About Egregora

What is Egregora?

Egregora is a system that transforms WhatsApp conversations into analytical blog posts. Using artificial intelligence (LLM), it:

1. Analyzes group conversations
2. Identifies topics and emerging narratives
3. Synthesizes discussions into structured posts
4. Preserves the complexity and divergences of collective thought

How does it work?

Ultra-Simple Pipeline

WhatsApp ZIP → Parse → Anonymize → Group → Enrich → LLM → Posts

Privacy First

• Automatic anonymization: All names converted to UUID5 pseudonyms
• Deterministic: Same person always gets the same UUID
• Full opt-out: Any participant can leave with /egregora opt-out
• Optional aliases: /egregora set alias "Name" for humanized identity

See ALIASES.md for details.

LLM Editorial Control

The LLM (Gemini) has complete control over:

• ✅ What's worth writing (filters noise automatically)
• ✅ How many posts (0-N per period)
• ✅ All metadata (title, slug, tags, summary)
• ✅ Content quality (editorial judgment)

Technology

• Parsing: Python + Ibis on DuckDB (DataFrames)
• LLM: Google Gemini (multi-turn tool calling)
• RAG: DuckDB VSS + Parquet (3072-dim embeddings)
• Site: MkDocs Material
• Privacy: UUID5 + opt-out + filtering

Open Source

This project is open source: github.com/franklinbaldo/egregora

---

Egregora v2 - Ultra-simple WhatsApp → Blog pipeline