[swayaam03]

Summary

Closes #1790

Fixes an issue where AsyncConversation.prompt() ignores the tools attached to the conversation instance when called without explicit tools passed into .prompt().

Cause

In llm/models.py:

• Conversation.prompt() uses tools=tools or self.tools
• Conversation.chain() uses tools=tools or self.tools
• AsyncConversation.chain() uses tools=tools or self.tools

However, AsyncConversation.prompt() was mistakenly passing tools=tools directly to Prompt(...), dropping self.tools.

Changes

1. llm/models.py: Changed tools=tools to tools=tools or self.tools in AsyncConversation.prompt().
2. tests/test_tools.py: Added unit test test_async_conversation_prompt_inherits_tools verifying that tools set on an AsyncConversation are properly passed through to the generated Prompt.

[Muguett-DBY]

I checked the base and this commit with an offline recording model. The model now receives the conversation's tool on both the first and second turns, and an explicit nonempty tools argument still takes precedence. An empty list inherits the conversation tools, matching the existing synchronous behavior. A model that doesn't support tools now rejects the inherited tool instead of silently losing it.

This exercised the exact models.py and message types with Codex assistance; the HTTP client and unused utility helpers were stubbed. I didn't run the full suite or a real provider.