Python tools for presenter research profiles, participant learning plans, project proposals, text visualizations, and calendar exports for the Active Inference Symposium.
Version 3.3 publishes the final-lineup 2026 Presenter Program. Explore the 2026 public program, including abstracts, consenting participant profiles, downloadable source dossiers, and the PDF and Word books. The book is published on Zenodo: 10.5281/zenodo.24161029 (version 3.3.0; the earlier 3.2.0 edition remains available). The GitHub v3.3.0 software release attaches the same six book-package files. See the release record for downloads, book hashes and verification scope. Private organizing inputs remain outside this repository; 2024/2025 archives and the offline 2026 intake templates are preserved.
- Find a method or repository
- Quick start for 2026
- Interactive interface
- Configure paid analysis
- Command-line workflows
- Features and package API
- Review the next intake before publication
- Development and verification
- Public Symposium program
- Documentation
- Citation
Start with the searchable methods hub, workflow catalog, or methods and evidence boundaries. Nine workflows link ten reviewed public repositories, with pinned source evidence and explicit handoff limits. The native Template renderer is an existing integration; other cards guide separate learning, modeling, transcription or collaboration work.
uv run symposium organize methods
uv run symposium organize methods --format json
uv run symposium organize capabilitiesThese commands run offline without API keys. The website publishes a page sitemap, canonical/structured metadata, public discovery index and source guidance. Participant self-reports are excluded from the additional discovery index. See indexing and optional credentials.
Requires Python 3.12 or newer and uv. The development runtime is pinned to Python 3.12; installations use the committed dependency lock.
git clone https://github.com/ActiveInferenceInstitute/Symposium.git
cd Symposium
uv sync --locked --extra dev
uv run symposium --version
# The supplied 2026 scaffold is safe to initialize again.
uv run symposium edition init --year 2026
uv run symposium edition validate --year 2026
# Check the approved public snapshot and report pending program content.
uv run symposium organize readiness --public-dir data/public/2026Initialization and validation run offline without API credentials. Edition initialization currently supports macOS and Linux. Default validation checks structure and reports empty inputs as warnings. Before processing real data:
uv run symposium edition validate --year 2026 --require-dataThe last command deliberately exits nonzero until both eligible participants and valid presenter exports are present. Follow the 2026 organizer guide to populate the workspace, validate it, and run analysis.
data/inputs/aif_2026/
├── Public_Participant_Information.csv # Header only until organizers add data
├── presenters/ # One folder per new presenter
└── ... # Onboarding instructions and templates
outputs/2026_symposium/ # Generated locally when processing runs
├── people/
├── presenters/
├── proposals/
└── visualizations/
uv run python run.py # Default: 2026
uv run python run.py --year 2025 # Explicit historical edition
uv run python launch.py --year 2026
./symposium.sh --year 2026The menu offers data inspection, participant analysis, column summaries and word clouds, single-person background research and curricula, visualizations, and configuration checks. All edition-specific inputs and reports use the selected year. Importing the runner does not change the working directory or create files.
Copy .env.example to .env and set the credentials needed for your chosen workflow.
Copy config.example.json to the ignored config.json to customize models and paths.
if [ ! -e .env ] && [ ! -L .env ]; then cp .env.example .env; fi
if [ ! -e config.json ] && [ ! -L config.json ]; then cp config.example.json config.json; fiConfiguration precedence is defaults → explicitly loaded JSON → environment.
The runner loads project-local config.json; the CLI accepts --config config.json
before the command. Explicit CLI paths select the input and output for that command.
SYMPOSIUM_YEAR selects the runner edition; its --year option overrides that value.
SYMPOSIUM_DATA_DIR and SYMPOSIUM_OUTPUTS_DIR change the base storage paths.
Perplexity is used for web-assisted background research; OpenRouter supports profile analysis and curricula. Existing model defaults remain configurable; verify provider availability before a paid run. Setup and tests make no paid API calls. AI-generated research profiles and proposals require organizer review before sharing.
All existing explicit-path commands remain available. These examples process the new edition once authorized data has been added:
# Profile analysis (choose the provider explicitly).
uv run symposium --config config.json analyze participants \
--csv-file data/inputs/aif_2026/Public_Participant_Information.csv \
--output-dir outputs/2026_symposium/people --api-provider openrouter
# Background-only research and curriculum-only generation.
uv run symposium analyze backgrounds \
--csv-file data/inputs/aif_2026/Public_Participant_Information.csv \
--output-dir outputs/2026_symposium/people
uv run symposium analyze curricula \
--csv-file data/inputs/aif_2026/Public_Participant_Information.csv \
--output-dir outputs/2026_symposium/people --api-provider openrouter
# Presenter publications: OpenAlex topic and works CSVs, one folder per person.
uv run symposium analyze presenters \
--data-dir data/inputs/aif_2026/presenters \
--output-dir outputs/2026_symposium/presenters
uv run symposium generate profiles \
--data-dir data/inputs/aif_2026/presenters \
--output-dir outputs/2026_symposium/presenters --include-methods
# Visualizations read Markdown reports, rather than registration CSVs.
uv run symposium visualize all \
--input-dir outputs/2026_symposium/people \
--output-dir outputs/2026_symposium/visualizations --method pca --layout springgenerate projects accepts --profiles-dir, --output-dir, a user-supplied
--domain-file, and a catechism such as KarmaGAP, EUGrants, or Synthetic.
