src/metadata
Build-generated Databricks model retirement, capability, rate-limit, and reasoning-effort metadata, plus a daily write-through cache for refreshed docs and error-learned reasoning ladders.
Committed snapshots seed every process. refreshModelMetadata pulls documentation through the shared memory-first createMetadataCache and rebuilds the in-memory index. Error-learned efforts use the same store via rememberReasoningLevels. Known-good values are written behind to the shared model project cache.
Functions
Section titled “Functions”- hydrateModelMetadata
- learnedReasoningCatalogue
- learnReasoningLevelsFromError
- modelCapabilitiesFor
- modelMetadataFor
- modelMetadataGeneratedAt
- modelRateLimitsFor
- modelReasoningLevelsFor
- modelStatusFor
- reasoningModelCatalogue
- refreshModelMetadata
- rememberReasoningLevels
- retiredModelNames
References
Section titled “References”MODEL_METADATA_TTL_MS
Section titled “MODEL_METADATA_TTL_MS”Re-exports MODEL_METADATA_TTL_MS
MODEL_RATE_LIMITS_URL
Section titled “MODEL_RATE_LIMITS_URL”Re-exports MODEL_RATE_LIMITS_URL
OPENAI_RESPONSES_MODELS_URL
Section titled “OPENAI_RESPONSES_MODELS_URL”Re-exports OPENAI_RESPONSES_MODELS_URL
QUERY_REASON_MODELS_URL
Section titled “QUERY_REASON_MODELS_URL”Re-exports QUERY_REASON_MODELS_URL
RETIRED_MODELS_URL
Section titled “RETIRED_MODELS_URL”Re-exports RETIRED_MODELS_URL
WEB_SEARCH_MODELS_URL
Section titled “WEB_SEARCH_MODELS_URL”Re-exports WEB_SEARCH_MODELS_URL