The problem
The documents sit in PDFs, wikis, tickets and chats. The team searches by hand. Sometimes they paste a chunk into a public ChatGPT. That model cannot see your files. It also cannot run anything in your environment.
What I leave running
A RAG over the docs we pick. Or an agent for one concrete case: queries, classification, reports.
The system searches those documents. It answers with a source. Whoever asks comes in with a company account.
- Load the documents (folder, wiki or API).
- Answers with title, fragment and link.
- An API or an internal chat.
- A log of questions. Reindex when the files change.
Usual stack: OpenAI, LangChain or LangGraph, pgvector, Pinecone or Chroma, a Python backend. Pick what the team can keep.
FAQ
What problem does this job solve?
The documents sit in PDFs, wikis, tickets and chats. A public ChatGPT cannot see them. It also cannot run anything in your environment.
What do you leave running?
A RAG or GPT over the docs we pick. Or an agent for one concrete case. Chat or API. A log of questions.
How do we start?
Email david@davidtovar.dev or call +34 657 23 05 07. We bound the case and set it up.
Notes on how I set this up are in the guides.
If you want me to set it up, email david@davidtovar.dev or call +34 657 23 05 07. Madrid.