The error you're encountering,

APIStatusError 400

, with the message "Invalid parameter: messages with role 'tool' must be a response to a preceding message with 'tool_calls'" occurs when the conversation flow between tool calls and tool responses is not properly structured in the LiveKit agents framework. Steps to Resolve: 1. Ensure Proper Message Sequencing: Verify that your conversation flow follows this pattern: ◦ An assistant message with

tool_calls

◦ Followed by tool response messages with role 'tool' and matching

tool_call_id

1. Verify Tool Call ID Matching: Ensure the

tool_call_id

in your tool response exactly matches the

call_id

from the corresponding function call. 2. Use Built-in Validation: Utilize the framework's

group_tool_calls

function to automatically group related messages, function calls, and outputs together. This function also removes invalid tool calls and outputs. 3. Check Your ChatContext Structure: Ensure your

ChatContext

contains properly structured items where: ◦

FunctionCall

items have valid

call_id

values ◦

FunctionCallOutput

items reference the correct

call_id

from their corresponding function calls 1. Review Message Building Logic: If you're manually constructing messages, ensure that: ◦ Tool response messages have the role 'tool' ◦ They include the correct

tool_call_id

matching a previous tool call ◦ Tool responses are only sent after assistant messages containing

tool_calls

The framework's built-in validation and grouping mechanisms should handle most cases automatically, but this error typically indicates that the conversation flow has been disrupted or tool call IDs are mismatched. Sources: livekit-agents/livekit/agents/llm/_provider_format/aws.py

@refined-appointment-81829 Can you please help?