docs-mcp-server

# Tasks: Search Quality Evaluation 1. [ ] **Install Dependencies** - Add `promptfoo` to `devDependencies`. - Add `evaluate:search` script to `package.json`. - Validation: `npm run evaluate:search` runs (even if empty). 2. [ ] **Implement Search Provider Adapter** - Create `src/tools/search-provider.ts`. - Implement the logic to connect to `DocumentStore` and run `DocumentRetrieverService.search`. - Validation: Unit test the provider to ensure it returns strings from the DB. 3. [ ] **Configure Promptfoo** - Create `promptfoo.yaml`. - Define the metrics (Relevance, Compactness, Integrity, Ranking) using assertions. - Validation: Run `promptfoo eval` with a dummy prompt and verify metrics appear. 4. [ ] **Create Evaluation Dataset** - Create `tests/search-eval/dataset.yaml`. - Add at least 5 "Golden Pairs" for React documentation (Query + Expected URL + Key Concepts). - Validation: The dataset is parsable by promptfoo. 5. [ ] **Wire Up Evaluation Runner** - Configure the `evaluate:search` npm script to run `promptfoo eval` using `promptfoo.yaml` and the dataset in `tests/search-eval/`. - Ensure the flow checks for the existence of the React index before running the evaluation. - Validation: `npm run evaluate:search` runs the full evaluation using `tests/search-eval/dataset.yaml`.