Locally running LLM & browser orchestration to automate sales prospecting. Reuses a persistent Chrome session so LinkedIn sees a normal user, not automation. Ollama generates personalized connection notes, first messages, and follow-ups; reply detection reads message-thread DOM classes to decide when to stop. No third-party SaaS, no API fees, nothing leaves the machine.

STACK

Python · FastAPI · Playwright · Ollama / Gemini · WebSockets · Vanilla JS UI

STATUS

LIVE · CONNECT + MESSAGE + FOLLOW-UP + REPLY DETECTION

PRIVACY

Data stays local · CSVs only · no telemetry

// CLICK ANY ENTRY — FULL SPEC OF WHAT WAS BUILT

• 00◆
  Home

  Project status + quick nav
  →
• 01▸
  Project Overview

  What it does, tech stack, file map
  →
• 02▸
  Architecture

  System diagram, request flows, WebSocket protocol
  →
• 03▸
  Features

  Connection, first message, follow-up detail
  →
• 04▸
  Development Roadmap

  What's built, what's planned
  →
• 05▸
  Configuration & Setup

  .env, Chrome profile, Ollama + Gemini
  →
• 06▸
  Outreach Playbook

  Campaign strategy + safety rules
  →
• 07▸
  Debugging & Logs

  Selectors, bugs fixed, known issues
  →
• 08▸
  Data & CSV Schema

  Schema for connections / messages / followups
  →