Lighthouse 3 is a high-reasoning autonomous research agent designed to provide C-Suite executives with high-signal, "noise-filtered" market intelligence. Developed for the Gemini 3 Hackathon, it leverages deep synthesis to transform raw global data into boardroom-ready strategic briefings.
The agent's output is rendered through a minimalist, executive-facing web portal: View Live Strategic Briefing
- High-Reasoning Synthesis: Leverages
gemini-3-pro-previewwiththinking_level=HIGHto identify "Hidden Connections" and secondary effects across global markets. - Grounded Research: Integrated with Google Search Grounding to ensure every briefing is anchored in real-time, verified data.
- Auditability: Every report includes a "Thought Signature", a transparent log of the AI's internal reasoning process to build executive trust.
- ROI-First Lens: Engineered to bypass "hype cycles," focusing instead on capital allocation, revenue drivers, and infrastructure shifts.
- LLM: Gemini 3 Pro Preview (Google GenAI SDK)
- Search: Google Search Tool (Grounding)
- Backend: Python 3.10 / Flask
- Deployment: Google Cloud Run (Containerized via Docker)
- Orchestration: Gunicorn (Production-grade WSGI server)
- Clone Repo:
git clone https://github.com/agbro/lighthouse-3 - Install Dependencies:
pip install -r requirements.txt - Configure Environment: Create a
.envfile with yourGEMINI_API_KEY. - Execute Research Agent:
python researcher.py(Generates the.mdbriefing) - Launch Local Portal:
python app.py(Serves the briefing at localhost:8080)
This project is optimized for Google Cloud Run. It utilizes a tailored .gcloudignore configuration to ensure lean container builds by excluding virtual environments and documentation while prioritizing mission-critical report assets.