How kans23-cloud/rag_based_environmental_recommendation_system works
This is a Streamlit assistant for personalized CO₂-reduction recommendations. A user submits an activity query; the app calls the recommendation agent, which retrieves relevant activity emissions from Chroma and uses an Ollama language model to produce recommendations. Separately, src/vector_store.py turns the CSV dataset into embedded documents and stores them in Chroma; the sampled files do not show this preparation step being invoked by the app. The graph reflects those visible relationships; the README is empty.
Main entry points: agent.py Recommendation agent · vector_store.py Vector-store builder · app.py Streamlit app · dataset.csv Activity dataset
Components 8
User interface
- Streamlit app
app.py
Recommendation runtime
- Recommendation agent
src/agent.py - Query embeddings
src/agent.py - Ollama LLM
Emissions knowledge
- Activity dataset
data/dataset.csv - Vector-store builder
src/vector_store.py - Chroma collection
Other components
- User
Connections 7
- User → Streamlit app: submits query
app.py - Streamlit app → Recommendation agent: requests recommendations
app.py - Recommendation agent → Query embeddings: configures encoder
src/agent.py - Recommendation agent → Chroma collection: retrieves context
src/agent.py - Recommendation agent → Ollama LLM: generates answer
src/agent.py - Vector-store builder → Activity dataset: reads rows
src/vector_store.py - Vector-store builder → Chroma collection: upserts documents
src/vector_store.py