Enterprise RAG with self-drawing charts

Ask a company’s documents a question and get the answer. Ask an analytical one and the answer comes with a chart, titled in the language you asked in.

Status: Phases 1–5 delivered
Stack: OpenAI: text-embedding-3-small, gpt-4o-mini · Pinecone · S3
Disclosure: Client withheld by agreement
Metrics: Shipped; metrics confidential.

What it is

Retrieval-augmented generation over a company’s own files, for people who want answers rather than search results. Analytical questions come back with a chart the system draws for that answer.

Process: what we built

PDF, DOCX, XLSX, CSV and JSON files are ingested to S3, embedded with text-embedding-3-small into Pinecone, per tenant, and answered by gpt-4o-mini. Charts are generated automatically and titled in the language of the question. Access is organised in role-based data rooms, and chats can be shared.

Ingest
- PDF, DOCX, XLSX, CSV, JSON
- S3
Embed
- text-embedding-3-small
- Pinecone, per tenant
Answer
- gpt-4o-mini
Chart
- Answer
- Auto-generated chart, titled in the asker’s language
Share
- Role-based data rooms
- Shared chats

Fig. 1 System schematic, drawn from the certificate’s facts; client details removed. The outlined step in blue is where the system decides.

Need something built to this standard?

Tell us what the system should decide or do. We’ll tell you plainly if AI isn’t the right tool.

Start a project [email protected]