tuanduy35399/chat_service_nckh

How tuanduy35399/chat_service_nckh works

Mai Vang Chat Service is a Vietnamese plant-disease assistant: users submit a question (optionally with conversation history or image-detection results), the API prepares the request, and retrieval-augmented generation returns an answer. The main knowledge workflow loads Markdown files, chunks and embeds them into a persistent Chroma store; chat retrieves context and calls Google Gemini. A separate indexing script in rag_indexing describes another Markdown-to-Chroma approach, but its excerpt contains an undefined splitter, so its runtime status is uncertain. Links below are limited to documented or excerpted relationships.

Main entry points: rag.py RAG answering · ingest.py Text chunking · chunking.py Alternate indexer · main.py Image enrichment

Components 15

Chat service

Chat API google_model/main.py
Image enrichment google_model/main.py

Knowledge retrieval

RAG answering google_model/rag.py
Top-3 retriever google_model/rag.py
Chroma store google_model/vector_db

Knowledge indexing

Markdown knowledge data
Document loading google_model/ingest.py
Text chunking google_model/ingest.py
Google embeddings google_model/ingest.py
Alternate indexer rag_indexing/chunking.py
Sentence embeddings rag_indexing/chunking.py
Chroma collection rag_indexing/chunking.py

Other components

User
YOLO detections
Gemini chat model

Connections 15

User → Chat API: submits question google_model/main.py
User → Image enrichment: provides request google_model/main.py
YOLO detections → Image enrichment: supplies results google_model/main.py
Image enrichment → RAG answering: calls ask google_model/main.py
Chat API → RAG answering: calls ask google_model/main.py
RAG answering → Top-3 retriever: invokes retrieval google_model/rag.py
Top-3 retriever → Chroma store: retrieves documents google_model/rag.py
RAG answering → Gemini chat model: invokes model google_model/rag.py
Markdown knowledge → Document loading: loads files google_model/ingest.py
Document loading → Text chunking: splits documents google_model/ingest.py
Text chunking → Google embeddings: embeds chunks google_model/ingest.py
Text chunking → Chroma store: indexes chunks google_model/ingest.py
Markdown knowledge → Alternate indexer: reads Markdown rag_indexing/chunking.py
Alternate indexer → Sentence embeddings: encodes chunks rag_indexing/chunking.py
Alternate indexer → Chroma collection: adds chunks rag_indexing/chunking.py

Drawn by GitDiagram on November 11, 2026 from the repository's file tree, README and source files. Read as Markdown ·