Najjar07/localbuka-ai-case-study

AI-powered restaurant recommendation assistant for kano state, Nigeria. # LocalBuka – AI Restaurant Recommendation Assistant ## Overview LocalBuka is a prototype restaurant discovery and recommendation assistant focused on restaurants and food in Kano State, Nigeria. ## Features - Cuisine preference matching - Exact and partial meal matching - Budget matching - Distance-based filtering - Restaurant rating consideration - Spicy-food preference - Explainable recommendations - Free-text command-line conversational assistant ## Project Structure ```text localbuka-ai-case-study/ ├── data/ │ └── restaurants.json ├── src/ │ ├── recommender.py │ ├── test_recommender.py │ └── chat_assistant.py ├── README.md ├── REFLECTION.md └── requirements.txt ``` ## Requirements Python 3.9+ is recommended. The current prototype uses only Python's standard library and does not require a paid LLM API. ## How to Run From the project root: ```bash python src/test_recommender.py python src/chat_assistant.py ``` ### Example ```text You: I need something spicy and cheap near me ``` Type `exit` to close the assistant. ## Recommendation Approach The system scores restaurants using cuisine, preferred meals, budget, distance, rating, and spicy-food preference. Exact meal combinations receive a stronger match than partial matches. Results are sorted by score and accompanied by explanations. ## Conversational Assistant The assistant accepts free-text requests, extracts supported preferences using lightweight keyword/rule matching, and passes them to the same recommendation engine. ## Testing The system was tested with different user profiles, including: - Hausa/Northern Nigerian food preferences with Tuwo Shinkafa + Miyan Kuka - Budget, spicy Beef Suya preference - International-food preference with Grilled Chicken ## Production Considerations Possible failures include misunderstood free-text requests and outdated restaurant information. Mitigations include better NLP, clarification questions, verified and regularly refreshed data, and monit …