DocsStart
Installation
Install BBM-Atlas from PyPI on Linux, macOS or Windows: one package for both editions, with optional extras for local models and tracing.
Requirements
- Python 3.11 or newer, on Linux, macOS or Windows.
- Nothing else to start with: with no configuration, BBM-Atlas keeps repositories, the knowledge graph and learned facts in one local database file.
- To share one server between many people or replicas, Postgres, Neo4j, Qdrant and Redis (Production).
Install from PyPI
BBM-Atlas is one package, bbm-atlas. The Enterprise features are in it too, switched off until a licence key turns them on. Install it into a virtual environment:
terminal
python -m venv .venvsource .venv/bin/activate # Windows: .venv\Scripts\activatepip install bbm-atlasbbm-atlas --versionThe install adds the bbm-atlas command. bbm-atlas --help lists its commands, and bbm-atlas <command> --help describes one.
Optional extras
The core install is light. Heavier capabilities are extras, each switched on by a setting once installed:
| Extra | Adds | Switch it on with |
|---|---|---|
embeddings | A local embedding model (BGE-M3 through sentence-transformers) for semantic search | BBM_ATLAS_EMBEDDING_PROVIDER=sentence_transformers |
llm | A GGUF language model run in-process with llama.cpp, for the agents | BBM_ATLAS_LLM_PROVIDER=llamacpp |
otel | OpenTelemetry traces of every request | BBM_ATLAS_OTEL_ENABLED=true |
detect | An ML content detector that tells code from prose when compressing | BBM_ATLAS_COMPRESSION_DETECTOR=magika |
terminal
pip install "bbm-atlas[embeddings,llm]"Upgrading
terminal
pip install --upgrade bbm-atlasStores are brought up to date the first time a new version uses them. With your own databases, you can apply the changes before starting the new version, for example ahead of a rolling update:
terminal
bbm-atlas migrateNext
- Index and search your first repository.
- Connect your AI assistant over MCP.
- Configure BBM-Atlas with environment variables.