Robog

Robog

Conversational access to robot documentation — hackathon project for the Telmekom Challenge 2026.

Turn buried robot docs (manuals, datasheets, specs, accessories guides) into accurate, clear, source-grounded answers for both technicians and non-technical users.

My github - https://github.com/swop-tech/Robog.git

🎯 Brief vs. behaviour — what the judges ask, what the bot answers, what is still open: docs/brief-alignment.md

📄 Challenge analysis & suggested path: docs/challenge-analysis.md

🔍 Review of the team's pipeline diagram: docs/pipeline-review.md

🧰 Stack decisions (embeddings, RAG framework, vector DB): docs/stack-decisions.md

🧹 Data processing (clean, chunk, metadata, parameter index): docs/data-processing.md

🔎 Retrieval engine (routing, language policy, evaluation): docs/retrieval.md

💬 Answer composer + evidence validation (G5, Bonus 02): docs/answering.md

🔌 FastAPI backend (/search, /ask, /products, /media): docs/api.md

🖥️ Web chat UI (ui/, served at GET /): ui/README.md

🧠 Vector database from zero + the Gemini → Chroma run: docs/vector_database_explanation.md, docs/gemini_chromadb_explanation.md

📦 Dataset layout: data/README.md

🎯 Approach: hybrid RAG + structured parameter index, retrieval engine exposed as an API (Bonus 01), multilingual ask/answer (Bonus 02), web chat as the required interface.