CivicScore AI Framework
Real-time accessibility & compliance remediation
- Continuous WCAG 2.2 + Section 508 auditing
- Web + PDF/Word + media coverage
- Guided authoring and in-platform fixes
- Automated remediation layers
AI-first products for real-world operations
The Digital Garden builds AI solutions at the center of your workflows.
We ship practical automation: accessibility compliance that stays current, RFP triage that produces decision-ready summaries, cited municipal code research for contractors, and ML scoring pipelines teams can operate with confidence.
Evidence-first
Auditable outputs over guesses
Guardrails
Policy-driven automation
Deployable
Built for production ops
Featured
Municipal & County Code Intelligence
Cited ordinance answers for contractors, planners, and civic teams
- Jurisdiction-scoped Q&A with section citations
- Multi-city compare for permitting deltas
- Source adapters (Municode, AmLegal, municipal sites)
- Live pilot — try it now with demo access
Products
Four AI-first offerings designed to fit real workflows: compliance remediation, procurement triage, cited ordinance research, and batch scoring for student performance signals.
Municipal & County Code Intelligence
Cited ordinance research for contractors & civic teams
- Jurisdiction-scoped Q&A with section citations
- Multi-city compare for permitting deltas
- Source adapters (Municode, AmLegal, municipal sites)
- Honest coverage warnings + freshness signals
RFP Strategist Agent
Automated Go/No‑Go decisioning
- Outlook + Gmail ingestion with filters
- Deduped runs + traceable JSONL logs
- Evidence gathering behind portals
- Policy guardrails + fit scoring
Student Dropout Prediction
Event-driven batch scoring + drift awareness
- Validated feature contracts + schema checks
- Batch predictions with student ID joins
- Timestamped artifacts for traceability
- Drift metrics and investigate signaling
CivicScore AI Framework
Continuous accessibility and compliance remediation across enterprise content.
What it does
Continuous audit
- Scans web experiences, portals, and interfaces for accessibility barriers
- Audits uploads like PDFs and Word documents (including dense text)
- Tracks changing standards (WCAG 2.2 AA/AAA, Section 508)
Guided remediation
- In-platform guided authoring tools to fix issues at the source
- Clear “what to change” guidance with policy-aware guardrails
- Remediation workflows for content owners and reviewers
Automated fixes
- Automated remediation layers for common issues (where safe)
- Consistent output formatting for audits and reporting
- Audit trails for governance and compliance teams
Core features
Coverage
- Web content: pages, components, forms, and navigation
- Documents: PDF and Word upload pipelines
- Media: video/audio workflows (captions and alternatives)
Governance
- Compliance scorecards by site, department, and content type
- Prioritized remediation queues and ownership assignment
- Prompt guardrails to keep remediation on-policy
Reporting
- Regulatory-aligned findings with severity and impact
- Exportable reports for audits and procurement requirements
- Continuous monitoring to prevent regressions
RFP Strategist Agent
Turn procurement digests into decision-ready summaries—automatically.
How it works
1) Ingest
- Connects to Outlook (Microsoft Graph) or Gmail
- Applies filters, sender exclusions, and dedupe state
- Schedules runs or executes dry runs (no email send)
2) Evidence
- Extracts opportunities from listing rows (preferred)
- Navigates portals with Playwright + stealth
- Falls back to public web crawl when blocked
3) Decide
- GO / NO‑GO / MAYBE with policy guardrails
- Capability-fit scoring against your team profile
- Traceable reasoning with JSONL orchestration logs
Use cases
RFP triage
- Structured opportunity lists instead of scrolling emails
- Deduped processing so nothing gets reviewed twice
- One clean summary per opportunity
Capability fit
- Resume/capability matching against your services
- Hard policy overrides for consistent NO‑GO rules
- Memory search over past decisions (ChromaDB)
Procurement monitoring
- Scheduled runs that keep stakeholders in the loop
- Evidence URLs and snippets for quick verification
- Optional delivery via Resend (or silent dry runs)
Deployment
Private worker
- Designed to run as a private VPS worker
- Environment-managed secrets for OAuth/API keys
- Local dedupe state and local decision memory
Production scheduling
- DigitalOcean systemd service + timer path
- Nightly schedules or hourly procurement sweeps
- Dry-run mode for safe validation
Observability
- JSONL orchestration trace logs per run
- Configurable filters and policy profiles
- Artifacts suitable for audit and reviews
Docs
Interfaces
- MCP tools and orchestrator CLI
- Provider integrations (Outlook / Gmail)
- Resend delivery and dry-run outputs
Traceability
- JSONL event schema and example traces
- Evidence URL selection and ranking
- Policy guardrails configuration
Security posture
- Secrets, OAuth tokens, and least-privilege guidance
- Private worker deployment recommendations
- Data handling boundaries and retention
Municipal & County Code Intelligence
Plain-language questions, jurisdiction-specific answers, and citations back to official ordinance text — built for contractors, planners, and civic teams who cannot afford to guess. Try the live demo (login required — email us to request access).
