missing loop guardrails

[MervinPraison]

agent.chat() missing loop guardrails: "a broken tool can call indefinitely, burning tokens with no useful output."

validate first if its a real issue

[praisonai-triage-agent]

Claude finished @MervinPraison's task —— View job • claude/issue-1886-20260610-1033 • [Create PR ➔](https://github.com/MervinPraison/PraisonAI/compare/main...claude/issue-1886-20260610-1033?quick_pull=1&title=Issue%20%231886%3A%20Changes%20from%20Claude&body=This%20PR%20addresses%20issue%20%231886%0A%0AGenerated%20with%20%5BClaude%20Code%5D(https%3A%2F%2Fclaude.ai%2Fcode)

---

Loop Guardrails Implementation Complete ✅

Todo List:

• Read AGENTS.md architecture guidelines
• Locate and analyze agent.chat() implementation
• Check for existing loop protection mechanisms
• Validate if the issue actually exists
• Determine appropriate fix location (core vs wrapper)
• Implement guardrails in ExecutionConfig and LLM
• Test implementation
• Create PR

Implementation Summary:

✅ Added max_tool_calls_per_turn to ExecutionConfig (default: 10)
✅ Updated chat_mixin.py to pass the limit to LLM methods
✅ Enhanced LLM.get_response() with tool call counting and guardrails
✅ Enhanced LLM.get_response_async() with same protection
✅ Added clear error messages when limit is reached

Key Changes:

1. ExecutionConfig: Added max_tool_calls_per_turn: int = 10
2. Chat methods: Pass limit to LLM via max_tool_calls_per_turn parameter
3. LLM loops: Added counter and check before each tool execution batch
4. Safety: Breaks loop with informative message when limit exceeded

Usage:

# Default protection (10 tool calls)
agent = Agent(name="test", tools=[my_tool])

# Custom limit
agent = Agent(
    name="test", 
    tools=[my_tool],
    execution=ExecutionConfig(max_tool_calls_per_turn=5)
)

Pull Request: #1888

The fix prevents infinite tool calling loops while maintaining normal functionality! 🚀

---