How salamfarukh13-blip/rag-dip-diet works
This is a Streamlit nutrition assistant that combines a DIP knowledge-base retrieval workflow with personalized meal guidance. In the chat flow, the dashboard collects a question and profile, retrieves relevant ChromaDB chunks, and calls the dual-LLM agent: IBM Granite analyzes DIP rules and quantities, then Groq formats the response. The dashboard also provides weight-based targets and a visual meal plan. The excerpts do not show the dashboard’s collection-loading implementation, so its retrieval connection to the store is evidenced by the displayed query call; other unsampled details are omitted.
Main entry points: agent.py RAG Orchestrator · app.py Streamlit Dashboard · ingest.py Chunk Ingestion · dip_knowledge.txt DIP Knowledge Base
Components 15
DIP Knowledge
- DIP Knowledge Base
dip_knowledge.txt
Knowledge Retrieval
- Context Retrieval
app.py - Chunk Ingestion
ingest.py - ChromaDB Store
ingest.py
Answer Agent
- RAG Orchestrator
agent.py - Targets And Prompt
agent.py - Granite Analysis
agent.py - Groq Formatting
agent.py
User Dashboard
- Streamlit Dashboard
app.py - Profile Inputs
app.py - Visual Meal Plan
app.py - Chat Experience
app.py
Other components
- User
- IBM watsonx.ai
- Groq API
Connections 18
- User → Streamlit Dashboard: uses
README.md - Streamlit Dashboard → Profile Inputs: collects
app.py - Streamlit Dashboard → Chat Experience: presents
app.py - Streamlit Dashboard → Visual Meal Plan: presents
app.py - Profile Inputs → Visual Meal Plan: supplies weight
app.py - Chat Experience → Context Retrieval: queries
app.py - Context Retrieval → ChromaDB Store: searches
app.py - DIP Knowledge Base → Chunk Ingestion: loaded from
ingest.py - Chunk Ingestion → ChromaDB Store: writes chunks
ingest.py - Chat Experience → RAG Orchestrator: invokes
app.py - Profile Inputs → RAG Orchestrator: provides profile
app.py - Context Retrieval → RAG Orchestrator: provides context
app.py - RAG Orchestrator → Targets And Prompt: builds prompt
agent.py - Targets And Prompt → Granite Analysis: analyzes
agent.py - Granite Analysis → IBM watsonx.ai: calls
agent.py - Granite Analysis → Groq Formatting: passes analysis
agent.py - Groq Formatting → Groq API: calls
agent.py - RAG Orchestrator → Chat Experience: returns answer
app.py