ui: chat HTML rework — quiet mode + smaller fonts + contextual ASCII flourishes

[dreamrec]

Summary

Three asks from live testing on v2.0.1:

1. The aA font-size toggle felt out of place / weird scaling — remove.
2. The chat is too busy reading every tool call — add a toggle that hides tool detail and shows just user prompts + assistant text.
3. Make the fonts a bit smaller everywhere, and add some little contextual ASCII art that fits the existing retro CRT aesthetic.

All three, in one branch.

Removed: aA font-size toggle (ex-PR-27)

HTML button	gone
CSS (#font-toggle, .fs-a, .fs-A, :root.fs-small, :root.fs-large)	gone — ~50 LOC
JS ($fontToggle, FONT_KEY, applyFontMode, click, keyboard shortcut)	gone
tests/test_chat_html_font_toggle.py (23 tests)	deleted

Added: quiet-mode toggle

Same far-right slot in the status bar. Hollow ○ (off) / filled ● (on) glyph. State persists per browser via localStorage["tdpilot.quietMode"]. Three CSS rules hide tool surfaces:

:root.quiet-mode .msg.tool_call,
:root.quiet-mode .msg.tool_result,
:root.quiet-mode details.tool-pair { display: none; }

User prompts, assistant text, errors, hints stay visible. The status bar at the top still shows the agent's per-turn state ("thinking" / "calling td_create_node…") so progress is never blacked out — quiet mode hides detail, not signal.

Keyboard shortcut: Cmd/Ctrl + . (replaces the v2.0.0 font-zoom shortcuts).

Smaller default fonts

Element	Was	Now
body / chat content	13px	12px
status bar	11px	10px
input textarea	13px	12px

Status-bar letter-spacing softened 0.1em → 0.08em (compensates for the tighter base). line-height bumped 1.5 → 1.55.

Contextual ASCII flourishes

On agent turn-end (state transitions from working → idle), attachFlourishToLastAssistant() walks back through #history, finds the most recent assistant message that doesn't already carry a flourish, and appends a .msg-flourish div with a topic-matched glyph from a 9-bucket pool:

default  ▰▰▰   ∴   ◇ ◇ ◇   ═══   · · ·   ──◆──   ·:·:·   ▱▰▱
light    ✺ ✸ ✺   \●/   ─ ✷ ─   ※ ✺ ※
camera   [◉]   ▤ ▥ ▤   ▣ ─ ▣   ┌─◎─┐
particle · ∴ · ∴ ·   ⋆ ✦ ⋆   ·:·∴·:·   ∗ ⋆ ∗
material ▓▒░ ░▒▓   ░▒▓▒░   ▓ ─ ▓
audio    ~~~ ∿ ~~~   ∿∿∿∿   · ∿ · ∿ ·
geom     ◆ ─ ◇ ─ ◆   △ ▽ △   □ ▣ □   ◯ ◉ ◯
error    × × ×   !? ─ ?!   × ─ ✗ ─ ×
success  ✓ ✓ ✓   ✦ ─ ▰ ─ ✦   ── ✓ ──   · · ✦ · ·

8 topic-detection regexes against the last user message + assistant body. If any non-default matches, 75% chance to draw from that bucket; 25% fallback to default for variety. Chosen topic name appears in tiny uppercase under the glyph as one-word post-context.

CSS aesthetic:

• --accent-dim color (purple-grey, not the bright primary accent)
• centered, 0.5em letter-spacing — reads as punctuation
• 0.55 opacity
• 1px dashed top border separating from the message body
• no animation, no emoji, no extra weight

Test plan

• uv run --extra dev pytest tests/test_chat_html_v210_ui.py -v → 18 pass
• uv run --extra dev pytest tests/ -q → 1635 pass / 12 deselected (was 1640: −23 font-toggle tests, +18 v2.1.0 tests)
• uv run --extra dev ruff check . → clean
• uv run --extra dev ruff format --check . → 298 files OK
• bash scripts/check_no_personal_paths.sh → clean
• python scripts/check_versions.py → in sync at v2.0.0 (no version bump in this PR — v2.1.0 bump happens in the release commit)

.tox rebuild

Required (chat HTML is in standalone tox _SOURCE_FILES). Pairs with the v2.0.1 release rebuild that's already pending in PR #21 — single TD Textport rebuild covers both.

Versioning

This is a new minor release (2.1.0 — feature additions: quiet mode + flourishes). Suggest sequencing:

1. Land PR #21 (v2.0.1 patch — bugfixes only)
2. Rebuild .tox + tag v2.0.1 + npm publish
3. Land this PR (UI rework)
4. Open a v2.1.0 release commit (similar to PR release: 2.0.1 — 12 v2.0 audit fixes + live model-tier propagation #21 pattern)
5. Rebuild .tox again + tag v2.1.0 + npm publish

OR collapse: bump straight to v2.1.0, ship everything together. Your call.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 830bf1819e

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Gate flourish insertion on actual assistant output

Only checking wasWorking && newState === 'idle' means a flourish is appended even when the just-finished transition produced no new assistant message (for example, a stop/cancel path). Because chat history rehydration (fullSync) rebuilds plain assistant messages without prior flourish DOM, this can decorate an older reply and make it look like a fresh completion happened. Please tie flourish insertion to whether a new assistant message was appended in the current turn, not just the status transition.

Useful? React with 👍 / 👎.