Larry-has/ai_10022200083

CS4241 AI Exam 2026: Custom RAG chatbot (manual pipeline) for Ghana election results and 2025 budget analysis. # Academic Insights Assistant (Next.js + FastAPI Manual RAG) > **CS4241 – Introduction to Artificial Intelligence** > Academic City University — End of Second Semester Examination, 2026 **Student Name:** Lawrence Effah **Index Number:** 10022200083 This project is a full custom RAG chatbot system for academic evaluation, built without LangChain, LlamaIndex, or any prebuilt RAG framework. Frontend: - Next.js App Router UI for chat, retrieval inspection, prompt inspection, and logs. Backend: - FastAPI with manual ingestion, chunking, embeddings, FAISS indexing, hybrid retrieval, prompt construction, generation, and request-level logging. ## Live Deployment - Frontend (production): academic-insights-assistant… - Backend API (production): backend-lake-delta-78.verce… - Backend API Docs (Swagger): backend-lake-delta-78.verce… ## Backend API Quick Links - Swagger Docs: backend-lake-delta-78.verce… - OpenAPI JSON: backend-lake-delta-78.verce… ## Video Walkthrough - 2-minute walkthrough link: drive.google.com ## Rubric Coverage (A-G) | Part | Requirement | Status | Primary Evidence | |---|---|---|---| | A | Comparative chunking impact analysis write-up | Complete | part_a_chunking_analysis.md | | B | Advanced retrieval | Complete (hybrid TF-IDF + FAISS) | retrieval.py | | C | Prompt experiments + evidence | Complete | part_c_prompt_experiments.md | | D | End-to-end API pipeline and logs | Complete | api.py, pipeline.py | | E | Adversarial testing + RAG vs pure-LLM | Complete | part_e_adversarial_eval.md | | F | Architecture diagram + explanation | Complete | part_f_architecture.md | | G | Innovation component | Complete (feedback-aware retrieval memory) | part_g_innovation.md | ## Repository Structure ```text academic-insights-assistant-nextjs/ app/ # Next.js app entry src/ …