LlamaIndex

Latest Audit Result

LlamaIndex ExaToolSpec v0.5.1 is outdated. It depends on exa-py 1.x (resolves 1.16.1) while latest is 2.12.1. The integration uses deprecated SDK methods (search_and_contents, use_autoprompt, autoprompt_string) that are removed in 2.x. It exposes search, get_contents, find_similar, text extraction, and highlights but lacks answer, streaming, category filters, include/exclude text filters, summary, freshness controls, subpages, and deep-search synthesis features available in the current API. The /research endpoint is deprecated (sunset May 1 2026) so its absence is correct. The context content option is also deprecated in SDK 2.x so its absence is correct.

Missing Features (9)

• exa-py pinned to 1.x (>=1.0.8,<2); latest is 2.12.1 — blocks access to new SDK features and modern API surface
• Uses deprecated search_and_contents() SDK method (replaced by search() with contents param in 2.x)
• Uses deprecated use_autoprompt parameter (removed in exa-py 2.x search() signature)
• Accesses response.autoprompt_string which is removed in exa-py 2.x SearchResponse
• search_and_retrieve_highlights passes only highlights=True — should expose configurable max_characters and query for highlights
• search_and_retrieve_documents hardcodes max_characters from constructor — should allow per-call text max_characters override
• retrieve_documents (get_contents) retrieves only .text with no content options — should support text max_characters, highlights, or summary
• find_similar returns only metadata (title, url, id) with no content retrieval — should support retrieving text/highlights for similar results
• Text content include_html_tags option not exposed despite being supported by the SDK