axm-bib 0.4.0

axm-bib — Bibliographic tools: search papers, resolve DOIs, download & extract PDFs

axm-bib searches academic sources, resolves scholarly identifiers to BibTeX, and downloads or extracts papers. Its AXMTools share the same Python pipeline. Network results depend on provider availability; file-producing operations need an explicit destination when the default paper library is not appropriate.

Features

• Parallel search across Semantic Scholar, CrossRef, DBLP, OpenReview, arXiv and OpenAlex, with source diagnostics and date filters.
• DOI, arXiv, HAL, DBLP and OpenReview resolution; merge coherent identifiers for one paper or batch several papers.
• Venue enrichment and citation/reference traversal with OpenAlex fallback.
• Open-access PDF download, Markdown extraction and embedded image extraction.
• Topic digests with persistent seen identifiers and a dry-run mode.
• The same tools through the generic axm CLI and MCP discovery.

Installation

uv add axm-bib

Python 3.12 or newer is required. These pages describe the checkout; the published release can lag behind it. For this exact revision, use the workspace installation under Development. Commands below assume its environment is activated.

Set a contact email before using the network pipeline:

export AXM_BIB_EMAIL="you@example.org"

Replace this address with your own. The current pipeline constructs its Unpaywall client eagerly, so email is required for search, resolution and graph traversal as well as PDF downloads. Local PDF extraction does not require it. Tools do not prompt for missing credentials.

Quick Start

axm bib_search --query "attention is all you need" --limit 3

Read papers, sources_queried and sources_failed. Results vary with providers. Choose an identifier from a result, then resolve it:

axm bib_resolve --ref arXiv:1706.03762

The result includes bibtex, key, doi and entry_type; this call does not write a file unless --output-path is supplied. A paper title is not an identifier.

Usage

Tools

axm bib_extract --pdf ./paper.pdf --output-dir ./extracted
axm bib_digest --topics "diffusion models" --since 7d --dry-run
axm bib_graph --ref arXiv:1706.03762 --direction both --limit 10

Extraction writes <pdf-stem>.md and figures/; reruns overwrite matching files. bib_pdf normally writes under ~/axm/papers/{key}/. Its --output-path changes only the PDF destination: BibTeX and extracted content remain in the paper library. See the PDF guide before choosing destinations.

Python

from axm_bib.pipeline import BibPipeline

pipeline = BibPipeline(mailto="you@example.org")
result = pipeline.search("attention is all you need", limit=3)
for paper in result.papers:
    print(paper.title, paper.doi or paper.arxiv_id)

The root axm_bib namespace exports exception types and __version__. Pipeline, models and clients use module-qualified imports. MCP discovery requires axm-bib to be installed in the server's environment; see the MCP guide.

Documentation

• Published documentation
• Getting started
• Task guides
• Tool reference and Python API
• Architecture

Development

git clone https://github.com/axm-protocols/axm-knowledge.git
cd axm-knowledge
uv sync --all-groups
source .venv/bin/activate
make test-bib
cd packages/axm-bib
mkdocs build --strict

Workspace dependencies may require the sibling checkouts declared in the root pyproject.toml. The standalone build uses this package's mkdocs.yml and docs dependencies. It does not validate the workspace's combined navigation.

License

Licensed under Apache-2.0. See LICENSE.