gapit

Mass screening of contigs and reads for antimicrobial resistance and virulence genes. An agent-first Python reimplementation of abricate: byte-compatible TSV plus first-class JSON and Markdown.

Why gapit

• Drop-in abricate replacement. Same BLAST pipeline, same hit rules, abricate-format TSV on stdout. Parity with abricate 1.4.0 is a release gate, checked by diffing gene calls against real abricate (pixi run -e parity parity).
• Machine-readable by design. Versioned output schemas (gapit.report/1, gapit.reads/1, gapit.summary/1), self-describing via gapit schema, typed JSON error envelopes on stderr, documented exit codes. An agent can discover the whole contract without reading docs.
• Reads, not just contigs. gapit screen --r1/--r2 screens FASTQ through minimap2, and --r1 accepts assembly FASTA directly (content-detected, map-ont forced) for a fast presence survey. abricate cannot screen raw reads.

Install

pixi manages the environment, including the external binaries (BLAST+, minimap2):

git clone https://github.com/indexofire/gapit.git
cd gapit
pixi install
pixi run gapit --version

The package supports Python 3.11+; the dev environment pins 3.14.

Quick start

# Screen contigs (abricate-compatible TSV on stdout, --db defaults to ncbi)
gapit screen contigs.fa --db ncbi

# Agent- and human-readable outputs
gapit screen contigs.fa --db ncbi --format json
gapit screen contigs.fa --format md

# Screen FASTQ reads; --read-type picks the minimap2 preset
gapit screen --r1 sample_R1.fastq.gz --r2 sample_R2.fastq.gz --read-type sr --db card
gapit screen --r1 ont_reads.fastq.gz --read-type map-ont --format json

# Summarize report tables into a gene presence/absence matrix
gapit summary *.tsv

# Databases: card + vfdb install offline from bundled snapshots, others fetch on demand
gapit db fetch
gapit db fetch ncbi
gapit db list
gapit db outdated                # flag stale databases and newer bundled snapshots
gapit db search "tet(M)"         # look up genes across every installed database
gapit db build mydb my_genes.fa --tsv my_meta.tsv   # custom db from any FASTA

# Introspection
gapit list              # installed databases (abricate --list compatible)
gapit schema report     # JSON Schema of gapit.report/1

Contig screening follows abricate defaults (--minid 80, --mincov 80). Reads-mode presence defaults to 90% alignment breadth (--min-breadth 90).

Databases

A database is a directory under the datadir, resolved from $GAPIT_DATADIR, then ~/.local/share/gapit/db (override per call with --datadir). Twelve providers exist; card and vfdb ship inside the wheel and install with zero network, the rest download from upstream when fetched.

gapit also reads datadirs built by abricate itself (legacy ~~~ headers). The reverse does not hold: gapit-native databases use the gapit/v1 header format (see SPEC.md §11), which abricate cannot read. Protein databases such as bacmet2 screen through blastx.

Output contract

• stdout purity. Data on stdout, diagnostics on stderr, always. --quiet silences stderr only.
• Deterministic. Stable sort orders, fixed tool parameters, no wall-clock timestamps inside data payloads.
• Errors print a JSON envelope to stderr and exit nonzero:

{"schema": "gapit.error/1", "code": "...", "message": "...", "context": {...}}

MCP server

gapit ships an MCP (Model Context Protocol) stdio server so agent runtimes can screen assemblies without parsing CLI output: gapit mcp or the gapit-mcp console script speaks newline-delimited JSON-RPC 2.0 on stdin/stdout (no extra dependencies — the protocol is hand-rolled). It exposes nine tools: screen (gapit.report/1 by default; aligner minimap2 for a fast assembly survey), screen_reads (FASTQ via minimap2, gapit.reads/1), summary (gapit.summary/1), schema, db_list, plus the database tools db_fetch, db_build, db_search, and db_outdated so an agent can self-provision and inspect databases mid-session. Tool failures return isError: true with the gapit.error/1 envelope as text. Register it with an MCP client:

{"mcpServers": {"gapit": {"command": "gapit-mcp"}}}

Development

User documentation lives in docs/index.md, rendered at https://indexofire.github.io/gapit/.

SPEC.md is the parity contract, PLAN.md the roadmap, AGENTS.md the contributor guide, CHANGELOG.md the change history.

License

gapit is MIT-licensed. It is a behavioral reimplementation of abricate (GPL-2.0) and copies no Perl code; abricate itself remains GPL-2.0. Bundled database content retains its original upstream licenses (SPEC.md §9).