feat: add Tavily search agent to Heurist Mesh

[tavily-integrations]

Summary

• Added a new TavilySearchAgent mesh agent (mesh/agents/tavily_search_agent.py) that provides web search and URL content extraction via Tavily's API
• Exposes two tools: tavily_web_search (advanced search with domain/topic/time filtering) and tavily_extract_url (content extraction from up to 20 URLs)
• Uses AsyncTavilyClient from tavily-python SDK with @with_retry and @with_cache decorators
• Follows existing agent patterns (mirrors ExaSearchAgent structure)

Files changed

• mesh/agents/tavily_search_agent.py — New agent implementation
• mesh/tests/tavily_search_agent_example.yaml — Test YAML with example inputs/outputs
• mesh/pyproject.toml — Added tavily-python>=0.5.0 dependency

Dependency changes

• Added tavily-python>=0.5.0 to mesh/pyproject.toml

Environment variable changes

• Requires TAVILY_API_KEY environment variable

Notes for reviewers

• The agent uses search_depth="advanced" by default for highest relevance (2 credits per search)
• An agent image at mesh/images/Tavily.png is referenced in metadata but not included — needs to be added separately
• No existing agents were modified; this is a purely additive change

🤖 Generated with Claude Code

Automated Review

• Passed after 1 attempt(s)
• Final review: The TavilySearchAgent implementation is correct, complete, and consistent with the existing ExaSearchAgent pattern. It introduces a new mesh agent with two tools (tavily_web_search and tavily_extract_url), uses the proper AsyncTavilyClient SDK, adds the dependency to mesh/pyproject.toml, and includes a test YAML. No regressions or unintended changes. Only minor style and design issues found.