A public research-data and exploratory tooling archive from the Active Inference Institute. The repository brings together literature searches, PDF keyword extraction scripts, a PubMed query notebook, ResearchRabbit exports, and an ontology snapshot related to active inference and the free energy principle.
This is an archive of research inputs and exploratory analysis artifacts rather than a packaged Python library. The data files are preserved as collected snapshots; scripts and notebooks may require adjustment for current APIs and dependencies.
- Contents
- Quick start
- Documentation
- Reproducibility and data provenance
- Canonical ontology home
- Citation
- License
- Contributing
- Audit trail
- InstituteOS sidecar
| Area | What it contains |
|---|---|
Literature_Analysis/PDF_Scrape/ |
Python scripts and CSV parameters for counting keywords in a local PDF corpus. |
Literature_Analysis/Pubmed_to_Coda/ |
A PubMed/NCBI Entrez notebook plus PMID, DOI, and exported article metadata CSVs. |
Literature_Analysis/ResearchRabbit/ |
ResearchRabbit graph data, DOI mapping, image export, and the original bundle archive. |
Literature_Analysis/CitationSearch_PublishOrPerish/ |
Publish-or-Perish Google Scholar export in TSV, ODS, and XLSX formats. |
Ontology_Snapshot/ |
December 2022 export of the ontology vocabulary: 408 terms and proposed definitions. See Canonical ontology home for the current release. |
The repository has no build step. Start with the area README files above, then inspect the CSV, spreadsheet, JSON, PNG, and notebook artifacts directly.
Open the notebook only for inspection until the research owner selects a revival cohort and non-overwriting output. Its abstract/author/join limitations and relative export can affect a rerun or overwrite historical FEP_ActInf.csv; read the complete preservation warnings.
The PDF scraper expects a local PDF corpus at Literature_Analysis/PDF_Scrape/resc/pdfList/, which is not included in this repository. Its parameter files and invocation details are documented in the PDF_Scrape README.
The default fast arm combines archive checks, recorded Publishing-copy digest agreement, dependency-free gate contracts, and controlled scraper regressions. It checks relative Markdown targets, tracked Python syntax, output-free notebook JSON, and tracked checkpoint exclusion. This describes configured behavior, not a current full-gate or scientific-validation receipt:
python .github/scripts/gate.py fastList arms with python .github/scripts/gate.py --list. Plain gnn verifies an already prepared installed parser; --install explicitly requests installation, while --gnn-source-root /exact/checkout/src/gnn deliberately selects narrower source qualification, never fallback. See the GNN guide, run guide, and historical timing page.
The documentation hub provides a quickstart, a data provenance inventory, and an architecture overview alongside the area READMEs above.
The committed literature files are dated exports and snapshots. Their exact query settings and source metadata are not fully recorded in the repository, so they should not be treated as a continuously updated literature index. Preserve the original files when producing a new analysis and record any new query date, filters, or transformations alongside the result.
The PubMed notebook queries NCBI's public Entrez API and writes FEP_ActInf.csv. NCBI service usage should follow the current Entrez documentation.
The ontology snapshot in this repository is historical evidence. The ontology
itself is published from
Active_Inference_Ontology,
which holds the current v5 content release (May 25, 2023, 429 terms) with stable
identities and typed relations, archives the
December 12, 2022 snapshot
of the same vocabulary export, and publishes the
GNN vocabulary bridge
that resolves ActInfOntologyAnnotation labels used by Generalized Notation
Notation models.
402 of the 408 terms in Ontology_Snapshot/ also appear in
the v5 release, and the definition columns remain proposed text in the source
export. Use the canonical repository when you need current terms, definitions,
or identifiers.
Use the archived Zenodo record and DOI when citing this repository:
- DOI: 10.5281/zenodo.7449027
- Record: Zenodo 7449027
A machine-readable citation is also provided in CITATION.cff.
Unless a source artifact states otherwise, repository documentation and research artifacts are made available under CC BY 4.0. Attribution and indication of changes are required. Upstream source services and exported records may have their own terms.
See CONTRIBUTING.md. In particular, do not commit credentials or private datasets, do not silently rewrite frozen data exports, and verify documentation against the files and code that actually exist.
The historical documentation review is recorded in REVIEW_LOG_2026-08-02.md; current decisions and recorded evidence are in TO-DO.md. The historical review is not a current gate receipt.
This repository carries a committed .aii/ sidecar with curated
InstituteOS metadata (aii-sidecar/v1): category, maturity, ecosystem relations,
key artifacts, provenance, and portable tasks. It is read by
InstituteOS tooling and
does not affect the project's build or runtime.