Refactor function calling

[braisedpork1964]

No description provided.

[Copilot]

Pull request overview

This PR is working toward adapting AgentMessage/agent execution to better interoperate with OpenAI ChatCompletion-style responses, adding function-calling + tool execution plumbing and some new utility/actions to support that workflow.

Changes:

• Add AgentMessage fields/methods for tool-calls, finish reasons, and conversion to/from ChatCompletion.
• Update agent/memory/aggregator flows to support aborted-rollout resumption and tool-result messages.
• Introduce new utilities and actions (text truncation, function-call agent, web visitor, MCP client, Async OpenAI wrapper).

Reviewed changes

Copilot reviewed 10 out of 10 changed files in this pull request and generated 16 comments.

Show a summary per file

File	Description
lagent/utils/util.py	Extends create_object and adds truncate_text.
lagent/utils/__init__.py	Exposes truncate_text via utils public API.
lagent/schema.py	Extends AgentMessage + conversion helpers; adds tool_call_id to ActionReturn.
lagent/memory/base_memory.py	Adds model-spec serialization/deserialization for stored messages; tweaks add/delete/load/save.
lagent/llms/openai.py	Adds an AsyncOpenAIWrapper and new SDK imports.
lagent/agents/fc_agent.py	Introduces function-call selection/env agents and tool prompt builder.
lagent/agents/aggregator/default_aggregator.py	Adjusts aggregation to emit assistant/tool/user messages and includes extra_info.
lagent/agents/agent.py	Adds aborted-rollout scroll/resume behavior and ChatCompletion-to-AgentMessage conversion.
lagent/actions/web_visitor.py	Adds a web-visit + LLM-based extraction action.
lagent/actions/mcp_client.py	Adds an MCP-backed async tool wrapper with rate limiting and concurrency control.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.