Skip to content

Repository files navigation

Fathom AI Clone

What it is

A production-oriented prototype of an AI meeting assistant inspired by Fathom. It focuses on turning recorded meetings into searchable transcripts, summaries, decisions, highlights, and action items.

How it works

  1. Google Auth: Secure authentication and identity provisioning.
  2. Calendar: Read-only integration to pull upcoming events.
  3. Meeting Capture: Autonomous recording layer (simulated for assessment).
  4. Transcript: Speaker diarization and timestamped transcription.
  5. Gemini: Generative AI processing of the raw transcript via @google/genai.
  6. Structured Intelligence: Extraction of action items, executive summaries, and key decisions using JSON schemas.
  7. Searchable Workspace: A clean, light-themed responsive dashboard to review past meetings and search transcripts.

Evaluation Note

The meeting-capture layer is simulated for this submission, per the challenge allowance. The post-meeting intelligence workflow is implemented natively in the application.

After signing in, the workspace is automatically provisioned with representative meeting data. Clicking "Join" on an upcoming meeting on the dashboard triggers a real-time simulation that injects a mocked transcript and routes you to the meeting intelligence view.

🛠️ Architecture & Tech Stack

The codebase follows a service-layer architecture with feature-oriented UI organization:

  • Services (Model): src/services/ (Prisma DB operations and Gemini logic)
  • Controllers: Thin /api/ routes that delegate business logic to services.
  • Views: Server Components that fetch data and orchestrate pure Feature Slices (src/features/).

Tech Stack:

💻 Running Locally

  1. Clone the repository

    git clone https://github.com/Reh1t/fathom-ai-clone.git
    cd fathom-ai-clone
  2. Install dependencies

    npm install
  3. Environment Variables Create a .env file in the root directory:

    DATABASE_URL="postgresql://..."
    DIRECT_URL="postgresql://..."
    NEXTAUTH_URL="http://localhost:3000"
    NEXTAUTH_SECRET="..."
    GOOGLE_CLIENT_ID="..."
    GOOGLE_CLIENT_SECRET="..."
    GEMINI_API_KEY="..."
  4. Database Setup

    npx prisma db push
    npx prisma generate
  5. Start the development server

    npm run dev

📝 Assessment Compliance

  • Agent capture verification is documented in CAPTURE-TEST.md.
  • Original agent transcripts are preserved in .agent-logs/.

About

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages