https://blog.margrop.net/en/tag/streaming/ Recent content in Streaming on Margrop Blog Hugo -- gohugo.io en-US Tue, 02 Jun 2026 10:42:00 +0800 https://blog.margrop.net/en/post/openclaw-tui-duplicate-thinking-stream-patch/ Tue, 02 Jun 2026 10:42:00 +0800 https://blog.margrop.net/en/post/openclaw-tui-duplicate-thinking-stream-patch/ <blockquote> <p><strong>The short version</strong></p> <p>If OpenClaw TUI shows the same thinking block twice and repeats the final answer, do not immediately delete every session or rebuild every model provider. In many OpenAI-compatible setups, the real issue is a streaming compatibility edge case: the upstream stream emits normal <code>delta</code> chunks and then an extra full <code>message</code> tail. Some OpenClaw paths may aggregate both, so the TUI and the persisted session both end up with duplicated content.</p> <p>The practical temporary fix is to back up the local OpenClaw installation, add a conservative dedupe guard in the OpenClaw aggregation layer, restart the gateway, and verify with a unique marker that both the visible reply and the session file contain only one copy.</p> </blockquote> <p>This is a privacy-safe incident note. All endpoints, model names, tokens, internal addresses, and session identifiers are replaced by placeholders. Use your own installation path and model identifiers when applying the workaround.</p> https://blog.margrop.net/en/post/newapi-streaming-duplicate-content-debugging/ Sun, 31 May 2026 18:00:00 +0800 https://blog.margrop.net/en/post/newapi-streaming-duplicate-content-debugging/ <blockquote> <p><strong>TL;DR</strong></p> <p>An AI Agent was producing duplicated replies in DingTalk — not two separate messages, but a single message whose content appeared twice. The root cause traced back to a streaming bug in NewAPI (the QuantumNous fork of OneAPI) v1.0.0-rc.10: when proxying MiniMax through the OpenAI Chat Completions protocol, the finish chunk contained <strong>both <code>delta.content</code> and <code>message.content</code></strong> with identical values. The Agent&rsquo;s stream handler consumed both as &ldquo;visible text,&rdquo; concatenating the content twice. The fix was straightforward: switch the OpenClaw provider protocol from <code>openai-completions</code> to <code>anthropic-messages</code>, which OneAPI natively supports and where streaming behaves correctly.</p> <p>No internal IPs, tokens, model IDs, or private paths appear in this article. All configuration snippets have been sanitized.</p> </blockquote>