Skip to content

Latest commit

 

History

History

README.md

Examples

Runnable examples for LodeDB. They assume you've installed from source (see the repo README) and run them inside the project environment with uv run. The first run downloads the embedding model from Hugging Face and caches it locally.

File What it shows Run
quickstart.py Core SDK: add, search, batched search_many, get, persist uv run python examples/quickstart.py
multimodal_clip.py Image + text search via the clip preset, and bring-your-own vectors (needs --extra image for the CLIP half) uv run python examples/multimodal_clip.py
langchain_store.py LodeDB as a LangChain VectorStore (needs --extra langchain) uv run python examples/langchain_store.py
llama_index_store.py LodeDB as a LlamaIndex VectorStore (needs --extra llama-index) uv run python examples/llama_index_store.py
llama_index_graph_store.py LodeDB as a LlamaIndex PropertyGraphStore (needs --extra llama-index) uv run python examples/llama_index_graph_store.py
mem0_store.py LodeDB as a mem0 VectorStoreBase backend (needs --extra mem0) uv run python examples/mem0_store.py
cognee_provider.py Register LodeDB as a cognee vector_db_provider (needs --extra cognee) uv run python examples/cognee_provider.py
privategpt_provider.py Register LodeDB as PrivateGPT's vector store (needs --extra llama-index) uv run python examples/privategpt_provider.py
kotaemon_store.py LodeDB as a kotaemon vector store (no extra needed) uv run python examples/kotaemon_store.py
mcp_config.json Register LodeDB as an MCP-capable agent's local memory (needs --extra mcp --extra embeddings) drop into your agent's MCP config

Each script writes its index under a local ./data* folder, which is git-ignored.

MCP config note

pip install 'lodedb[mcp,embeddings]' puts the lodedb CLI on your PATH (the server embeds text, so it needs the embedding extra), so the normal MCP entry is just:

{ "mcpServers": { "lodedb": { "command": "lodedb", "args": ["mcp", "--path", "./data"] } } }

If you installed into a virtual environment (including a uv project) where lodedb isn't on PATH, use the uv run --project form in mcp_config.json instead and edit the two absolute paths. The main README also covers lodedb mcp install, which writes the entry for you.