How harshavr8/release-readiness-copilot works
Release-Readiness Copilot evaluates seeded commit and test-run data, classifies failures, optionally reconsiders uncertain results using later-commit context, and produces a scored go/no-go report. The graph definition owns workflow routing; its node implementations call the MCP client, classifier, and report logic. The MCP server reads the seeded JSON inputs, while the entry point saves report files. The README documents these capabilities; connections not shown in the excerpts are omitted.
Main entry points: nodes.py Classify Failures · run.py Run Entry Point · pipeline.py LangGraph Workflow · server.py MCP Tools Server
Components 18
Workflow
- Run Entry Point
run.py - LangGraph Workflow
graph/pipeline.py - Shared Graph State
graph/state.py - Fetch Commits
graph/nodes.py - Fetch Test Runs
graph/nodes.py - Classify Failures
graph/nodes.py - Gather More Context
graph/nodes.py - Reclassify Failures
graph/nodes.py - Build Report Node
graph/nodes.py
Evidence Inputs
- Seeded Commits
data/commits.json - Seeded Test Runs
data/test_runs.json
Analysis and Reporting
- Failure Classifier
graph/classifier.py - Scoring and Rendering
graph/report.py - Saved Reports
reports/.gitkeep
MCP Data Access
- MCP Client
graph/mcp_client.py - MCP Tools Server
mcp_server/server.py
Other components
- Release Engineer
- Claude API
Connections 21
- Release Engineer → Run Entry Point: runs
README.md - Run Entry Point → LangGraph Workflow: invokes
run.py - LangGraph Workflow → Shared Graph State: uses state
graph/pipeline.py - LangGraph Workflow → Fetch Commits: starts with
graph/pipeline.py - Fetch Commits → Fetch Test Runs: then runs
graph/pipeline.py - Fetch Test Runs → Classify Failures: then runs
graph/pipeline.py - Classify Failures → Gather More Context: routes if uncertain
graph/pipeline.py - Classify Failures → Build Report Node: routes if resolved
graph/pipeline.py - Gather More Context → Reclassify Failures: then runs
graph/pipeline.py - Reclassify Failures → Gather More Context: loops if uncertain
graph/pipeline.py - Reclassify Failures → Build Report Node: routes if resolved
graph/pipeline.py - Fetch Commits → MCP Client: requests commits
graph/nodes.py - Fetch Test Runs → MCP Client: requests test runs
graph/nodes.py - MCP Client → MCP Tools Server: calls tools
graph/mcp_client.py - MCP Tools Server → Seeded Commits: reads
mcp_server/server.py - MCP Tools Server → Seeded Test Runs: reads
mcp_server/server.py - Classify Failures → Failure Classifier: classifies with
graph/nodes.py - Reclassify Failures → Failure Classifier: reclassifies with
graph/nodes.py - Failure Classifier → Claude API: classifies via
README.md - Build Report Node → Scoring and Rendering: scores and renders
graph/nodes.py - Run Entry Point → Saved Reports: writes
run.py