Try the live demo
Explore jurisdiction-scoped Q&A, multi-city compare, and cited ordinance research on our hosted pilot. Demo access is gated — email us and we will send login credentials.
What it does
Cited research
- Ask natural-language questions about municipal ordinances
- Every answer links to section numbers and official source URLs
- Indexed excerpts with relevance scores and history notes
Jurisdiction-scoped
- Rules vary by city — pick a jurisdiction before you query
- Scope warnings when a question names a different municipality
- Coverage warnings for partial indexes (e.g. single-ordinance pilots)
Multi-city compare
- Ask one question across 2–3 cities side by side
- Surface permitting deltas: fences, setbacks, permits, and more
- Designed for bid prep and cross-jurisdiction planning
How it works
1) Catalog & ingest
- Jurisdiction catalog seeded from authoritative legal libguides
- Host-specific fetchers for Municode, American Legal, and municipal sites
- Repeatable CLI pipeline: catalog build → fetch → parse → embed
2) Retrieve & answer
- Vector search over indexed ordinance chunks (pgvector)
- Hybrid LLM mode: local embeddings + cloud generation
- Policy prompts that refuse legal advice and flag missing excerpts
3) Verify & audit
- Citation cards with section detail modals and source links
- Freshness signals from last ingest timestamps
- Query logs and eval harness for answer quality
Use cases
Contractors & builders
- Permit requirements, fence heights, setbacks, and advertising rules
- Fast pre-bid research without hunting across dozens of city sites
- Evidence you can attach to proposals and submittals
Planners & code officials
- Internal reference tool for ordinance lookup during reviews
- Compare how peer cities handle similar requirements
- Honest gaps surfaced instead of confident hallucinations
Civic & procurement teams
- Document intelligence layer for municipal compliance workflows
- API-first (FastAPI + OpenAPI) for integration into existing tools
- Local-first demo deploy; architecture supports hosted rollout
Pilot coverage
Iowa municipalities (v1)
- Des Moines, Cedar Rapids, Iowa City, Ames, Dubuque
- Moulton and Ute (partial coverage where noted)
- County codes on the roadmap — municipal pilot first
Stack
- FastAPI RAG API + Next.js contractor UI
- PostgreSQL + pgvector for catalog, chunks, and audit trail
- Provider interfaces for embeddings, LLM, and host-specific parsers
Guardrails
- Informational research only — not legal advice
- Rate limiting and structured JSON responses
- Always verify against official adopted ordinances before building
Student Dropout Prediction
Event-driven scoring pipeline with traceable artifacts and drift signals.
Highlights
Pipeline
- Feature uploads trigger preprocessing and batch scoring
- Predictions join back to student IDs for actionable output
- Stable file contracts for downstream use
Traceability
- Timestamped artifacts per run (inputs, outputs, summaries)
- Operator-friendly troubleshooting with clear S3 layout
- Repeatable runs without overwriting prior results
Drift
- Early warning when production data shifts
- Metrics suitable for alarms and dashboards
- Investigate signaling for ops response
Ready to talk?
Tell us what you’re trying to operationalize. We’ll recommend the right product path and set up a live demo.
Email: outreach@thedigitalgarden.io