Medical GraphRAG Assistant

# Implementation Plan: Playwright UX Tests **Branch**: `002-playwright-ux-tests` | **Date**: 2025-12-10 | **Spec**: [spec.md](./spec.md) **Input**: Feature specification from `/specs/002-playwright-ux-tests/spec.md` ## Summary Implement automated UX tests for the Medical GraphRAG Assistant Streamlit application using the Playwright MCP server integrated with Claude Code. Tests will run as a single conversation session, executing sequentially with fail-fast behavior. Coverage includes page load verification, example button functionality, chat input/response, sidebar tool listing, and clear chat functionality. ## Technical Context **Language/Version**: Markdown (Claude Code prompts), Python 3.11 (target application) **Primary Dependencies**: @playwright/mcp (MCP server), Claude Code (execution host) **Storage**: N/A (stateless tests) **Testing**: Playwright MCP tool calls via Claude Code conversation **Target Platform**: Chromium browser via Playwright, targeting http://54.209.84.148:8501 **Project Type**: Single project (test suite only) **Performance Goals**: All P1 tests complete within 2 minutes total (SC-001) **Constraints**: 10s page load timeout, 30s AI response timeout, fail-fast on first failure **Scale/Scope**: 5 user stories, ~12 acceptance scenarios, 10 functional requirements ## Constitution Check *GATE: Must pass before Phase 0 research. Re-check after Phase 1 design.* | Principle | Compliance | Notes | |-----------|------------|-------| | I. Authorship & Attribution | PASS | Test artifacts authored by Thomas Dyar | | II. MCP-First Architecture | PASS | Using Playwright MCP server for all browser automation | | III. Vector Database Purity | N/A | Tests do not interact directly with database | | IV. Medical Data Integrity | PASS | Tests verify UI behavior, do not modify data | | V. Graceful Degradation | PASS | Tests use fail-fast per spec, timeouts configured | **Gate Status**: PASSED - No violations. Proceed to Phase 0. ## Project Structure ### Documentation (this feature) ```text specs/002-playwright-ux-tests/ ├── plan.md # This file ├── spec.md # Feature specification ├── research.md # Phase 0 output (Playwright MCP capabilities) ├── data-model.md # Phase 1 output (test case model) ├── quickstart.md # Phase 1 output (how to run tests) ├── contracts/ # Phase 1 output (test protocol) │ └── test-protocol.md # Test execution contract ├── checklists/ │ └── requirements.md # Quality checklist └── tasks.md # Phase 2 output (/speckit.tasks command) ``` ### Source Code (repository root) ```text tests/ └── ux/ └── playwright-mcp/ └── test-prompts.md # Claude Code prompts for each test ``` **Structure Decision**: Tests are executed via Claude Code conversation, not traditional test files. Test prompts define the verification steps. No compiled code required. ## Complexity Tracking No constitution violations requiring justification.