How kutralakumarank/ai-ecommerce-review-summarizer works
AIReviewSummarizer is a Streamlit product that turns a selected product’s reviews into overall, positive-only, and negative-only summaries. Its main workflow prepares sentiment-split review data, builds or loads three retrieval indexes, then invokes a LangGraph agent separately for each question; the agent may retrieve from multiple tools, grade retrieved context, and generate or rewrite before results are rendered. The README documents key workflow links that are not all visible in the sampled source. The graph preserves those documented capabilities without treating unsampled implementation details as inspected.
Main entry points: ai_review_summarizer_nodes.py Workflow assembly · Rag.py RAG indexing · main.py App coordinator · etlpipeline.py ETL pipeline
Components 25
Streamlit interface
- Streamlit entry
app.py - App coordinator
src/main.py - Input controls
src/ui/streamlitui/loadui.py - UI configuration
src/ui/uiconfigfile.py - Summary display
src/ui/streamlitui/display_result.py
Review preparation
- Raw review CSV
src/ETL/ecommerce_product_reviews_dataset.csv - ETL pipeline
src/ETL/etlpipeline.py - Clean reviews
src/ETL/transformed_product_reviews_dataset.csv - Positive reviews
src/ETL/positive_reviews.csv - Negative reviews
src/ETL/negative_reviews.csv
Retrieval layer
- RAG indexing
src/RAG/Rag.py - All-review index
src/RAG/vectorstore - Positive index
src/RAG/vectorstore_positive - Negative index
src/RAG/vectorstore_negative - Retriever tools
src/ui/streamlitui/loadui.py
Agent workflow
- Graph builder
src/graph/graph.py - Workflow assembly
src/nodes/ai_review_summarizer_nodes.py - Message state
src/states/state.py - Agent decision
src/RAG/Rag.py - Tool retrieval
src/nodes/ai_review_summarizer_nodes.py - Relevance grading
src/RAG/Rag.py - Question rewrite
src/RAG/Rag.py - Answer generation
src/RAG/Rag.py
Model configuration
- Groq model setup
src/LLMS/groqllm.py
Other components
- Product requester
Connections 39
- Product requester → Streamlit entry: starts app
README.md - Streamlit entry → App coordinator: calls
app.py - App coordinator → Input controls: loads inputs
src/main.py - Input controls → UI configuration: reads options
src/ui/streamlitui/loadui.py - Input controls → ETL pipeline: runs products ETL
src/ui/streamlitui/loadui.py - Raw review CSV → ETL pipeline: extracts
src/ETL/etlpipeline.py - ETL pipeline → Clean reviews: writes cleaned
src/ETL/etlpipeline.py - ETL pipeline → Positive reviews: writes positive
src/ETL/etlpipeline.py - ETL pipeline → Negative reviews: writes negative
src/ETL/etlpipeline.py - Input controls → RAG indexing: builds indexes
src/ui/streamlitui/loadui.py - Clean reviews → RAG indexing: loads reviews
README.md - Positive reviews → RAG indexing: loads subset
README.md - Negative reviews → RAG indexing: loads subset
README.md - RAG indexing → All-review index: persists index
README.md - RAG indexing → Positive index: persists index
README.md - RAG indexing → Negative index: persists index
README.md - All-review index → Retriever tools: loads retriever
src/ui/streamlitui/loadui.py - Positive index → Retriever tools: loads retriever
src/ui/streamlitui/loadui.py - Negative index → Retriever tools: loads retriever
src/ui/streamlitui/loadui.py - App coordinator → Groq model setup: configures model
src/main.py - App coordinator → Graph builder: builds workflow
src/main.py - App coordinator → Retriever tools: gets tools
src/main.py - Graph builder → Workflow assembly: delegates build
src/graph/graph.py - Workflow assembly → Message state: uses state schema
src/nodes/ai_review_summarizer_nodes.py - Workflow assembly → Agent decision: registers agent
src/nodes/ai_review_summarizer_nodes.py - Workflow assembly → Tool retrieval: registers tool node
src/nodes/ai_review_summarizer_nodes.py - Workflow assembly → Relevance grading: routes for grading
src/nodes/ai_review_summarizer_nodes.py - Workflow assembly → Question rewrite: registers rewrite
src/nodes/ai_review_summarizer_nodes.py - Workflow assembly → Answer generation: registers generation
src/nodes/ai_review_summarizer_nodes.py - Retriever tools → Agent decision: binds tools
README.md - Agent decision → Tool retrieval: dispatches calls
README.md - Tool retrieval → Relevance grading: returns tool results
README.md - Relevance grading → Answer generation: accepts relevant
README.md - Relevance grading → Question rewrite: rejects context
README.md - Question rewrite → Agent decision: retries question
README.md - Groq model setup → Agent decision: supplies model
README.md - App coordinator → Summary display: renders results
src/main.py - Summary display → Agent decision: invokes graph
README.md - Answer generation → Summary display: returns summaries
README.md