malakazz18/Tunisia-Renewable-Energy-Transition-Assistant

A document-grounded chatbot that answers questions about Tunisia's renewable energy incentive programs (PROSOL, PROSOL Elec, PROSOL Elec Économique) using Retrieval-Augmented Generation (RAG). # Tunisia Renewable Energy RAG Chatbot A document-grounded chatbot that answers questions about Tunisia's renewable energy incentive programs (PROSOL, PROSOL Elec, PROSOL Elec Économique) using Retrieval-Augmented Generation (RAG). ## Problem / Motivation General-purpose LLMs don't reliably know the specifics of niche national programs like Tunisia's solar subsidy schemes, and will often generate plausible-sounding but incorrect details (wrong subsidy amounts, made-up eligibility rules). This project demonstrates a RAG pipeline that grounds every answer in real source documents, and explicitly says "I don't know" when the answer isn't present in the retrieved chunks — rather than guessing. ## Documents Used - **`cahier_des_charges_prosol_elec_2023.pdf`** — ANME's official program specification for PROSOL Elec (2023) - **`guide_sections_1_2_5.txt`** — Extracted sections (energy context, regulatory framework/Law 2015-12, and investment/incentive environment) from ANME's "Projets d'Énergie Renouvelable en Tunisie — Guide Détaillé" (2019), originally a 329-slide presentation; trimmed to the sections relevant to subsidies, financing, and eligibility ## Architecture / Pipeline ``` Documents (PDF + TXT) ↓ Text extraction & loading (PyPDFLoader / TextLoader) ↓ Chunking (RecursiveCharacterTextSplitter, chunk_size=1000, overlap=200) ↓ Embeddings (Gemini embedding model) ↓ Vector storage (ChromaDB) ↓ Similarity retrieval (top-k=4 chunks per question) ↓ Grounded generation (Gemini LLM, answers restricted to retrieved context) ↓ Answer + cited source chunks ``` ## Technologies - **LangChain** — document loading, text splitting, orchestration - **ChromaDB** — vector storage and similarity search - **Google Gemini** — embeddings (`gemini-embedding-001`) and generation (`gemini-3.6-flash`) - **Python / Google Colab** ## Installation & Usage 1. Open the notebook in Google Colab 2. Run the install cell to set up dependencies 3. Enter your Gemini API key when prompted (`get …