How aksharma-15/flipkart-order-intelligence-and-support-assistant works
This capstone combines order-return risk modeling, product-image classification, and a conversational support assistant. The documented analytics workflow generates an orders dataset, trains and evaluates a return-risk model, and saves it for use by the Part 3 agent. Part 2 trains and saves an image classifier, which the README identifies as loadable for Part 3 predictions. Transcript filenames indicate policy support, multi-turn state, and grounding concerns. Part 3 source was not sampled, so its internal dispatch and policy-retrieval wiring cannot be established here.
Main entry points: part3_agent.ipynb Support Agent · orders_dataset.csv Orders Dataset · part1_return_risk.ipynb Risk Pipeline · part2_vision_model.ipynb Vision Pipeline
Components 10
Order Data
- Order Generator
src/generate_orders.py - Orders Dataset
data/orders_dataset.csv
Risk Analytics
- Risk Pipeline
part1_return_risk.ipynb - Risk Model
models/return_risk_model.pkl
Image Classification
- Vision Pipeline
part2_vision_model.ipynb - Sample Images
data/sample_images - Product Classifier
models/product_classifier.pt
Support Assistant
- Support Agent
part3_agent.ipynb - Support Reply
part3_agent.ipynb
Other components
- Customer
Connections 9
- Order Generator → Orders Dataset: writes dataset
- Orders Dataset → Risk Pipeline: training data
- Risk Pipeline → Risk Model: saves model
- Sample Images → Vision Pipeline: training images
- Vision Pipeline → Product Classifier: saves model
- Customer → Support Agent: asks
- Support Agent → Risk Model: uses risk model
- Support Agent → Product Classifier: uses classifier
- Support Agent → Support Reply: returns reply