Development

Commands

make install              # install uv + dependencies
make test                 # unit + integration + contract tests (full suite)
make test-fast            # unit tests only (< 3s)
make test-v               # full suite verbose
make test-all             # everything including golden evaluators
make lint                 # lint with ruff
make format               # format with ruff
make run                  # run the CLI (ARGS="--flag")
make run-dry              # TUI with fake delays, no LLM calls
make eval                 # golden dataset evaluators
make contract             # contract tests (recorded API responses)
make smoke-test           # live API smoke tests (requires credentials)
make snapshot-update      # update syrupy snapshot baselines
make budget-report        # show prompt token counts
make graph                # generate agent graph PNG
make build                # build sdist + wheel into dist/
make publish              # publish to PyPI
make clean                # remove build artifacts and caches

Project Structure

src/yeaboi/
├── agent/                      # LangGraph state & graph
│   ├── graph.py                #   Graph compilation & wiring
│   ├── llm.py                  #   LLM provider selection (Anthropic/OpenAI/Google)
│   ├── nodes.py                #   Node functions (intake, analyze, generate, etc.)
│   └── state.py                #   ScrumState, QuestionnaireState, artifact dataclasses
├── prompts/                    # Prompt templates per node
│   ├── analyzer.py             #   Project analyzer prompt
│   ├── feature_generator.py    #   Feature generation prompt
│   ├── intake.py               #   30 questions, smart/standard modes, adaptive templates
│   ├── sprint_planner.py       #   Sprint planning prompt
│   ├── story_writer.py         #   Story writing prompt with few-shot examples
│   ├── system.py               #   Base system prompt
│   └── task_decomposer.py      #   Task decomposition prompt
├── tools/                      # Tool definitions (23 total)
│   ├── azure_devops.py         #   Azure DevOps repo/file/work items
│   ├── calendar_tools.py       #   Bank holiday detection
│   ├── codebase.py             #   Local repo scanning
│   ├── confluence.py           #   Confluence search/read/write
│   ├── notion.py               #   Notion search/read/write (own token)
│   ├── github.py               #   GitHub repo/file/issues/readme
│   ├── jira.py                 #   Jira board/velocity/sprint/epic/story
│   └── llm_tools.py            #   LLM-powered estimation and AC generation
├── ui/                         # Full-screen TUI system
│   ├── mode_select/            #   Mode selection screens
│   ├── provider_select/        #   LLM/tool provider setup
│   ├── session/                #   Main session (phases, editor, pipeline)
│   ├── shared/                 #   Animations, ASCII font, components, input
│   └── splash.py               #   Animated intro
├── repl/                       # Legacy REPL (CLI-flag-driven flows)
│   ├── _intake_menu.py         #   Intake mode selection
│   ├── _io.py                  #   Artifact rendering, file import/export
│   ├── _questionnaire.py       #   Questionnaire UI (one-at-a-time flow)
│   ├── _review.py              #   Review checkpoint UI
│   └── _ui.py                  #   Pipeline progress, streaming, spinner
├── cli.py                      # CLI entry point (argparse, 20 flags)
├── config.py                   # Environment/config management
├── setup_wizard.py             # First-run credential flow
├── sessions.py                 # SQLite session store
├── persistence.py              # State serialization helpers
├── formatters.py               # Rich rendering (dark/light themes)
├── input_guardrails.py         # 4-layer input validation
├── output_guardrails.py        # 4-layer output validation
├── questionnaire_io.py         # Markdown questionnaire import/export
├── html_exporter.py            # Self-contained HTML reports
├── json_exporter.py            # JSON export for CI/CD
├── jira_sync.py                # Batch Jira creation with idempotency
└── __init__.py                 # Version, LangSmith noise suppression

Testing Conventions

• One test file per source module: repl.py → test_repl.py, state.py → test_state.py
• Group related tests in classes: TestGracefulExit, TestStreaming, TestPriority
• Node tests live in tests/unit/nodes/ (split into 9 files)
• Shared node test helpers in tests/_node_helpers.py
• Pytest markers: slow, eval, vcr, smoke

Environment Variables

Git Conventions

• Commit messages: lowercase imperative (e.g., "add streaming output", "fix import sorting")
• Branch naming: feature/<description> for feature work
• PRs: feature branches merge to main via pull request
• Include Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> on AI-assisted commits

Evaluation & Testing

Red Teaming Checklist

• Prompt injection ("Ignore your instructions and...")
• Jailbreaking (roleplay scenarios to bypass safety)
• Messy inputs (typos, slang, code-switching)
• Extremely long or empty project descriptions
• Contradictory requirements
• Adversarial inputs designed to trigger hallucination or bias

Graceful Degradation

Tech Stack

For contributor workflow — worktrees, hooks, CI conventions — see CLAUDE.md in the repo root.