Skip to content

CONTEXT.md

Latest commit

 

History

History
81 lines (61 loc) · 3.73 KB

CONTEXT.md

File metadata and controls

81 lines (61 loc) · 3.73 KB
Raw

DevArch Task Routing

Route user intent to the appropriate workspace. Each workspace loads only its own context.

Workspaces

Workspace Commands Mental Mode
Extraction mine, build-db, serve, ingest-pipeline Get raw data from git repos
Analysis signals, analyze, cascade, extract-sessions Find patterns and insights
Reporting visualize, export-report, audit, validate, public-case-study Generate human-readable output
Multi-Project sync, global-viz, fetch-github, benchmark Aggregate across repos
Setup init, demo, status Configure new projects

Task Routing

Your Task Go Here Read You'll Also Need
Mine a repo stages/02-mine/CONTEXT.md Stage contract —
Build the database stages/03-build/CONTEXT.md Stage contract —
Run signal detection stages/04-detect/CONTEXT.md Stage contract shared/signal-heuristics.md
Run analysis vectors stages/05-analyze/CONTEXT.md Stage contract analysis-vectors/.md
Detect eras stages/04-detect/CONTEXT.md Stage contract shared/concepts.md (Era System)
Generate visualization stages/06-visualize/CONTEXT.md Stage contract —
Export report stages/07-report/CONTEXT.md Stage contract —
Run audit stages/08-audit/CONTEXT.md Stage contract —
Sync multiple projects Multi-project section below config/profile.json —
Create demo stages/01-setup/CONTEXT.md Stage contract setup/questionnaire.md

Pipeline Flow

setup → mine → build-db → signals → analyze → visualize → report → audit
                                                                    ↓
                                                              strategy (optional)

Optional at any point:
  cascade, serve, public-case-study, sync, global-viz

Cross-Workspace Flow

Setup (init, demo)
    ↓ project.json created
Extraction (mine → build-db)
    ↓ archaeology.db created
Analysis (signals → cascade → analyze)
    ↓ detected-signals.json + analysis-*.json created
Reporting (visualize → report → audit)
    ↓ HTML + markdown reports + audit result
Strategy (optional GTM from archaeology data)

Each workspace consumes output from the previous workspace. An agent in extraction never needs analysis vectors. An agent in analysis never needs visualization templates.

Checkpoints

  • After signals: Review detected signals before running analysis. Human checks threshold values.
  • After analyze: Review findings before generating visualizations. Human checks for meaningful results.
  • After audit: Review severity issues before considering pipeline complete. Human addresses CRITICAL/HIGH.

Supplementary Data

Any external data with timestamps can be added at any stage. Configure in project.json under supplementary_sources. Correlation runs automatically during signal detection (stage 04) and analysis (stage 05).

Supported: fitness data, YouTube history, calendar events, weather, lunar phases, any timestamped CSV/JSON.

Multi-Project

For multi-project analysis, configure config/profile.json with project paths, then run devarch sync or devarch global-viz.

Analysis Vectors

devarch analyze supports 6 vectors (run all or specify with -v):

  • sdlc-gap-finder — SDLC practice gaps
  • ml-pattern-mapper — ML pattern detection
  • agentic-workflow — AI/agent workflow identification
  • formal-terms-mapper — Formal methods terminology
  • source-archaeologist — Deep code archaeology
  • youtube-correlator — YouTube watch history correlation