> ## Documentation Index
> Fetch the complete documentation index at: https://docs.vimaspatial.tech/llms.txt
> Use this file to discover all available pages before exploring further.

# Architecture

> How Vima stays small enough for the production VPS.

Production is a single Vultr VPS:

| Service  | Internal Port | Role                       |
| -------- | ------------- | -------------------------- |
| Caddy    | `80`, `443`   | TLS and reverse proxy      |
| Frontend | `3000`        | landing page and dashboard |
| Backend  | `8765`        | FastAPI evidence API       |
| MCP      | `8766`        | remote agent tool endpoint |

## Routing

```text theme={"dark"}
https://vimaspatial.tech/       -> frontend:3000
https://vimaspatial.tech/api/*  -> backend:8765
https://vimaspatial.tech/mcp*   -> mcp:8766
```

## No-GPU Constraint

The production box is 1 vCPU / 2 GB RAM. Anything that runs there must be tiny.

That is why expensive perception is precomputed offline:

```text theme={"dark"}
Depth Anything + SAM + reconstruction artifacts
  -> static inference outputs
  -> backend evidence context
  -> dashboard / API / MCP responses
```

<Warning>
  Do not add Torch, Transformers, Open3D, SAM, Depth Anything, or local COLMAP
  runtime dependencies to production containers.
</Warning>

## Deploy

Pushes to `main` run `.github/workflows/deploy.yml`, rsync to the VPS, rebuild
Docker Compose, and verify backend, MCP, and public Caddy health.