Architecture Governance
Internal control surface for EAR platform architecture. Inspect system domains, release states, protected systems, migration history, and build queue before making any platform changes.
Foundational builds establishing the EAR intelligence platform: Supabase integration, n8n orchestration, review workflow, forum intelligence, AI disclosure, source credibility, role model, event history, operator mode, jobs intelligence scaffolding, hybrid ingestion, multi-AI ratings, category specialist, multi-brand OS, entity intelligence, entity resolution, signal-entity pipeline, and pipeline consolidation.
Established the /jobs route with 60 seed job records, job card components, filtering UI, and the job_sources schema. ADR-022 governs this build.
UI and filtering improvements on the robotics jobs board.
Live ATS ingestion from Greenhouse, Lever, and Ashby. Ingestion orchestrator, provider adapters, source health tracking, and manual trigger API. ADR-024 governs this build.
Skills extraction (80+ taxonomy), vertical classifier (11 verticals), freshness decay engine, dedup engine, trend snapshots, company intelligence profiles, and admin source panel. ADR-025 governs this build.
Internal governance and architecture awareness layer. Domain registry, protected systems list, migration registry, build registry, ADR index, dependency map, architecture dashboard, and documentation. Fixes RISK-4 and RISK-6 from the Build 026 audit.
Full CAP pillar: funding tracker (/funding), startup profiles (/startups + /startups/[slug]), investor intelligence (/investors), defense & government monitor (/defense), capital heatmap (/capital), and nav CAP pillar. Schema: cap_investors, funding_rounds, ma_events, defense_contracts, startup_profiles.
Public job feed endpoints for SEO (trending today, highest-paying, by vertical), email alert subscriptions for skill and company tracking.
blocked by build 026AI-powered job matching and recommendations using embedding similarity on job descriptions and user skill profiles.
blocked by build 027Seeded international news sources (Asia, China, EU) into job_sources table. RSS and API sources only.
ATS ingestion foundation. Added Greenhouse, Lever, and Ashby provider support. Created ingestion orchestrator. Added source health tracking fields (last_checked_at, failure_count, credibility_score) to job_sources.
Workforce intelligence schema. Created ingestion_runs, job_skills, company_metrics, skill_trends, and trend_snapshots tables. Added vertical, freshness_score, content_hash, and dedup_group_id columns to robotics_jobs.
Pre-026 fix: wired actual ingestion_runs row persistence into the /api/jobs/ingest POST route. Previously the table was created but never written to. The GET route now reads recent runs from the table.
Capital intelligence schema. Created 5 new tables for the CAP pillar: cap_investors, funding_rounds, ma_events, defense_contracts, startup_profiles. Seeded with 10 investors, 15 funding rounds, 7 M&A events, 10 defense contracts, and 10 startup profiles.
WARNING: Systems listed below cannot be rewritten, renamed, or structurally refactored without an explicit human instruction referencing the protected system ID. Additive changes only.
Core coordination layer that wires all ingestion adapters, the dedup engine, vertical classifier, skills extractor, and source health updates. Any structural change risks breaking the entire live ingestion pipeline.
Defines the canonical Zod schema and TypeScript types for all job data. Used by adapters, orchestrator, API routes, and UI components. Breaking this schema breaks the entire jobs pipeline.
Each adapter implements the JobIngestionProvider interface against a live external API. Structural changes risk breaking production ingestion from real ATS platforms serving real robotics companies.
Content-hash based deduplication prevents duplicate jobs from appearing in the feed. The 30-day lookback window, normalization logic, and dedup_group_id assignment are calibrated for production data. Altering these will corrupt the dedup_group_id lineage in the database.
Classifies every ingested job into one of 11 canonical robotics verticals. The classifier patterns are calibrated against real job data. Changing vertical IDs would orphan existing database records.
80+ skill taxonomy with weighted confidence scores. Skills are persisted to job_skills table. Renaming or removing skills would orphan existing skill_trends and company_metrics records.
Exponential decay freshness scoring with a 14-day half-life. The decay formula is calibrated to keep jobs visible for ~60 days. Altering the formula would invalidate all existing freshness_score values in robotics_jobs.
Canonical taxonomy definitions used across all EAR domains. Any rename cascades across briefs, signals, entity records, and the jobs pipeline. Changes must be coordinated with a schema migration.
Approved canonical entity definitions that underpin the company resolver, signal classifier, and entity pipeline. Modifications require cross-domain impact assessment.
Company profile pages aggregate jobs-intelligence, funding, signals, and timeline data into structured public-facing profiles. These pages are indexed by search engines. Structural rewrites risk breaking SEO and external links.
| Domain | State | Owner | Protected | Tables | Dependencies | Last Updated |
|---|---|---|---|---|---|---|
Ontologyontology | FROZEN | platform | YES | 2 | none | 2026-05-12 |
Ingestion Pipelineingestion | VERIFIED | platform | YES | 3 | ontologysource-health | 2026-05-13 |
Jobs Intelligencejobs-intelligence | VERIFIED | platform | YES | 5 | ingestionsource-healthcompany-intelligence | 2026-05-12 |
Company Intelligencecompany-intelligence | STABLE | platform | no | 4 | ontologyingestion | 2026-05-12 |
Source Healthsource-health | VERIFIED | platform | YES | 2 | none | 2026-05-13 |
Capital Intelligencecapital-intelligence | EXPERIMENTAL | platform | no | 5 | company-intelligence | 2026-05-13 |
Research Intelligenceresearch-intelligence | STABLE | platform | no | 2 | ontologyingestion | 2026-05-01 |
Educationeducation | EXPERIMENTAL | platform | no | 0 | ontologyresearch-intelligence | 2026-04-01 |
AI Model Libraryai-model-library | SANDBOX | platform | no | 1 | ontology | 2026-04-15 |
Admin & Governanceadmin-governance | STABLE | platform | no | 0 | none | 2026-05-13 |
Canonical definitions, taxonomy, and approved vocabularies for all EAR verticals, skill categories, and entity types. The ground truth layer that all other domains must reference.
Real-time job and news ingestion from ATS providers (Greenhouse, Lever, Ashby), RSS feeds, ArXiv, GitHub Trending, HuggingFace, and Reddit. Handles fetching, normalization, deduplication, and persistence.
Workforce intelligence layer over the ingested job corpus. Includes skills extraction, vertical classification, freshness decay, trend snapshots, company hiring metrics, and admin analytics surfaces.
Entity intelligence profiles for robotics companies. Aggregates signals, job data, funding events, and timeline history into structured company pages.
Monitors the health, reliability, and credibility of all ingestion sources. Tracks failure counts, last check times, and credibility scores per source.
Robotics funding intelligence layer: funding rounds, M&A events, defense contracts, investor profiles, and startup intelligence profiles. The PitchBook pillar of the EAR OS.
ArXiv paper ingestion, academic signal detection, model release tracking, and research trend analysis. Feeds into intelligence briefs and trend signals.
Learning paths, glossary, plain-language briefs, and structured course content for the EAR EDU pillar.
Curated index of open-weight robotics AI models, foundation models, and deployment tooling. Includes model cards, benchmark comparisons, and hardware requirements.
Internal platform governance layer. Architecture dashboard, domain registry, protected systems list, ADR index, migration history, dependency map, and build state registry.