How rathna070-ai/ado-llm-defect-analyzer works
ADO Defect LLM Analysis imports defects from Azure DevOps or Excel/CSV, classifies them with a configurable LLM, persists results, and produces summaries and exports for BI or a Streamlit dashboard. The graph follows documented stages and caller-owned relationships shown in excerpts, the SOURCE INDEX, or README. Some dashboard, report, provider, and storage details are unsampled; their links are limited to those references and documented capabilities.
Main entry points: cli.py CLI · fetch.py Fetch stage · categorize.py Categorization stage · ado_client.py Azure DevOps client
Components 30
User entry points
- CLI
src/ado_defect_analysis/cli.py - Streamlit app
src/ado_defect_analysis/dashboard/streamlit_app.py - Source and run view
src/ado_defect_analysis/dashboard/views/home.py - Leadership results
src/ado_defect_analysis/dashboard/views/results.py - Background run manager
src/ado_defect_analysis/background.py
Defect ingestion
- Saved-query URL parser
src/ado_defect_analysis/ado_query.py - Azure DevOps client
src/ado_defect_analysis/ado_client.py - Excel/CSV parser
src/ado_defect_analysis/excel_source.py - Fetch stage
src/ado_defect_analysis/pipeline/fetch.py
Analysis pipeline
- Defect data models
src/ado_defect_analysis/models.py - Categorization stage
src/ado_defect_analysis/pipeline/categorize.py - Aggregation and review metrics
src/ado_defect_analysis/pipeline/aggregate.py - CAPA metrics
src/ado_defect_analysis/capa.py - Narrative report
src/ado_defect_analysis/pipeline/report.py - BI export
src/ado_defect_analysis/pipeline/export.py
LLM and configuration
- Provider selection
src/ado_defect_analysis/llm/factory.py - LLM interface and usage
src/ado_defect_analysis/llm/base.py - OpenAI-compatible transport
src/ado_defect_analysis/llm/openai_compatible.py - Azure provider
src/ado_defect_analysis/llm/azure_provider.py - Groq provider
src/ado_defect_analysis/llm/groq_provider.py - Configuration
src/ado_defect_analysis/config.py - Credential store
src/ado_defect_analysis/secrets.py - Retrying HTTP session
src/ado_defect_analysis/http.py
Persistence and results
- Defect store
src/ado_defect_analysis/storage.py
Other components
- Operator
- Azure DevOps
- ADO export file
- LLM service
- SQLite database
- CSV/XLSX and review outputs
Connections 46
- Operator → CLI: runs commands
README.md - Operator → Streamlit app: uses dashboard
README.md - CLI → Fetch stage: dispatches fetch
src/ado_defect_analysis/cli.py - CLI → Categorization stage: dispatches categorize
src/ado_defect_analysis/cli.py - CLI → Narrative report: dispatches report
src/ado_defect_analysis/cli.py - CLI → BI export: dispatches export
src/ado_defect_analysis/cli.py - CLI → Configuration: loads settings
src/ado_defect_analysis/cli.py - Streamlit app → Source and run view: renders
src/ado_defect_analysis/dashboard/streamlit_app.py - Streamlit app → Leadership results: renders
src/ado_defect_analysis/dashboard/streamlit_app.py - Source and run view → Fetch stage: starts source fetch
src/ado_defect_analysis/dashboard/views/home.py - Source and run view → Background run manager: starts run
src/ado_defect_analysis/dashboard/views/home.py - Background run manager → Categorization stage: runs analysis
src/ado_defect_analysis/background.py - Source and run view → Configuration: uses settings
src/ado_defect_analysis/dashboard/views/home.py - Saved-query URL parser → Azure DevOps client: resolves query reference
src/ado_defect_analysis/ado_client.py - Fetch stage → Saved-query URL parser: parses query URL
src/ado_defect_analysis/pipeline/fetch.py - Fetch stage → Azure DevOps client: fetches work items
src/ado_defect_analysis/pipeline/fetch.py - Fetch stage → Excel/CSV parser: parses export
src/ado_defect_analysis/pipeline/fetch.py - Fetch stage → Defect store: upserts defects
src/ado_defect_analysis/pipeline/fetch.py - Azure DevOps client → Azure DevOps: queries API
src/ado_defect_analysis/ado_client.py - Excel/CSV parser → ADO export file: reads file
- Azure DevOps client → Defect data models: constructs defects
src/ado_defect_analysis/ado_client.py - Excel/CSV parser → Defect data models: produces defects
src/ado_defect_analysis/excel_source.py - Categorization stage → Defect store: loads and saves
src/ado_defect_analysis/pipeline/categorize.py - Categorization stage → Provider selection: gets provider
src/ado_defect_analysis/pipeline/categorize.py - Categorization stage → LLM interface and usage: uses interface
src/ado_defect_analysis/pipeline/categorize.py - Categorization stage → Defect data models: uses defect shapes
src/ado_defect_analysis/pipeline/categorize.py - Aggregation and review metrics → Defect store: reads categorized defects
src/ado_defect_analysis/pipeline/aggregate.py - Aggregation and review metrics → CAPA metrics: uses metrics
src/ado_defect_analysis/pipeline/aggregate.py - Narrative report → Aggregation and review metrics: builds summary input
src/ado_defect_analysis/pipeline/report.py - Narrative report → Provider selection: gets provider
src/ado_defect_analysis/pipeline/report.py - BI export → Aggregation and review metrics: uses aggregates
src/ado_defect_analysis/pipeline/export.py - BI export → CSV/XLSX and review outputs: writes reports
README.md - Leadership results → Aggregation and review metrics: renders metrics
src/ado_defect_analysis/dashboard/views/results.py - Leadership results → BI export: offers export
src/ado_defect_analysis/dashboard/views/results.py - Provider selection → Configuration: selects backend
src/ado_defect_analysis/llm/factory.py - Provider selection → Azure provider: constructs provider
src/ado_defect_analysis/llm/factory.py - Provider selection → Groq provider: constructs provider
src/ado_defect_analysis/llm/factory.py - Azure provider → OpenAI-compatible transport: extends transport
src/ado_defect_analysis/llm/azure_provider.py - Groq provider → OpenAI-compatible transport: uses transport
src/ado_defect_analysis/llm/groq_provider.py - OpenAI-compatible transport → LLM interface and usage: implements interface
src/ado_defect_analysis/llm/openai_compatible.py - OpenAI-compatible transport → Retrying HTTP session: builds session
src/ado_defect_analysis/llm/openai_compatible.py - OpenAI-compatible transport → LLM service: sends completions
src/ado_defect_analysis/llm/openai_compatible.py - Azure DevOps client → Retrying HTTP session: builds session
src/ado_defect_analysis/ado_client.py - Configuration → Credential store: resolves credentials
src/ado_defect_analysis/config.py - Defect store → SQLite database: reads and writes
src/ado_defect_analysis/storage.py - Background run manager → Configuration: uses settings
src/ado_defect_analysis/background.py