Pulsar

Pulsar is an intelligent, automated lead generation and outreach platform designed to bridge the gap between Ideal Candidate Profiles (ICPs) and high-quality leads. By combining a Next.js frontend with a distributed agent architecture powered by FastAPI and Playwright, Pulsar automates the discovery, validation, and outreach process across multiple platforms.

---

Architecture

Pulsar is built as a modular ecosystem:

• Frontend (web-app): A Next.js application handling the standard CRUD operations, authentication, and campaign management. Integration with Supabase ensures robust data persistence and real-time updates.
• Agent Service (ml-service): A FastAPI-based backend that orchestrates specialized agents. It handles short-lived API requests directly and manages long-running background tasks.
• Task Queue: Powered by BullMQ and Redis. Long-running discovery and analysis jobs are offloaded to this queue to ensure reliability and scalability.
• Real-time Feedback: Progress and results are emitted from the server to the frontend via Server-Side Events (SSE), providing a seamless user experience.

---

Agents

At the heart of Pulsar is a network of agents:

• Master Agent (WIP): Responsible for interpreting product descriptions and ICPs to create and delegate campaign tasks.
• Sub-Agents: Specialized workers for specific platforms.
  • Instagram Agent (Functional): An autonomous web-agent that:
    • Authenticates and manages sessions securely.
    • Scours hashtags and profiles based on specific criteria.
    • Intercepts GraphQL/REST traffic for high-fidelity data extraction.
    • Matches profiles against the defined ICP.
    • Injects visual indicators (thinking status) into the controlled browser for observability.

---

Status & Roadmap

Completed

• Next.js + Supabase Foundation: Complete CRUD for users and campaigns.
• Distributed Infrastructure: Redis + BullMQ integration functional.
• FastAPI Agent Framework: Dedicated service for agent orchestration.
• Instagram Web Agent: Fully functional lead discovery and profile analysis via web automation.
• Real-time Streaming: SSE implementation for live result updates in the web UI.

Work in Progress / Next Steps

• Master Agent: Full automation of campaign creation from product descriptions.
• Additional Platform Agents: Expanding to LinkedIn, Twitter, and other strategic platforms.
• API Integrations: Transitioning to official APIs where available and free to improve efficiency.
• CSV Export: Ability to download discovered leads in a structured format.
• Outreach Automation: Direct messaging and follow-up sequences.

---

Local Development

Pulsar is (containerized) for easy setup.

Prerequisites

• Docker & Docker Compose
• Node.js (for web-app local dev)
• Python 3.10+ (for agent local dev)

Getting Started

1. Clone the repository:

   git clone <repository-url>
   cd Pulsar

2. Environment Setup: Create .env files in both web-app/ and ml-service/ based on their respective .env.example templates.

3. Launch Infrastructure:

   docker-compose up -d

   This will spin up the Redis cache and the essential services.

---

Philosophy

Pulsar is built to be deterministic where possible and agentic where necessary. We prioritize high-fidelity data extraction and human-like interaction patterns to ensure long-term sustainability and quality over quantity in lead generation.