Skip to main content

(-o-) alit

Your agent forgets every paper it's ever read.

alit is a local knowledge base for AI agents doing literature review. One agent reads 50 papers and stores structured summaries, citations, and reading status. The next agent — or the same agent next week — queries that knowledge instantly.

The agent builds the knowledge. You set the taste. Knowledge compounds.

pip install agent-lit → zero dependencies, SQLite-only, works with any coding agent.

The core loop

alit taste "vision-language grounding, embodied AI"   # you set direction
alit recommend 5                                       # agent picks what to read
alit summarize <id> --l4 "..." --model claude          # agent stores findings
alit ask "what approaches exist for X?" --depth 2      # agent synthesizes

Run alit --help for the full command list (25 commands for search, import, export, citations, and more).

What alit does and doesn't do

alit The agent
Stores papers, summaries, citations, taste ✓
Ranks recommendations (PageRank + taste + recency) ✓
Retrieves context for synthesis ✓
Persists across sessions ✓
Reads papers ✓
Writes summaries ✓
Decides what to cite ✓
Answers research questions ✓

alit stores and retrieves. The agent thinks. You set the taste.

Setup

pip install agent-lit    # or: uv add agent-lit

Spin up Claude Code, opencode, Cursor, or whatever you use, then prompt:

Set up alit for literature review in this project. See https://github.com/Zhou-Hangyu/alit

From there:

Find and add the top 10 papers on vision-language grounding from the last 2 years.
Read the next 5 recommended papers and summarize each one.
What does the literature say about cross-modal attention mechanisms?

How it works

.alit/
├── papers.db    ← one SQLite file, entire knowledge base
└── pdfs/        ← auto-downloaded from arXiv

No servers. No API keys. No vector databases.

Update

pip install --upgrade agent-lit

Under the hood

  • Search: BM25 via SQLite FTS5
  • Ranking: PageRank on citation graph (pure Python)
  • Recommendations: PageRank + recency + taste matching
  • Synthesis: multi-stage funnel retrieval (~5K tokens to query 10K papers)
  • Enrichment: arXiv API (batched) with Semantic Scholar fallback
  • Backward compatible: schema auto-migrates on upgrade

License

MIT

Metadata

Release files for agent-lit 0.5.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for agent-lit 0.5.0
File Size Uploaded Provenance Actions
agent_lit-0.5.0.tar.gz 43.0 kB Without provenance Details for agent_lit-0.5.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for agent-lit 0.5.0
File Interpreter ABI Platform Actions
agent_lit-0.5.0-py3-none-any.whl Python 3 none any Details for agent_lit-0.5.0-py3-none-any.whl

Total release size: 82.6 kB

Release files / agent_lit-0.5.0.tar.gz

Download URL agent_lit-0.5.0.tar.gz
Size 43.0 kB
Tags Source
SHA-256 checksum
How to use checksums
8845876bfdc4b43da6267578887aa3f48218a9457b5f521384200f8f246be126
BLAKE2b-256 checksum
How to use checksums
125b87e9f3c896556cec2cdf0eb6de72eeda96b1b79d7f1271e6bd0ffa340b85
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.8.8

Release files / agent_lit-0.5.0-py3-none-any.whl

Download URL agent_lit-0.5.0-py3-none-any.whl
Size 39.6 kB
Tags Python 3
SHA-256 checksum
How to use checksums
1c015e699d77c4550d639af8f13532d993610ff2095b871540ce4147bbffcb8e
BLAKE2b-256 checksum
How to use checksums
63129f450a8d2ba3a91039b76a208b624e6e1679e79054c391546725cebc68f0
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.8.8

Security for agent-lit 0.5.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page