kb-ai

Proof of concept. This package exists to demonstrate the shape of the API and the packaging pipeline. It is not production-hardened — expect breaking changes.

A lightweight, dependency-free knowledge base toolkit for AI applications. Store documents, search them with tf-idf ranking, and persist everything as plain JSON — no vector database or model server required to get started.

Maintained by 4us71n0.

Installation

pip install kb-ai

Quick start

from kb_ai import Document, KnowledgeBase

kb = KnowledgeBase()
kb.add_text("faq-1", "Refunds are issued within 5 business days.", {"topic": "billing"})
kb.add(Document(id="faq-2", text="Password resets are sent by email."))

for hit in kb.search("how long do refunds take"):
    print(hit.score, hit.document.id, hit.document.text)

kb.save("kb.json")
restored = KnowledgeBase.load("kb.json")

Command line

kb-ai add faq-1 "Refunds are issued within 5 business days." --metadata '{"topic": "billing"}'
kb-ai search "refund policy" -n 3
kb-ai list
kb-ai remove faq-1

Every command accepts --store PATH to point at a specific JSON file (default: kb.json in the current directory).

API

Development

python -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"
pytest
ruff check .
mypy

Releasing to PyPI

1. Bump __version__ in src/kb_ai/__init__.py and update CHANGELOG.md.
2. Build and verify locally:

   python -m build
   twine check dist/*
   

3. Tag and push — the publish workflow uploads to PyPI via Trusted Publishing:

   git tag v0.1.0 && git push origin v0.1.0
   

See PUBLISHING.md for the one-time PyPI setup.

License

MIT — see LICENSE.