How bala-kopparthi/ragpilot works
RAGPilot is a local document-question-answering assistant. Its workflow has two stages: ingesting supported QA documents into a persistent vector index, then retrieving relevant chunks for a question and generating a grounded answer with source previews. The graph preserves document loading, chunking, embedding, vector storage, retrieval, prompting, and answer presentation. Relationships are based on the README and the sampled scripts; the local Ollama service and generated ChromaDB directory are runtime dependencies, not source-tree modules.
Main entry points: query.py Question CLI · ingest.py Index Embedding · documents QA Documents
Components 13
Document Ingestion
- QA Documents
documents - Document Loading
ingest.py - Chunking
ingest.py - Index Embedding
ingest.py
Question Answering
- Question CLI
query.py - Query Embedding
query.py - Chunk Retrieval
query.py - Context Prompt
query.py - Answer Generation
query.py - Answer and Sources
query.py
Local AI Services
- Ollama
Other components
- User
- Vector Store
Connections 15
- User → Question CLI: asks
query.py - QA Documents → Document Loading: loads
ingest.py - Document Loading → Chunking: splits
ingest.py - Chunking → Index Embedding: embeds chunks
ingest.py - Index Embedding → Ollama: requests embeddings
ingest.py - Index Embedding → Vector Store: stores vectors
ingest.py - Question CLI → Query Embedding: configures
query.py - Query Embedding → Ollama: requests embeddings
query.py - Question CLI → Chunk Retrieval: invokes
query.py - Chunk Retrieval → Vector Store: searches
query.py - Chunk Retrieval → Context Prompt: supplies chunks
query.py - Question CLI → Context Prompt: builds chain
query.py - Context Prompt → Answer Generation: formats prompt
query.py - Answer Generation → Ollama: generates answer
query.py - Question CLI → Answer and Sources: prints answer and sources
query.py