Snowflake

OutdatedAudit Done

other

View Session

Health

Outdated

Benchmark

40/100

SDK Version

No SDK dependency

Audit Status

Audit Done

Started May 11, 2026

Approval

Not requested

Latest Audit Result

Snowflake calls /search and /contents directly using current endpoint URLs and does not use deprecated /research or deprecated request/response fields (startCrawlDate, endCrawlDate, resolvedSearchType, highlightScores). The declared search and text capabilities still map to valid API request fields, but the Cortex schema/docs are outdated for category formatting and search type coverage, and several current API options are not exposed. The prior pdf/github category gap should not be carried forward because the current docs/OpenAPI sources checked list categories as company, research paper, news, personal site, financial report, and people.

Health: outdated

Missing Features (2)

Category enum format mismatch in Cortex Agent tool_spec: uses underscored names (research_paper, personal_site, financial_report) but the Exa API expects space-separated names (research paper, personal site, financial report). Cortex Agent calls with these categories will fail or be silently ignored. (docs/integrations/snowflake.mdx lines 691 and 718-719)
Search type enum incomplete in Cortex tool_spec and docs parameter table: declares [auto, neural, fast, deep] but current API docs/spec include newer search types including instant and deep-reasoning; local generated docs/OpenAPI also include deep-lite. (docs/integrations/snowflake.mdx lines 191 and 715)

Marked outdated since May 11, 2026

Missing Features (7)

Category enum format mismatch in Cortex Agent tool_spec: uses underscored names (research_paper, personal_site, financial_report) but the Exa API expects space-separated names (research paper, personal site, financial report). Cortex Agent calls with these categories will fail or be silently ignored. (docs/integrations/snowflake.mdx lines 691 and 718-719)
Search type enum incomplete in Cortex tool_spec and docs parameter table: declares [auto, neural, fast, deep] but current API docs/spec include newer search types including instant and deep-reasoning; local generated docs/OpenAPI also include deep-lite. (docs/integrations/snowflake.mdx lines 191 and 715)
text.maxCharacters hardcoded with no user-facing parameter: EXA_SEARCH hardcodes 3000 and EXA_GET_CONTENTS hardcodes 5000, so users cannot control text length. (docs/integrations/snowflake.mdx lines 138 and 233)
highlights passed as boolean true only: no advanced query or maxCharacters options exposed, though the API supports object form.
summary passed as boolean true only: no query or schema options exposed, though the API supports customized and structured summaries.
No content freshness control: maxAgeHours and livecrawlTimeout are not exposed for /search or /contents. Deprecated livecrawl is correctly not used.
excludeText filter not available: EXA_SEARCH exposes includeText via include_text but has no exclude_text parameter.