RAG system over the ALD lab archive — accumulated documents and data, searchable and questionable in natural language. In active development.
2026
Fig. 1 System Architecture
Team members use the archive through a web app or an MCP connector in Claude / ChatGPT. The server only stores and searches — all LLM work runs elsewhere, so it costs nothing extra.
Fig. 2 Indexing Pipeline (Flow A)
An uploaded paper is split into chunks and becomes searchable immediately. Related-paper candidates are queued at the same time — no LLM involved.
Fig. 3 Relationship Graph (Flow B)
Two Claude models judge how each pair of papers relates. If they agree, the edge is confirmed; if not, it waits for human review.
A retrieval-augmented generation (RAG) system built on the ALD lab archive — indexing the group’s accumulated documents and data so they can be searched and questioned in natural language.
In active development; details and repository link to follow.