feat(web): render rich Mermaid diagrams in chat

[yaacovcorcos]

Problem

Scient rendered Mermaid fences as ordinary code even when a relationship was much easier to understand visually, and supported agents had no reliable way to discover that the Scient chat surface could present diagrams.

What changed

• Render settled fenced mermaid blocks as polished inline diagram cards in chat and Markdown file previews.
• Keep streaming fences and unsupported clients on the ordinary readable code path.
• Add lazy, bounded, locally bundled Mermaid rendering with duplicate-SVG id rebasing and isolated loading/error recovery.
• Add expanded viewing with fit and 25-400% zoom, source inspection/copy, standalone SVG download, and bounded high-resolution PNG copy/download.
• Preserve Mermaid source as the canonical Markdown for whole-message copy and external readers.
• Advertise the capability through the existing narrow Codex and Claude instruction seams.
• Add user, internal, third-party, and manual fixture documentation.

User impact

Before this change, Mermaid appeared only as source code. After it, completed diagrams become responsive light/dark cards without hiding their source or breaking surrounding Markdown when a diagram is malformed. Mobile remains a readable source-code fallback.

Upstream boundary

The implementation lives under Scient-owned diagram and provider files. The inherited T3 ChatMarkdown change is limited to one import and one settled-fence branch, with a seam regression test. This PR is independent of the open chat-math PR #81 and includes none of its commits; if #81 lands first, this branch only needs that narrow seam reconciled during rebase.

Validation

• Focused web diagram suite: 4 files, 15 tests passed
• Focused provider suite: 2 files, 88 tests passed
• Web and server typechecks passed
• Targeted lint and formatting passed
• Production web build passed during the clean desktop candidate launch
• git diff --check passed
• Yaacov completed manual light/dark visual review of the diagram fixture in an isolated Scient (Dev) candidate

Implemented and verified with OpenAI Codex in the Codex app.

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	11.4 KiB	11.3 KiB	−51 B (−0.4%)	15.1 KiB	✅
Codex	Thread snapshot wire	5.5 KiB	5.5 KiB	−5 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	5.9 KiB	5.9 KiB	−46 B (−0.8%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	49.8 KiB	49.7 KiB	−59 B (−0.1%)	66.4 KiB	✅
Codex	Live turn messages	18	16	−2 (−11.1%)	21	✅
Claude	Total thread wire	11.4 KiB	11.4 KiB	+3 B (+0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	5.5 KiB	5.5 KiB	−4 B (−0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	5.9 KiB	5.9 KiB	+7 B (+0.1%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	50.6 KiB	50.6 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	18	18	0 (0.0%)	21	✅

Baseline: 5c1a63f · PR result: ae33c64 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 94.7 KiB
• Claude decoded thread snapshot: 95.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.