tracer-cloud/opensre:latest` | Most recent release |
| `ghcr.io/tracer-cloud/opensre:<version>` | A specific release, e.g. `0.1.2026.7.8` (matches the GitHub release tag without the `v`) |

```bash
docker pull ghcr.io/tracer-cloud/opensre:latest

# Web mode (FastAPI health app, default)
docker run -p 8000:8000 --env-file .env ghcr.io/tracer-cloud/opensre:latest

# Gateway mode (Telegram two-way messaging gateway)
docker run -e MODE=gateway --env-file .env ghcr.io/tracer-cloud/opensre:latest
```

## Run local gateway

Use the built-in gateway command when you want a local HTTP server for investigations.

| Command | What it does |
| --- | --- |
| `uv run opensre gateway start` | Starts the gateway daemon in the background (web app, Telegram chat, task scheduler). |
| `uv run opensre gateway start --foreground` | Runs the gateway attached to this terminal. |
| `uv run opensre gateway stop` | Stops the background daemon. |
| `uv run opensre gateway status` | Shows daemon and component status. |

The web app binds `0.0.0.0` on the port from the `PORT` environment variable
(default `8000`).

Quick local checks:

```bash
curl http://127.0.0.1:8000/ok
```

Example investigation request:

```bash
curl -X POST http://127.0.0.1:8000/investigate \
  -H "Content-Type: application/json" \
  -d '{
    "raw_alert": {
      "alert_name": "etl-daily-orders-failure",
      "pipeline_name": "etl_daily_orders",
      "severity": "critical",
      "message": "Orders pipeline failed with timeout."
    }
  }'
```

By default, `/investigate` (like `/alerts`) accepts only loopback callers. Set
`OPENSRE_ALERT_LISTENER_TOKEN` to allow remote callers with a bearer token:

```bash
curl -X POST http://127.0.0.1:8000/investigate \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $OPENSRE_ALERT_LISTENER_TOKEN" \
  -d '{"raw_alert":{"alert_name":"test"}}'
```

## Railway deployment

Railway is still supported as a hosted runtime option.

Before deploying on Railway, make sure your service has:

- `DATABASE_URI` pointing to your Railway Postgres instance
- `REDIS_URI` pointing to your Railway Redis instance

Then deploy the service using your Railway project workflow.