About EnergyFile
EnergyFile is a Retrieval-Augmented Generation (RAG) document platform built for an energy-market research and consulting practice, giving analysts a single searchable home for reports from major energy publishers.
The Challenge
The client’s research library had grown to thousands of PDF, Word, Excel, and PowerPoint reports scattered across folders, with no way to search by meaning or get a cited answer without manually opening dozens of files.
Our Solution
We built a document manager that:
- Indexes every format — PDF (with OCR), DOCX, XLSX, PPTX, TXT, and Markdown — through an async background worker
- Answers in natural language with Google Gemini embeddings and Qdrant vector search, always citing the source document and page
- Enforces subscription tiers with chunk-level access control, so a Reader-tier subscriber never sees Researcher-only content
- Ships a persistent AI chat widget on every page, so a question can be asked without leaving the current document
Technical Implementation
- Backend: FastAPI, MongoDB, Qdrant, arq (async workers), Redis
- AI: Google Gemini for embeddings and chat
- Frontend: Next.js, Ant Design
Results
The live library indexes 3,000+ documents across 8 publishers (including Argus Media, S&P Global Platts, and the IEA), with OTP and Google sign-in and a subscription panel showing each user’s tier and entitlements.
