Operational rulebook for the Active Fedference project.
Enforcement varies by file — see AGENTS.md's REQUIRED vs
AESTHETIC table for what is gate-checked versus convention.
Apply the software: application-guide.md ·
../examples/README.md
Contract & identity: ../ISA.md · ../README.md · ../STANDALONE.md
Current public release v1.0.4: DOI 10.5281/zenodo.21972644
· GitHub release
· public GitHub repository
· Zenodo record · prior v1.0.3 record · v1.0.2 record · v1.0.1 record · v0.1.0 record
| I want to… | Start here |
|---|---|
| Aggregate my own labeled categorical posteriors | application-guide.md; ../examples/README.md, example 05 |
| Compare the three server rules | ../examples/README.md, example 02 |
| Exercise direct, process, socket, and replay boundaries | ../examples/README.md, example 03 |
| Produce and verify an application receipt | ../examples/README.md, example 05; ../src/fedference_cli/README.md |
| Produce and verify a registered research receipt | ../examples/README.md, example 04; ../src/fedference_cli/README.md |
| Understand the mathematical scope | core/conceptual-foundations.md |
| Contribute software or documentation | development/quickstart.md; development/modularity.md |
| Reproduce the manuscript surfaces | manuscript/rendering_pipeline.md |
| Audit a release boundary | reference/verification-commands.md; reference/zenodo-release.md |
| Folder | Purpose |
|---|---|
core/ |
Science narrative, architecture, experiments & artifacts |
development/ |
Agent rules, quickstart, modularity, testing, style |
manuscript/ |
PDF/web pipeline, accessibility boundary, tokens & labels |
operations/ |
Output layout, troubleshooting, FAQ |
reference/ |
Verification, API/schema stability, and Zenodo release boundary |
research/ |
Indexed scholarship, claim, statistical, visual, complexity, RedTeam, runtime/composability, and CLI/orchestration audits |
security/ |
Runtime/development trust boundaries, abuse paths, mitigations, and no-claim constraints |
todo/ |
Scoped forward roadmap pages linked from TODO.md |
../README.md— project pitch and run commandsapplication-guide.md— adapt the software to caller-owned posteriors../examples/README.md— deterministic public-API example ladderdevelopment/quickstart.md— contributor and reviewer-snapshot reproductioncore/conceptual-foundations.md— why this project existsoperations/faq.md— common questions
development/agent_instructions.md— repository editing rules and visual QAcore/architecture.md— layer boundariesdevelopment/testing_philosophy.md— zero mocks + coveragedevelopment/modularity.md— before adding software, scripts, reports, figures, or documentationmanuscript/rendering_pipeline.md— before editing outputsmanuscript/accessibility.md— before accessibility or publication-surface claimsdevelopment/style_guide.md— before editing sourcemanuscript/tokens-and-labels.md— before editing manuscript prosereference/zenodo-release.md— before reserving, updating, uploading, or publishing a DOI deposition
- Active-inference readers: begin with the Friston belief-sharing mechanism
and categorical recovery identity in
core/conceptual-foundations.md, then follow the POMDP and belief-sharing modules incore/architecture.md. - Federated-learning readers: begin with PVI/FedGVI client objectives,
source-protocol parity, site factors/cavities, and the three distinct server
aggregation rules in
core/architecture.md, then inspect the registry/receipt boundary incore/experiments-and-artifacts.md. - Implementation-first readers: run the numbered programs in
application-guide.mdand../examples/README.md, then usereference/verification-commands.mdfor the example, distribution, and release-boundary probes.
| Task | Read first |
|---|---|
src/fedference/* |
core/architecture.md, development/style_guide.md |
src/fedference_cli/*, src/analysis/*, or src/publication/* |
development/modularity.md, core/architecture.md |
tests/* |
development/testing_philosophy.md |
manuscript/*.md |
manuscript/tokens-and-labels.md, ../manuscript/SYNTAX.md |
scripts/* |
core/architecture.md, ../scripts/AGENTS.md |
examples/* |
../examples/README.md, development/modularity.md |
| Federation transport / deployment | security/active_fedference-threat-model.md, todo/true-multi-machine-federation.md |
Pipeline / output/ |
operations/output-layout.md |
| Roadmap / TODO scope | todo/README.md, ../TODO.md |
uv run --locked --extra dev pytest tests/ \
--cov=src --cov-fail-under=90 -qFull probe list: reference/verification-commands.md.
Mermaid diagrams are deliberately kept in the Markdown source for GitHub and
local readers. Every block in README.md and docs/ has one non-empty
accTitle, one non-empty accDescr, quoted punctuation-bearing labels, and an
immediately adjacent plain-language text equivalent. The validator discovers
the live diagram count rather than relying on a fixed inventory. Every block
must pass the static contract and, before a release render, an actual Mermaid
CLI SVG probe:
uv run --locked python scripts/validate_mermaid.py
uv run --locked python scripts/validate_mermaid.py --render --renderer npx \
--output-dir .tmp/mermaid-renderTechnical file inventory and REQUIRED vs AESTHETIC table:
AGENTS.md.