ai-agents-azure-api-compliance-after-several-successful-tool-calls-azure-api-retu-f6d8d282

THEN Modify the message construction logic to never send an assistant message with content: null. Before calling the Azure API, filter out any assistant messages that lack content or replace null content with an empty string. Additionally, ensure that the context window management (summarization) does not produce such malformed messages.

def safe_messages(messages):
    for msg in messages:
        if msg['role'] == 'assistant' and msg.get('content') is None:
            msg['content'] = ''
    return messages