How chowdary0507/zepto-ai-ml-capstone works
This repository is a three-part AI/ML capstone, not one connected runtime: a book-scraping and SQLite pipeline, Titanic analysis and modeling scripts, and a policy-question RAG service. The graph shows their documented workflows and the sampled call/data relationships. The analytics modeling excerpt is partial, so uncertain internal steps and integrations are omitted.
Main entry points: pipeline.py Pipeline Runner · 02_modeling.py Model Training · main.py Assistant Graph · 01_eda.py EDA and Cleaning
Components 26
Book Data Pipeline
- Pipeline Runner
data_pipeline/pipeline.py - Book Extraction
data_pipeline/pipeline.py - Price Conversion
data_pipeline/pipeline.py - Relational Load
data_pipeline/pipeline.py - Books SQLite
data_pipeline/zepto_books.db - SQL Queries
data_pipeline/pipeline.py - Join Comparison
data_pipeline/pipeline.py
Titanic Analytics
- EDA and Cleaning
analytics/01_eda.py - Cleaned Dataset
analytics/01_eda.py - Model Training
analytics/02_modeling.py - Survival Classifier
analytics/02_modeling.py - Fare Regression
analytics/02_modeling.py - Model Metrics
analytics/02_modeling.py
Policy Assistant
- Ask API
support_assistant/main.py - Assistant Graph
support_assistant/main.py - Intent Routing
support_assistant/main.py - Policy Documents
support_assistant/docs - Chroma Store
support_assistant/main.py - Query Embeddings
support_assistant/main.py - Policy Retrieval
support_assistant/main.py - Answer Validation
support_assistant/main.py - Answer Response
support_assistant/main.py
Other components
- Operator
- Books To Scrape
- Titanic Dataset
- Customer
Connections 32
- Operator → Pipeline Runner: runs
- Pipeline Runner → Book Extraction: calls
- Book Extraction → Books To Scrape: fetches pages
- Book Extraction → Price Conversion: returns records
- Pipeline Runner → Price Conversion: calls
- Pipeline Runner → Relational Load: calls
- Price Conversion → Relational Load: provides rows
- Relational Load → Books SQLite: writes tables
- Pipeline Runner → SQL Queries: calls
- SQL Queries → Books SQLite: queries
- Pipeline Runner → Join Comparison: runs
- Join Comparison → Books SQLite: reads SQL join
- Operator → EDA and Cleaning: runs
- Titanic Dataset → EDA and Cleaning: loads dataset
- EDA and Cleaning → Cleaned Dataset: writes cleaned data
- Operator → Model Training: runs
- Cleaned Dataset → Model Training: provides features
- Model Training → Survival Classifier: trains classifier
- Model Training → Fare Regression: trains regressor
- Model Training → Model Metrics: writes metrics
- Customer → Ask API: submits question
- Ask API → Assistant Graph: invokes
- Assistant Graph → Intent Routing: dispatches
- Assistant Graph → Policy Documents: loads documents
- Assistant Graph → Chroma Store: bootstraps collection
- Intent Routing → Policy Retrieval: routes policy queries
- Intent Routing → Answer Response: routes general queries
- Policy Retrieval → Query Embeddings: encodes query
- Policy Retrieval → Chroma Store: queries
- Policy Retrieval → Answer Validation: provides chunks
- Answer Validation → Answer Response: builds response
- Ask API → Answer Response: returns