Academic Inbox
A local intelligence layer for academic WhatsApp groups.
A Windows-first system for bringing academic messages and files into one local archive through Beeper Desktop. The project brings together deduplication, organization, search, and local AI around information that is otherwise scattered across group conversations. This overview describes the system direction; individual capabilities remain under development.
Useful information. Everywhere but together.
Announcements, documents, and repeated messages arrive across several academic WhatsApp groups. Finding the original file or recovering the context of an announcement requires navigating a stream built for conversation, not retrieval.
One archive, kept local.
The intended workflow connects selected groups through Beeper Desktop to a local archive. Each stage has a clear responsibility, from preserving source material to making it discoverable.
- Selected academic groups via Beeper Desktop
- Ingest messages and files, retaining source context
- Deduplicate and organize in SQLite
- Retrieve through FTS5 and local embeddings
- Explore through a FastAPI / Jinja interface with Ollama
Retrieval before generation.
Reliable ingestion, traceable source material, and predictable search are the foundation. Full-text search and semantic retrieval address different needs; local AI should help interpret retrieved material without obscuring its origin. The proposed stack includes SQLite FTS5, local embeddings, and Ollama.
An evolving local system.
The project is in development, with attention on ingestion consistency, duplicate handling, and useful local retrieval. Detailed implementation notes and measured results will follow as those pieces are verified.