How We Built an AI-Powered Legal Timeline Generator

TL;DR: At Hack the Law Cambridge, we built an AI-powered Event Timeline Generator using Momen, ChatGPT-4o, and Gemini 2.5—no code required. It extracts events from legal text, detects conflicts between witness statements, and visualizes everything in a timeline. This post breaks down the full architecture so you can reverse engineer it yourself.

⚖️ The Legal Pain Point: Reconstructing Timelines from Testimonies

Litigation teams often spend hours parsing through witness statements, deposition transcripts, and contradictory accounts to rebuild a factual timeline. It's manual, high-stakes, and prone to human error.

So we asked: Can we automate this?

🛠️ The Tool: AI-Powered Timeline Generator (Demo App)

We built this app for:

• 🧑‍⚖️ Litigation associates
  


• 🧑‍💼 Paralegals
  


• 🧑‍💻 Legal tech teams
  

Core Features:

• Upload or paste raw text (statements or transcripts)


• AI extracts events and timestamps


• AI detects and highlights contradictions


• Interactive timeline view with source quotes


• Structured database for case history

🚫 Why Not Just Use ChatGPT?

Sure, you can paste text into ChatGPT—but legal workflows need:

• Structure


• Repeatability


• Traceability


• Collaboration

That’s why we used Momen, a no-code platform where you can build full-stack apps with backend logic, UI, databases, and AI agents.

⚙️ How It Works (Full Stack Breakdown)

🧩 Database Schema

We created 6 interrelated tables:

• Statement


• Timeline_event


• Conflict


• Event_evidence


• Event_in_conflict


• 
  Analysis (parent record for each case)

🧠 Two AI Agents:

1. timeline_extractor (ChatGPT-4o)
   
   
   → Parses each statement, extracts events, timestamps, witnesses




2. conflict_detector (Gemini 2.5)
   
   
   → Compares statements and flags conflicting events

🔁 Actionflows = No-Code Backend Logic

✅ process_statements

• Triggered when user clicks “Generate Timeline”


• Creates an analysis_id
  


• Saves input statements


• Calls insert_events for each statement

✅ insert_events

• Calls timeline_extractor
  


• Saves structured events to timeline_event table

✅ insert_conflicts

• Triggered when user clicks “Detect Conflicts”


• Calls conflict_detector
  


• Stores results in Conflict and Event_in_conflict tables

🖼️ Frontend (Also No Code)

Built using Momen’s visual UI components:

• Text inputs for witness statements


• Button to trigger timeline generation


• Interactive timeline list view (subscribed to timeline_event)


• Sidebar to show conflict details

💡 Conflicting events show in red, with side-by-side quote comparison.

🚀 Timeline: Built in 2 Days

Even under hackathon time pressure, the app:

• Parsed multi-page legal text


• Built a timeline in under 30 seconds


• Flagged conflicts with full traceability

📉 Total cost? ~$99 (includes both LLM agents + infra).

🔍 Want to Reverse Engineer It?

You can explore the full setup—database, agents, and Actionflows—here:

👉 Reverse engineer the full Momen build (view-only link)

Yes, you can view every detail of how it works.

For a more detailed breakdown, check it here

🧭 What’s Next?

Because it’s built with Gemini 2.5 (multi-modal), future versions could:

• Analyze deposition videos or CCTV


• Merge transcripts + video into one timeline


• Integrate with document management systems

Drop questions or ideas in the comments!