WBDOCS-1656: Document chat-style rendering in Weave Trace view

[anastasiaguspan]

Summary

Adds a section to View and customize trace display explaining when the Trace view Call tab shows chat-style (bubble) rendering versus raw JSON, with links from the Trace and Threads guides.

Related

• WBDOCS-1656

Checklist

• SME accuracy review (Weave UI / provider formats)
• Editorial pass

Made with Cursor

[anastasiaguspan]

Sources and decision log

JIRA: WBDOCS-1656 — document request formatting that triggers chat UI in the trace; ticket comments note OpenAI-style schema and that Zach Birenbaum may have a list of recognized provider formats.

Existing docs cited or aligned with:

• weave/guides/tracking/threads.mdx (Chat view behavior, extracted messages from LLM calls)
• weave/guides/integrations/openai.mdx (Chat Completions messages shape)
• weave/guides/integrations/huggingface.mdx (chat view in Trace UI for chat_completion)

Placement: New content lives under View and customize trace display (view-call.mdx) because that page already covers how trace data is shown and customized. Cross-links added from Navigate the Weave Trace view and Trace threads so readers find it from Trace navigation and Threads.

Intentionally omitted: An exhaustive enumerated list of every recognized provider and JSON path. That list is not in public docs and should be confirmed with engineering before publishing as reference-grade detail.

---

Needs SME verification

Please review and check off after confirming with the Weave UI / tracing implementation.

• Complete provider list: What provider formats and field paths does the Trace Call tab recognize for chat-style rendering beyond the OpenAI Chat Completions-style messages / choices pattern and the Hugging Face Inference Client example already documented?
• Parity with Threads: For the single-call Trace view (not the Threads pane), are Anthropic and other integrations recognized the same way as in Chat view behavior (for example anthropic.Anthropic.completion), or only in specific contexts?
• Edge cases: Any calls that log valid LLM payloads but still render as raw JSON (nesting, postprocessors, multimodal content arrays, and so on) that we should mention in one sentence?

---

Resume prompt (for a follow-up session)

Continue work on #2470 for WBDOCS-1656. The draft documents chat-style rendering in the Trace view in weave/guides/tracking/view-call.mdx with cross-links from trace-tree.mdx and threads.mdx. Remaining work: SME answers to Needs SME verification above; then editorial pass; add label author-docs-plugin to the PR if the label exists in the repo (creation failed: label not found).

---

{
  "skill": "author-docs",
  "version": "unknown",
  "depends_on": {},
  "model": "unknown"
}

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated (UTC)
wandb	🟢 Ready	View Preview	Apr 16, 2026, 4:59 PM

[github-actions]

📚 Mintlify Preview Links

🔗 View Full Preview

📝 Changed (3 total)

📄 Pages (3)

File	Preview
weave/guides/tracking/threads.mdx	Threads
weave/guides/tracking/trace-tree.mdx	Trace Tree
weave/guides/tracking/view-call.mdx	View Call

---

🤖 Generated automatically when Mintlify deployment succeeds
📍 Deployment: f3bc29e at 2026-04-16 16:59:26 UTC

[github-actions]

🔗 Link Checker Results

✅ All links are valid!

No broken links were detected.

Checked against: https://wb-21fd5541-aguspan-wbdocs-1656-chat-trace-ui.mintlify.app