Lawgic Legal Track

AI-first platform with Context Engine: Lawgic Legal Track organizes cases, syncs filings & mail, and its Copilot drafts, analyzes and suggests next steps—so lawyers focus on strategy, not admin.

Comment

Inspiration

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned## Inspiration

Generative AI rewrote the rules of knowledge work overnight, yet 90 % of Spanish-speaking law firms in Latin America still run on Word macros and shared drives—an untapped blue-ocean market for digitalisation. We saw the chance to leapfrog legacy software and give lawyers an AI-native workspace that turns every new filing, email, or note into structured insight, without adding extra clicks.

How We Built It — Dynamic Context Engine (DCE)

The core is a Context Box that is born and retired with each matter (Track). All Track functions talk only to their own box.

Layer Purpose Tech highlights
Core Context Immutable facts—jurisdictions, parties, case IDs Strongly-typed tables
Dynamic Context Everything that changes—documents, events JSON-friendly columns for flexible filtering
AI Analysis Key dates, insights, impact scores Stored as versioned JSON blobs
Metadata Timestamps, users, update history Auto-filled on every change

Event-Driven Update Flow

  1. Single Source of Truth (SST) — an event engine inside our .NET backend; direct DB writes are forbidden.
  2. Each event type (DocumentAdded, DocumentDeleted, EmailReceived, …) has a dedicated EventHandler.
  3. Handlers update the Context Box, trigger OCR/LLM pipelines, and append fresh metadata.
  4. Two endpoints keep the API clean:
  • GET /contextbox to fetch current state
  • PUT /contextbox/update with { eventType, payload } to queue any update
    1. Heavy work (OCR + AI) runs in a background thread, so the user sees “upload successful” instantly while analysis finishes a few minutes later.

Stack Snapshot

  • Backend: .NET 8, PostgreSQL + pgvector
  • AI: Azure OpenAI (o3-mini) via Retrieval-Augmented Generation
  • Realtime & Auth: Firebase
  • CI/CD: GitHub Actions → Azure Container Apps / Vercel
  • Frontend: Next.js 14, TypeScript, Tailwind CSS

What We Learned

  • AI moves faster than any spec—design thin orchestration layers so models can be swapped without UI rewrites.
  • Every lawyer touchpoint is an opportunity to capture context automatically; the less they type, the more they adopt.

Next Steps

  1. Promote the Context Box to a plug-and-play module for a larger agent framework, where our native mini-tools become callable actions.
  2. Roll out connectors for additional LATAM jurisdictions to solidify first-mover advantage.
  3. Expose public webhooks so third-party apps can push events into the Context Box and receive timeline-aware suggestions in return.

Built With

Share this project:

Updates