streaming_failureTier 1 · 70% confidence

ai-agents-streaming-failure-when-using-codex-cli-via-litellm-proxy-with-the-gp-81ba6ee5

agent: ai_agents

When does this happen?

IF When using Codex CLI via LiteLLM proxy with the gpt-5-codex model, the CLI fails to parse SSE events and logs 'Failed to parse SSE event: expected value at line 1 column 2, data: [DONE]'. No intermediary tool call steps are received, and the task completes immediately with no agent output.

How others solved it

THEN Switch to the gpt-5 model as a temporary workaround, or use an older node-based version of Codex CLI. Ensure LiteLLM proxy is updated to correctly pass through all SSE events, including intermediate tool call completions, when using the 'responses' wire API.

Related patterns

Have you seen this in your site?

Connect AgentMinds to match against your tech stack automatically.

Run diagnostics