In simple English
This tool lets you build a private question-answering assistant over your own files without writing code. You pick a folder of documents and a database, and the system sets up a chat page and a search tool. When you edit, add, or delete a file, the system updates its memory within seconds. You can run it entirely on your own computer for free, or connect it to external AI services.
What it is
A no-code server for retrieval-augmented generation (a way to feed your files to a language model so it answers using your facts) that connects local files to eight different vector databases. It uses a background file watcher to keep the database in sync with your source folder without manual re-indexing.
What it is good for
Solo builders who want an up-to-date knowledge base over their project documents, customer support files, or local notes without building custom data-sync pipelines.
- Keeps search data updated in real time as files change on disk.
- Works with eight different vector databases including local DuckDB.
- Sets up easily with a single configuration file and a built-in wizard.
- Requires a free licence key from Pathway with a web sign-in.
- Installing local embeddings pulls in a large PyTorch dependency.
- Running local OCR and document parsers takes significant memory on large folders.
Writing custom file watchers and sync scripts for vector databases
DuckDB
- onyx-dot-app/onyx33k stars
An open-source enterprise search and chat system that connects to many document sources.
- ggml-org/llama.cpp
Runs language models locally from a C and C++ library and command-line tool.
- EleutherAI/lm-evaluation-harness
Runs standard question sets against a language model and reports scores comparable with published figures.
- mudler/LocalAI50k stars
Serves text, vision, speech, image and video models from one self-hosted service that imitates the paid interfaces.
pip install serviette[local] export PATHWAY_LICENSE_KEY=your_free_key serviette demo