aiminister errorlogy

Errorlogy

https://errorlogy.com — analytical platform for governance error patterns (non-accusatory framing).

Errors in governance as observable objects: the gap between what was declared, what was known, and what was decided.

Errorlogy models governance errors as observable objects (gap: declared / known / decided).
politic.bar is the first product: an analytical catalog of public error cards without accusatory language.

Status


MVP	Active — `errorlogy-mas/` + GUI clients in this repo
Deployment	Local development only (no production API host here)
errorlogy.com	Marketing site — not the MAS API
Docs	Last sync ~September 2026
Docker	Not required

Analysis layers

DATA → WMS → CB/SF/MP/GT/HM/... → α → ACC → PNO → FPD

More detail: docs/concept/ · Obsidian: ERRORLOGY_MVP_OBSIDIAN/

Components

Component	Path	Role
errorlogy-mas	`errorlogy-mas/`	FastAPI backend — 14-agent pipeline, taxonomy v16
errorlogy-gui	`errorlogy-gui/`	Electron + Vite + React desktop UI (v0.2.x)
errorlogy-gui-v2	`errorlogy-gui-v2/`	Browser UI (forecast, streams)
Umbrella contracts	ai-native-gov	Institutional topology & cross-layer schemas

Local development only

All services bind to localhost (127.0.0.1). There is no hosted API in this repo.

Backend (MAS API)

cd errorlogy-mas
python -m venv .venv && .venv\Scripts\activate   # Windows; use source .venv/bin/activate on macOS/Linux
pip install -r requirements.txt
cp .env.example .env                             # LLM keys locally — never commit
python api/main.py                               # → http://127.0.0.1:8000/docs

Desktop install (Electron): the packaged app expects Python 3.10+ with MAS deps in errorlogy-mas/. Prefer py -3 or an explicit venv — a generic python on PATH (e.g. Hermes/conda) may lack errorlogy-mas packages. See errorlogy-gui/README.md.

Smoke tests

Path	Command	Keys
Engine only	`python examples/run_challenger.py --engine-only`	None — deterministic numerics
Full pipeline	`python examples/run_challenger.py`	LLM keys in `.env`

Run from errorlogy-mas/. Output: examples/challenger_output.json.

Frontends

	errorlogy-gui	errorlogy-gui-v2
Path	`errorlogy-gui/`	`errorlogy-gui-v2/`
URL	`http://localhost:5173`	`http://localhost:5174`
Runtime	Electron or browser (Vite)	Browser only
Role	Full desktop UI — taxonomy, globe, MAS metrics, ingest	Focused forecast UI — streams, case analyze, methodology

Terminal 1 — API: cd errorlogy-mas && python api/main.py
Terminal 2 — v1: cd errorlogy-gui && npm install && npm run dev:vite (proxies /api → 127.0.0.1:8000)
Terminal 2 — v2: cd errorlogy-gui-v2 && npm install && npm run dev

Without LLM keys: use Engine only on the Analyze page (v1) or the engine-only smoke above.

Pipeline:

Scout → WMS → Classifier → Alpha → PNO → ACC → EGD → T4D → CAT → FPD → LBI
      → Red Team → Card Compiler → Neutrality Audit

Source of truth for code: errorlogy-mas/AGENTS.md, ontology: errorlogy-mas/data/errorlogy_unified_taxonomy_v16.json.

Cross-layer API (MVP iter 1) — institutional activation stub (INSTITUTIONAL_MODEL, no μ/analyze):

Method	Path	Description
`POST`	`/api/events/cross-layer`	Frame & persist cross-layer event
`GET`	`/api/events/cross-layer`	List events (`?story_id=`, `?event_type=`, `?limit=`)
`GET`	`/api/events/cross-layer/layers`	Valid `institution:*` layer enum
`GET`	`/api/events/cross-layer/{event_id}`	Single event

Schemas vendored from umbrella: errorlogy-mas/schemas/. OpenAPI: /docs.

Integrations

Optional services wired in MAS (all local). Details: errorlogy-mas/README.md.

Integration	Role
Exa	Ingest web search and source discovery before Scout
OpenRouter / GLM	Long-form narrative agents (`z-ai/glm-5.2`) via `OPENROUTER_API_KEY`
Z.ai	Direct GLM provider (`ZAI_API_KEY`)
zvec	Local KB demo — hybrid FTS + vector (`examples/zvec_kb_demo.py`)
Cross-layer stub	Institutional activation events (`/api/events/cross-layer`) — ai-native-gov contracts, no μ pipeline

Repository layout

The workspace combines active development (new MVP) and an archive of sketches (early iterations).

Path	Status	Purpose
`errorlogy-mas/`	ACTIVE	Multi-agent politic.bar backend: 14-agent pipeline, taxonomy v16, FastAPI, multi-LLM router (built by Claude)
`errorlogy-gui/`	ACTIVE	Electron + Vite + React desktop UI v0.2.4 (~90% API integration)
`errorlogy-gui-v2/`	ACTIVE	Browser UI v0.1 — forecast, streams, methodology (port 5174)
`ERRORLOGY/errorlogy_old_version/`	OLD / SKETCH	Early artifacts: politic.bar v0.6, AGIU, spec, taxonomy copies
`ERRORLOGY_MVP_OBSIDIAN/`	Documentation	Obsidian: concept, taxonomy, map, work journal

More on what Claude built: Obsidian — errorlogy-mas active MVP.

OLD SKETCH (do not extend without explicit request)

Path	Contents
`…/Windows_old_MVP/Politic Bar (pre errorlogy)/`	Methodology v0.6, seed cases, 8-agent pipeline
`…/AGIU/`	Hono health + demo FastAPI analytics
`…/Cursor_Project/`	Full web-MVP specification

Documentation

Agent rules: AGENTS.md, errorlogy-mas/AGENTS.md
Obsidian vault: ERRORLOGY_MVP_OBSIDIAN/ — home
Concept docs: docs/concept/
Umbrella contracts: ai-native-gov — institutional topology and cross-layer schemas
politic.bar methodology: Obsidian — politic.bar MVP sketch
Cursor: .cursor/rules/
OSS evaluation funnel: docs/oss-integration-funnel.md — open-source candidate evaluation funnel; tracker research/oss-candidates.yaml; checklist python research/score_candidate.py
Harness engineering: docs/reference/harness-engineering/README.md — eval/agent harness, principles and checklist for MAS

Idea

Errorlogy — governance errors as observable objects (gap: declared / known / decided).
politic.bar — first product: analytical catalog without accusatory language.

Navigation Menu