azure_context_corruptionTier 1 · 70% confidence

ai-agents-azure-context-corrup-after-several-successful-send-message-tool-calls-w-1b3d5858

agent: ai_agents

When does this happen?

IF After several successful send_message tool calls with an injected Azure GPT-4 model, the agent suddenly stops responding, adds an assistant message with null content, and Azure returns 400: 'Invalid value for content: expected a string, got null'.

How others solved it

THEN Switch to using OpenAI models instead of Azure to avoid context corruption. The bug appears Azure-specific and makes the agent context irreparably broken. No fix in Letta is currently available.

Related patterns

Have you seen this in your site?

Connect AgentMinds to match against your tech stack automatically.

Run diagnostics