How achyutak41/sastra_achyuta_227003005_cse works
This is an AUTOSAR HLD analysis assistant: engineers ask natural-language questions and receive evidence-grounded answers with citations and confidence. Its documented capabilities include ARXML analysis, semantic retrieval, architecture knowledge reuse, and human-supported decision-making. The visible runtime path is Streamlit → RAG pipeline → retrieval, with Ollama generating answers from retrieved evidence. Ingestion, chunking, embedding, and indexing are represented as product stages, but several connections are unverified because their implementations were not excerpted; evaluation scripts are omitted as supporting machinery.
Main entry points: rag_pipeline.py RAG Orchestrator · app.py Streamlit Assistant · search.py Semantic Search · build_faiss_index.py FAISS Index Builder
Components 17
ARXML Ingestion
- ARXML Parser
Code/ingestion/arxml_parser.py - ARXML Validation
Code/validation/validate_arxml.py - Structured JSON Export
Code/ingestion/export_json.py - Structured AUTOSAR Data
Input_Data/processed/EcuExtract.json
Knowledge Preparation
- Search Text Generation
Code/ingestion/create_search_text.py - Chunk Creation
Code/retrieval/create_chunks.py - Embedding Generation
Code/retrieval/create_embeddings.py - FAISS Index Builder
Code/retrieval/build_faiss_index.py
Evidence Retrieval
- FAISS Vector Index
Input_Data/processed/EcuExtract.faiss - Chunk Metadata
Input_Data/processed/EcuExtract_metadata.json - Semantic Search
Code/retrieval/search.py
Grounded Answers
- RAG Orchestrator
Code/rag/rag_pipeline.py - Grounding Prompt
Model_Prompts_Config/rag_prompt.txt - Ollama LLM
- Answer and Citations
Code/rag/rag_pipeline.py
Engineer Interface
- Streamlit Assistant
Code/app.py
Other components
- Engineer
Connections 20
- Engineer → Streamlit Assistant: asks question
Code/app.py - Streamlit Assistant → Semantic Search: loads retriever
Code/app.py - Streamlit Assistant → RAG Orchestrator: requests answer
Code/app.py - RAG Orchestrator → Semantic Search: retrieves evidence
Code/rag/rag_pipeline.py - Semantic Search → FAISS Vector Index: searches vectors
Code/retrieval/search.py - Semantic Search → Chunk Metadata: reads chunks
Code/retrieval/search.py - RAG Orchestrator → Grounding Prompt: loads template
Code/rag/rag_pipeline.py - RAG Orchestrator → Ollama LLM: generates answer
Code/rag/rag_pipeline.py - RAG Orchestrator → Answer and Citations: builds citations
Code/rag/rag_pipeline.py - Streamlit Assistant → Answer and Citations: displays response
Code/app.py - Engineer → ARXML Validation: submits ARXML
- ARXML Validation → ARXML Parser: checks input
- ARXML Parser → Structured JSON Export: parses document
Code/ingestion/export_json.py - Structured JSON Export → Structured AUTOSAR Data: writes JSON
Code/ingestion/export_json.py - Structured AUTOSAR Data → Search Text Generation: formats architecture
Code/ingestion/create_search_text.py - Search Text Generation → Chunk Creation: creates chunks
- Chunk Creation → Embedding Generation: embeds text
- Embedding Generation → FAISS Index Builder: builds index
- FAISS Index Builder → FAISS Vector Index: persists vectors
- FAISS Index Builder → Chunk Metadata: persists metadata