See uv run symposium generate projects --help for options.
Participant imports accept the 2025 survey questions, corresponding short labels,
and canonical field names. The future-impact question supports subsequent years.
Optional survey null markers are normalized to empty strings. Missing name/consent
headers, duplicate fields, malformed CSV rows, duplicate eligible
names, and colliding report filenames are rejected. Only explicit affirmative public
sharing consent (yes, y, true, 1, ignoring case) permits inclusion.
- Presenter analysis from OpenAlex topic and publication exports.
- Participant profiles, background research, curricula, and column summaries.
- Profile and methods synthesis, catechism-based project proposals.
- Markdown and JSON reports; PNG visualizations of text embeddings, networks, distributions, word clouds, and PCA diagnostics.
- Calendar utilities through
symposium.ScheduleExporter(ICS, JSON, CSV). - Generic explicit-path CLI commands and year-separated organizer workflows.
from pathlib import Path
from symposium.core.edition import Edition
from symposium.core.data_loader import DataLoader
edition = Edition(year=2026)
participants = DataLoader.load_participant_data(edition.participants_csv)
print(f"{len(participants)} eligible registrations")Python 3.10/3.11 support ended in version 3.0. Upgrade to Python 3.12+
and recreate the environment with uv sync --locked --extra dev; the development
pin remains 3.12. CI verifies 3.12, 3.13, and 3.14. The separate native Template
renderer continues to use its own Python 3.14 environment.
Keep raw registrations, real source mappings, consent decisions and candidate workspaces outside public checkouts. Build the candidate with the organizing workflow, then compare its verified public files with the current snapshot:
uv run symposium organize review-update \
--previous-public-dir data/public/2026 \
--candidate-public-dir "$SYMPOSIUM_PRIVATE_DIR/approved-public" \
--output "$SYMPOSIUM_PRIVATE_DIR/intake-review.json" --public-root .The report identifies additions, removals, changed public fields and artifact changes.
Participant removals require organizer review; the report does not guess why consent
or identity changed. Readiness reports content gaps separately from file integrity;
--require-complete fails when a published abstract or session title is missing.
Neither command certifies fresh consent or publishes changes. The v3.3.0 book uses
the November 12 final program lineup and its source abstracts. Presenter
profiles use only exact whole-label matches from reviewed current public sources.
See the
methods guide for the boundaries between source-grounded public
materials and optional AI-generated research drafts.
uv sync --locked --extra dev
./gate.sh # default arm: test-fast
./gate.sh all # every arm (see ./gate.sh list)gate.sh is the one entry point. The workflow jobs call these arms
instead of restating commands (test-fast is local-only), so the coverage gate and
the lint, build, program and documentation checks are the same text in both places.
The coverage regression gate is 80%. Tests cover historical
workflows, real temporary CSV files, offline onboarding, failure cases, and generated
reports. Tests requiring live providers are opt-in (uv run python -m pytest --run-real-api -m real_api) and require credentials. Recorded wall times per arm are
generated in gate measurements; the pinned GNN
conformance baseline is described in .github/GNN.md.
The Symposium Pages site
presents approved abstracts, consenting participant profiles, available dossiers and
abstract-book downloads, and explicitly scheduled sessions. The 2026 Symposium is
online and free, November 12–13 (UTC). The public build uses only the verified
data/public/2026 snapshot; it requires no organizer credentials or paid AI service.
The snapshot is dated November 12, 2026: 55 contributions, 19 explicit UTC
sessions, and 119 consenting participant profiles. Prerecorded talks without a
submitted video and withdrawn contributions are removed. The participant charts
still use the separately admitted 55-record public 2026 view of November 5;
they were not regenerated for this edition. The book keeps two missing abstracts
explicitly pending and credits the Institute followed by all 50 published
presenter/group display labels, without splitting composite credits.
See the publication and organizer guide for source
mapping, approval checks, native abstract-book production, and deployment boundaries.
- Method catalog, indexing, credentials.
- 2026 organizer guide: inputs, schema, readiness, analysis, review.
- Documentation index: user, API, architecture, and development guides.
- Data guide: historical and current data organization.
- Changelog, citation, license.
- Contributing, issues.
For the software:
@software{symposium2026,
title = {Symposium: Research Analysis Tools for Active Inference Symposium},
author = {Active Inference Institute},
version = {3.3.0},
year = {2026},
url = {https://github.com/ActiveInferenceInstitute/Symposium}
}For the book: Active Inference Institute and the published presenter/group credits (2026), 2026 Active Inference Symposium — Presenter Program, version 3.3.0, published November 12, 2026, DOI 10.5281/zenodo.24161029. Full book credits remain in the PDF/Word cover and citation metadata. The repository and release metadata use the MIT license.
Maintained by the Active Inference Institute.
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.