feat(exa): add highlights, max_age_hours, instant search type, and new categories

[theishangoswami]

adds support for:

• highlights param on both exa_search and exa_get_contents — token-efficient page excerpts (bool or dict with maxCharacters/query)
• max_age_hours param — content freshness control (0=always livecrawl, -1=cache only, N=livecrawl if older than N hours). old livecrawl enum kept for backward compat
• instant search type — lowest latency option alongside auto/fast/deep
• research paper and people categories

also:

• renders highlights in both format_search_response and format_contents_response
• adds highlights to test fixtures and 2 new highlight rendering tests
• all pre-commit hooks pass (format, lint, type-lint, tests)

[mvangara10]

@mkmeral could you please review the PR

[mkmeral]

/strands review

[agent-of-mkmeral]

Assessment: ⚠️ Request Changes

Clean feature addition that extends the exa tool with useful Exa API capabilities. The code follows existing patterns well. Requesting changes for missing documentation PR section and a few issues noted inline.

Review Details

• Documentation: No "Documentation PR" section in the PR description. Since this adds new public API parameters (highlights, max_age_hours, instant search type, new categories), either a docs PR link or a justification for why docs aren't needed is required.
• Testing: Highlight rendering tests are good additions ✅. However, there are no tests verifying the new parameters (highlights, max_age_hours) are correctly passed through to the API payload in either exa_search or exa_get_contents. Payload construction tests would strengthen confidence.
• Commit format: The second commit (add x-exa-integration header for usage tracking) doesn't follow conventional commit format. Should be prefixed with a type like feat(exa): or chore(exa):.
• Tracking header: The x-exa-integration: aws-strands-agent header is a partner integration tracking header. This is fine and common practice, but worth calling out for visibility since it's a behavioral change in an open-source tool.

Good work overall — the code is clean and follows existing patterns.

---

🤖 AI agent response. Strands Agents. Feedback welcome!