kaniska-praba007/hackerrank-orchestrate-june-edition-submission

How kaniska-praba007/hackerrank-orchestrate-june-edition-submission works

This repository is a multimodal insurance and damage-claim evidence review system for cars, laptops, and packages. Its main workflow loads claim, history, and evidence-requirement CSVs plus claim images; parses claim intent; analyzes images; checks evidence sufficiency and risk; verifies a decision; and writes a schema-shaped CSV result. A separate evaluation command reruns the pipeline against labelled claims and reports metrics. The README documents optional caching; its runtime wiring is not visible in the sampled source, so that relationship is marked unverified.

Main entry points: orchestrator.py Graph Orchestrator · claim_parser.py Claim Parser · visual_analyzer.py Visual Analyzer · schemas.py Typed Schemas

Components 23

CLI and Evaluation

CLI Commands code/main.py
Evaluation Runner code/evaluation/main.py

Claim Pipeline

Graph Orchestrator code/pipeline/orchestrator.py
CSV and Image Loading code/pipeline/ingestion.py
Pipeline State code/pipeline/state.py
Output Writer code/pipeline/output_writer.py

Review Agents

Claim Parser code/agents/claim_parser.py
Mock Claim Parser code/agents/mock_claim_parser.py
Visual Analyzer code/agents/visual_analyzer.py
Mock Visual Analyzer code/agents/mock_visual_analyzer.py
Evidence Engine code/agents/evidence_engine.py
Part Matcher code/agents/part_matcher.py
Risk Detector code/agents/risk_detector.py
Mock Risk code/agents/mock_risk.py
Claim Verifier code/agents/verifier.py

Models and Services

Typed Schemas code/models/schemas.py
LLM Retry code/services/retry_llm.py
Optional Cache code/services/cache.py

Other components

Operator
Claim CSVs
Gemini API
Prediction CSV
Evaluation Report

Connections 31

Operator → CLI Commands: runs code/main.py
Claim CSVs → CLI Commands: provides paths code/main.py
CLI Commands → Graph Orchestrator: runs pipeline code/main.py
CLI Commands → Evaluation Runner: runs evaluation code/main.py
Evaluation Runner → Graph Orchestrator: runs pipeline code/evaluation/main.py
Graph Orchestrator → CSV and Image Loading: loads records code/pipeline/orchestrator.py
CSV and Image Loading → Typed Schemas: creates records code/pipeline/ingestion.py
Graph Orchestrator → Pipeline State: uses state code/pipeline/orchestrator.py
Graph Orchestrator → Claim Parser: dispatches parsing code/pipeline/orchestrator.py
Graph Orchestrator → Mock Claim Parser: dispatches mock code/pipeline/orchestrator.py
Graph Orchestrator → Visual Analyzer: dispatches analysis code/pipeline/orchestrator.py
Graph Orchestrator → Mock Visual Analyzer: dispatches mock code/pipeline/orchestrator.py
Graph Orchestrator → Evidence Engine: checks sufficiency code/pipeline/orchestrator.py
Evidence Engine → Part Matcher: matches parts code/agents/evidence_engine.py
Graph Orchestrator → Risk Detector: detects risk code/pipeline/orchestrator.py
Graph Orchestrator → Mock Risk: selects mock code/pipeline/orchestrator.py
Graph Orchestrator → Claim Verifier: decides claim code/pipeline/orchestrator.py
Graph Orchestrator → Output Writer: writes output code/pipeline/orchestrator.py
Output Writer → Prediction CSV: writes CSV code/pipeline/output_writer.py
Evaluation Runner → Prediction CSV: reads predictions code/evaluation/main.py
Evaluation Runner → Evaluation Report: writes report code/evaluation/main.py
Claim Parser → Typed Schemas: returns intent code/agents/claim_parser.py
Visual Analyzer → Typed Schemas: returns findings code/agents/visual_analyzer.py
Risk Detector → Typed Schemas: uses risk types code/agents/risk_detector.py
Claim Verifier → Typed Schemas: returns decision code/agents/verifier.py
Claim Parser → LLM Retry: retries calls code/agents/claim_parser.py
Visual Analyzer → LLM Retry: retries calls code/agents/visual_analyzer.py
Claim Parser → Gemini API: calls model code/agents/claim_parser.py
Visual Analyzer → Gemini API: calls vision code/agents/visual_analyzer.py
Claim Parser → Optional Cache: may cache calls
Visual Analyzer → Optional Cache: may cache calls

Drawn by GitDiagram on November 10, 2026 from the repository's file tree, README and source files. Read as Markdown ·