← Back to Integrations

Google Sheets

OutdatedAudit Done
sheetsexa-labs/exa-for-sheets
View Session

Health

Outdated

Benchmark

64/100

SDK Version

No SDK dependency

Audit Status

Audit Done

Started May 11, 2026

Approval

Not requested

Latest Audit Result

Audit completed on exa-labs/exa-for-sheets at origin/master (v2.0.5). The declared capabilities are implemented via direct REST calls: /search supports auto/neural/keyword plus optional highlights; /contents fetches URL contents; /findSimilar returns related URLs; /answer returns grounded answers; advanced answer also uses /chat/completions. Main issues are API shape drift: /findSimilar is considered deprecated by current Exa API guidance, EXA_SEARCH still exposes keyword and sends legacy useAutoprompt, EXA_CONTENTS omits explicit text:true even though docs specify it, and README/docs lag implemented parameters and current search types. Report uploaded to https://kitchen-sink.exa.ai/content-host/view/google-sheets-audit-03cb4a2d.

Health: outdated

Missing Features (4)

  • DEPRECATED: EXA_FINDSIMILAR still calls /findSimilar; current Exa API guidance says to treat /findSimilar as deprecated and prefer /search, optionally after /contents on the seed URL.
  • EXA_SEARCH exposes type "keyword" and sends useAutoprompt, but current /search docs list search types as neural, fast, auto, deep-lite, deep, deep-reasoning, and instant; useAutoprompt is not documented in the current request schema.
  • EXA_CONTENTS calls /contents with only {urls:[url]} and relies on default text output; current /contents docs/examples specify passing top-level text:true for text retrieval.
  • EXA_SEARCH allows category values "github" and "pdf" that are not in the current documented category set (company, research paper, news, personal site, financial report, people), risking silent no-op behavior.

Marked outdated since May 11, 2026

Missing Features (9)

  • DEPRECATED: EXA_FINDSIMILAR still calls /findSimilar; current Exa API guidance says to treat /findSimilar as deprecated and prefer /search, optionally after /contents on the seed URL.
  • EXA_SEARCH exposes type "keyword" and sends useAutoprompt, but current /search docs list search types as neural, fast, auto, deep-lite, deep, deep-reasoning, and instant; useAutoprompt is not documented in the current request schema.
  • EXA_CONTENTS calls /contents with only {urls:[url]} and relies on default text output; current /contents docs/examples specify passing top-level text:true for text retrieval.
  • EXA_SEARCH allows category values "github" and "pdf" that are not in the current documented category set (company, research paper, news, personal site, financial report, people), risking silent no-op behavior.
  • Update EXA_SEARCH type validation and documentation to current search modes; preserve auto as default, remove/replace keyword, and consider adding fast/instant/deep-lite/deep/deep-reasoning if desired.
  • Remove useAutoprompt from /search payload construction because it is legacy/undocumented in current API docs.
  • Update EXA_CONTENTS to send top-level text:true explicitly and optionally expose highlights as a documented top-level contents option.
  • Update README and sidebar docs for implemented-but-undocumented parameters: EXA_ANSWER type, EXA_SEARCH highlightsMaxChars, outputSchemaJson, and current category/type values.
  • If keeping EXA_FINDSIMILAR temporarily, document it as legacy/deprecated and plan migration to /search-based related-page discovery.