Technical Document: How OpenGov Encyclopedia Would Work

1. High-Level Architecture

External Sources (Public Only)

├── Federal Register API / RSS

├── USAspending.gov V2 API

├── Curated high-signal agency pages (~200–300 URLs)

└── Agency-submitted URLs/PDFs (via simple form)

↓ (scheduled Lambda or webhook)

Orchestration Layer (AWS GovCloud Lambda or similar)

├── Grok (GSA OneGov API – inherited FedRAMP controls)

├── Optional secondary FedRAMP LLM (consistency check)

├── Cargo validation rules + Redis cache

└── Clearance Dashboard (simple internal MediaWiki page or lightweight app)

↓ (only approved changes)

MediaWiki Core (FedRAMP-authorized hosting)

├── Citizen-Centric Pages (USWDS-integrated skin)

├── Cargo tables (API-first knowledge graph)

├── MediaWiki API (for Grok bot edits)

└── Audit / revision tags table

2. Seeding the Initial Content (Phase 1: Weeks 1–4)

Step 1: Infrastructure Setup (Week 1)

Deploy MediaWiki + Cargo + USWDS-aligned skin on FedRAMP hosting.
Define core Cargo tables (Agency, Program, Organization, Topic).
Configure Grok bot account via GSA OneGov.

Step 2: Seed Agencies & Major Organizations (Weeks 2–3)

Use the official USA.gov A-Z Agency Index + agency “About” pages as the seed list (~150–200 URLs).
Grok batch job (one-time run):
- Prompt: “From this official agency page, extract: name, parent, mission summary, website, key sub-components, leadership. Output valid wikitext + Cargo fields. Cite source URL. Confidence score required.”
- Grok creates pages + populates Cargo tables.
All items route to Clearance Dashboard for batch attestation (high-confidence auto-pass after validation).

Step 3: Seed Programs & Initiatives (Weeks 3–4)

Sources: Federal Register notices, USAspending.gov API, curated agency program pages.
Grok prompt: “Create Program page from this source. Fill Cargo: sponsor, purpose, start_date, duration, funding, related agencies. Output wikitext + Cargo. Cite source.”
Review via Clearance Dashboard.

Step 4: Seed Task/Topic Pages (Week 4)

Grok uses seeded data: “Create ‘Prepare for a Disaster’ task page linking relevant programs/agencies. Add plain-language explanation and official USA.gov links.”
Links always point back to agency/USA.gov as source of truth.

3. Ongoing Updates (Post-Week 4)

Daily Grok orchestration monitors curated high-signal list + Federal Register + USAspending API.
Change detected → Grok proposes update → Clearance Dashboard.
Agency staff forward URLs via form → Grok processes → attestation.
No broad scraping — only targeted, public, high-value pages.

4. Technical Safeguards

Scalability — Cargo for narrative/relationships; PostgreSQL offload for high-volume numerics if needed.
Auditing — Mandatory revision tags (source URL, timestamp, Grok score, attestation ID) in separate table.
Transparency — Page badges show “AI-drafted – human-attested” + Trust Score tooltip.
Accessibility — Early WCAG 2.2 AA audit; ARIA landmarks.
Security — FedRAMP hosting, inherited controls from OneGov.

This setup ensures OpenGov Encyclopedia starts small, grows intelligently, stays accurate, and never competes with agency sites or USA.gov — it only adds context and connections that drive users back to the official sources.

OpenGov Technical

Contents