How karthickrajathandapani/healthcare-sales-insight-prediction-and-rag works
Mosaic HCP Assistant is a field-sales dashboard combining HCP follow-up workflows, predictive insights, PDF reporting, and a conversational assistant. A rep uses the dashboard or chatbot; Flask serves filtered HCP data and results from KPI calculations, forecasting, and risk scoring. Chat queries are routed to analytics, prediction, or retrieval, then answered through Groq; the retrieval branch uses hybrid search over a Pinecone index. The README documents UI capabilities and service integrations; some internal wiring is unsampled, so those links are omitted where not evidenced.
Main entry points: app.py Flask routes · chain.py Chat engine · index.html Dashboard UI · vectorstore.py Pinecone index
Components 21
Dashboard experience
- Dashboard UI
templates/index.html - Dashboard controls
static/js/dashboard.js - Chat client
static/js/chatbot.js - Prediction charts
static/js/predictions.js
App and HCP data
- Flask routes
app.py - HCP engagement CSV
data/cleaned_hcp_engagement_dataset.csv - Data and KPIs
rag/data_loader.py
Predictive insights
- Rx forecasting
ml/forecasting.py - Engagement risk
ml/risk_model.py - Prediction APIs
app.py
Chat and retrieval
- Document chunking
rag/chunking.py - Pinecone index
rag/vectorstore.py - Hosted embeddings
rag/pinecone_embeddings.py - Hybrid retrieval
rag/retriever.py - Cross-encoder reranker
rag/reranker.py - Query routing
rag/analytics.py - Chat engine
rag/chain.py
Reporting
- PDF report builder
reporting.py
Other components
- Field rep
- Groq LLM
- Pinecone service
Connections 26
- Field rep → Dashboard UI: uses
README.md - Dashboard UI → Dashboard controls: loads controls
README.md - Dashboard UI → Chat client: loads chat
README.md - Dashboard UI → Prediction charts: loads charts
README.md - Dashboard controls → Flask routes: calls dashboard APIs
README.md - Chat client → Flask routes: sends chat
README.md - Prediction charts → Flask routes: requests predictions
README.md - Flask routes → Data and KPIs: loads data and KPIs
app.py - Data and KPIs → HCP engagement CSV: reads
rag/data_loader.py - Flask routes → Rx forecasting: uses forecast model
app.py - Flask routes → Engagement risk: uses risk model
app.py - Flask routes → Prediction APIs: serves results
README.md - Flask routes → Document chunking: builds documents
app.py - Flask routes → Pinecone index: builds or loads index
app.py - Pinecone index → Hosted embeddings: configures embeddings
rag/vectorstore.py - Hosted embeddings → Pinecone service: requests vectors
rag/pinecone_embeddings.py - Pinecone index → Pinecone service: reads and writes index
rag/vectorstore.py - Flask routes → Hybrid retrieval: builds retriever
app.py - Chat engine → Hybrid retrieval: retrieves context
README.md - Chat engine → Query routing: routes and answers
rag/chain.py - Chat engine → Cross-encoder reranker: reranks results
rag/chain.py - Cross-encoder reranker → Hybrid retrieval: reranks candidates
README.md - Chat engine → Groq LLM: generates responses
README.md - Flask routes → Chat engine: streams chat replies
app.py - Flask routes → PDF report builder: builds export
app.py - Field rep → PDF report builder: exports report
README.md