Skip to main content
Glama
determinism.md1.32 kB
# Determinism Rules ## Output Requirements - **No probabilistic scoring without thresholds**: Scores like Lighthouse's 0-1 scale must be converted to explicit pass/fail/unknown with defined thresholds - **Explicit outcomes**: All results must have an explicit `outcome` field with value `'pass'`, `'fail'`, or `'unknown'` - **Unknown requires reason**: When `outcome` is `'unknown'`, a `reason_code` field is mandatory - **No ambiguous fields**: No "overall quality" or "accessibility score" fields without decomposition into specific rule outcomes ## Normalisation Schema All tool outputs must be normalised to the unified schema in `src/types/normalised.ts`: ```typescript interface NormalisedResult { rule_id: string; wcag_ref: string[]; severity: 'critical' | 'serious' | 'moderate' | 'minor'; confidence: 'high' | 'medium' | 'low'; outcome: 'pass' | 'fail' | 'unknown'; selector: string; dom_context: string; message: string; reason_code?: string; // Required when outcome is 'unknown' } ``` ## Thresholds - **Lighthouse scores**: `score >= 0.9` = pass, `score === 0` = fail, otherwise unknown - **Confidence mapping**: High impact = high confidence, moderate = medium, minor = low - **Unknown reason codes**: Must be explicit (e.g., 'INCOMPLETE_CHECK', 'SCORE_AMBIGUOUS', 'WAVE_ALERT')

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/Duds/accessibility-mcp'

If you have feedback or need assistance with the MCP directory API, please join our Discord server