Agent Deliberation: Deliberative Polyphony. Sequential multi-agent policy auditing & stakeholder impact digest.
Courtroom Sim: Adversarial Synthesis. Precedent-aware, turn-based Mock Trial sandbox with live API lookup.
Rights Checker: Jurisprudential Mapping. Maps user situations to statutes via fuzzy-grounded quote matching.
Cross-Bill Analysis: Legislative De-confliction. Detects statutory overlaps and literal conflicts across acts.
Consensus Map: Consensus Telemetry. Visual agreement heat maps charting ideological alignment between agents.
Personal Impact: Democratic Tailoring. Personalized, persona-steered impact digests for gig-workers & SMEs.
Fact Guardrails: Veracity Guardrails. Anti-hallucination quote-policing and validation using gemini judge.
State Bills: Federalist Ledger. Query, filter, and track regional state bills and assembly archives.

🏛️ CivicSync — Hackathon Submission

Democratizing access to complex legislation through open-source RAG, multi-agent deliberation, and zero-hallucination verification powered by Google Gemini and Elasticsearch, deployed on Hugging Face Spaces.

🌟 Inspiration

Legislation is complex, opaque, and difficult for the average citizen to navigate.

[!NOTE]
The Problem: The language of law is a barrier to justice and civic participation. Dense legal jargon alienates the public from understanding the very laws that govern their daily lives.

Our motivation was simple: bridge the gap between legalese and public comprehension. We built CivicSync as an open-source civic tech ecosystem designed to democratize access to statutory information and empower individuals with direct, understandable legal transparency.

🛡️ What it does

CivicSync is a transparent legislative analysis dashboard designed to parse, search, and simplify statutory legislation globally.

Translating Legalese: It processes dense legal drafts and statutory bills (such as the DPDP Act 2023 or Bharatiya Nyaya Sanhita used in our initial case studies) into plain, citizen-friendly language.
Persona-Oriented Impacts: Tailors summaries and impact assessments to specific citizen profiles (e.g., students, small business owners, local laborers) using Gemini 3.1 Flash.
Fact-Checking & Verification: Employs Gemini 3.1 Flash as an automated, fast-policing judge to compare generated summaries against the source text to ensure all citations and claims are 100% grounded.
Scalable Knowledge Search: Integrates a high-performance Elasticsearch engine to index, query, and search massive legal databases and precedent corpora (such as Indian Kanoon reference data) with sub-millisecond retrieval speeds.

⚙️ How we built it

We designed the codebase from the ground up for transparency, community collaboration, and production reliability:

flowchart TD
    subgraph "GitHub Repository"
        code["Source Code & Configs"]
    end

    subgraph "App Deployment: Hugging Face Spaces"
        fastapi["FastAPI Backend"]
        vite["Vite React SPA Frontend"]
        docker["Dockerized Environment"]
    end

    subgraph "Database & Retrieval"
        es["Elasticsearch Instance"]
        mcp["Elastic MCP Server"]
    end

    subgraph "AI Engine & Models"
        gemini["Google Gemini 3.1 Flash"]
    end

    code --> docker
    docker --> fastapi
    docker --> vite

    vite --> fastapi
    fastapi --> mcp
    mcp --> es
    fastapi --> gemini

Open Source: Hosted transparently on GitHub to invite peer review and collaborative enhancement.
Hosting Environment: Deployed directly on Hugging Face Spaces as a Docker space containing both the FastAPI backend and built Vite React SPA frontend.
Elasticsearch & MCP: Leveraged an Elasticsearch instance connected via the Elastic MCP Server to run optimized lexical DSL queries and hybrid vector search scoring over hundreds of thousands of legal sections and precedents.
Gemini-Powered Agents: Orchestrated multi-perspective analysis agents powered by Google Gemini 3.1 Flash for deliberation and courtroom simulation.

📂 Repository Structure

.
├── app/                 # FastAPI Backend (Gemini engine, MCP client, courtroom logic)
├── bills/               # Built-in legislation documents and reference bills (PDF & TXT formats)
├── civicsync-ui/        # Primary React + Vite + TS Frontend Dashboard (with Live Courtroom Simulator)
├── data/                # Regional/state bills datasets & local courtroom session transcripts
├── docs/                # Architecture design, PRD, and hackathon playbooks
├── frontend/            # Optional Streamlit demo user interface
├── public/              # Global static icons and assets
├── scripts/             # Setup and pre-caching helper scripts
├── tests/               # Backend validation test suites
├── CivicSync-v2-main/   # Legacy v2 codebase repository copy
├── CivicSync_v3/        # Git submodule pointer for V3 courtroom engine
├── Dockerfile           # Deployment container config for Hugging Face Spaces
├── render.yaml          # Render Blueprint deployment config
├── requirements.txt      # Python backend package dependencies
└── README.md            # Project documentation and developer instructions

⚠️ Challenges we ran into

[!WARNING]
Building robust legal tech requires navigating complex architectural transitions under tight timelines.

Search Migration: Transitioning the RAG pipeline from custom hybrid search algorithms to a production-grade Elasticsearch index while maintaining sub-second query performance and seamless index updates.
Deployment & Networking: Configuring the multi-stage Dockerfile inside Hugging Face Spaces to successfully compile the React/Vite client-side assets and serve them alongside the FastAPI backend on port 7860.
Marketing & Video Production: Translating a highly serious, technical legal application into an engaging, high-impact walk-through video under strict time constraints.

🏆 Accomplishments that we're proud of

True Civic Impact: Successfully launching an open-source tool that makes law understandable for everyday people.
Hugging Face Spaces & Elasticsearch Integration: Establishing a resilient, hybrid hosting architecture spanning Hugging Face Spaces, Elasticsearch, and the Google Gemini API.
Effective Advocacy: Creating compelling video and promotional content proving that technical civic tech can be marketed dynamically to a mainstream audience.

💡 What we learned

[!TIP]
User Experience (UX) is a core accessibility requirement when dealing with rigid, dense legal text.

Cloud Orchestration: Gained deep experience in managing cross-platform cloud deployments, environment syncing, and container pipelines inside Hugging Face Spaces.
Search Engine Optimization: Deepened our understanding of configuring index mappings, analyzers, and query parameters in Elasticsearch to retrieve highly context-relevant legal documents.
Social & Civic Engagement: Learned how to craft high-engagement social campaigns to drive public awareness and bring critical civic tech directly to the community.

🔮 What's next for CivicSync

Database Expansion: Incorporate pending bills, local municipal bylaws, and wider fields of global legal jurisprudence.
Elasticsearch Scaling: Implement advanced semantic vectors directly inside our Elasticsearch cluster to enable dense-sparse hybrid search directly at the database layer.
Multi-Lingual & Global Localization: Expand the interface support to multiple global and regional languages, adapting the parser schemas to serve the legislative frameworks of different democratic nations seamlessly.