agent-lit 0.3.0

Persistent memory for AI agents doing literature review. Zero dependencies. SQLite-only.

(-o-) alit

Your AI agent reads papers so you don't have to.

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

Got tokens to burn? Let your agent read 50 papers overnight and hand you a synthesis in the morning.

Tell your agent:

Use alit to manage my literature review. See https://github.com/Zhou-Hangyu/alit

Why alit

Your agent can web-search papers anytime — but forgets everything next session. alit is persistent memory. One agent reads 50 papers, another agent queries that knowledge instantly. Knowledge compounds across sessions.

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. No setup beyond pip install.

Install

pip install alit         # or: uv add alit
alit init                # creates .alit/ in your project

The agent skill auto-installs on first run.

Set your taste

alit taste "I'm into multimodal foundation models and how they learn cross-modal
representations. Love papers with clean ablations over pure benchmark chasing.
Especially interested in vision-language grounding and embodied AI."

Quick start

alit add "https://arxiv.org/abs/1706.03762"     # fetches metadata + PDF
alit import library.bib                          # or dump your Zotero/Mendeley
alit recommend 5                                 # ranked by your taste
alit ask "What are the key attention mechanisms?" --depth 2

Update

pip install --upgrade alit

Commands

Command	What it does
alit init	Initialize .alit/
alit add <title-or-url>	Add paper (auto-enriches arXiv, auto-tags)
alit find <query>	Search arXiv/S2 for papers by topic
alit import <file>	Bulk-add from URL file or BibTeX (.bib)
alit sync	Import from remembered BibTeX source (Zotero, etc.)
alit enrich	Batch-fetch metadata for papers missing abstracts
alit search <query>	BM25 full-text search
alit recommend [N]	Reading queue ranked by score
alit ask <question>	Cross-paper synthesis via funnel retrieval
alit read <id>	Guided reading view
alit show <id>	Paper details + citations
alit list	List papers
alit note <id> <text>	Append reading notes
alit summarize <id>	Store summary with model provenance
alit cite <from> <to>	Add citation edge
alit status <id> <s>	Set reading status
alit tag <id> <tags>	Set tags
alit taste [text]	Set or show your research taste
alit progress	Visual progress dashboard
alit stats	Collection overview
alit orphans	Find citations to missing papers
alit attach <id> <pdf>	Attach local PDF
alit fetch-pdf <id>	Download PDF from arXiv
alit delete <id>	Remove paper + citations
alit export [--format X]	Export as JSON or markdown

All commands support --json.

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

Development

git clone https://github.com/Zhou-Hangyu/alit
cd alit
uv sync
uv run pytest

License

MIT