Refactor Agents

[jdchawla29]

Note

Medium Risk
Large refactor of core agent execution, tool-call plumbing, and gateway model resolution; regressions could break tool execution, citations, or provider-specific message formatting. Mostly internal API reshaping, but touches critical runtime paths for Claude/Gemini agents.

Overview
Refactors the core agent architecture by replacing the old EvalContext-bound MCPAgent with a run-scoped AgentContext + generic provider message typing, and moving tool execution/formatting into shared AgentTools abstractions used by Claude and Gemini.

Updates Claude and Gemini agents to the new loop: provider-specific format_messages(), unified AgentResponse return type, centralized tool preparation/execution, and revised citations handling (including tool-result citation documents for Claude and grounding citations for Gemini).

Moves/modernizes gateway routing: create_agent() now lives in hud/agents/gateway.py and can resolve models via HUD /models/ metadata, selecting the right agent type/provider client automatically. Documentation is updated to reflect terminology changes ("ResponseAgent" → response automation, InferenceResult → AgentResponse) and changelog wording.

^{Reviewed by Cursor Bugbot for commit 9366a1a. Bugbot is set up for automated code reviews on this repo. Configure here.}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 4ac8139764

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[jdchawla29]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a56c22e894

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 9366a1a. Configure here.}