Install

Get Sombra running locally in three commands.

Prerequisites

• uv — recommended Python installer
• Rootless Podman with podman-compose
• Node.js 18+ and npm — for the docs site (optional)

1. Clone and install

git clone <your-repo-url> sombra
cd sombra
uv sync

This installs the Sombra recognizer library plus the test tooling.

2. Create your environment file

cp .env.example .env

Then open .env and set your real OPENCODE_ZEN_API_KEY — the key for the remote LLM provider (OpenCode Zen). The default LITELLM_MASTER_KEY is fine for local use.

3. Start the stack

make up

make up loads .env and runs podman-compose up -d --build, starting three services:

Service	Port	Role
presidio-analyzer	:5002	Detects PII in prompts
presidio-anonymizer	:5001	Masks the detected tokens
litellm	:4000	OpenAI-compatible proxy with the PII guardrail

Just want to build the analyzer image without starting the stack? Use make build.

Next

Your stack is running. Now configure OpenCode.