fix(tui): comprehensive TUI improvements and bug fixes by junhoyeo · Pull Request #4 · junhoyeo/tokscale

junhoyeo · 2025-12-13T02:15:24Z

Summary

Major TUI improvements including visual enhancements, bug fixes, and new features.

New Features

Copy to Clipboard (y key): Copy selected row data (vim-style keybind)
Export Feedback: Status messages for copy/export operations (shown in footer for 2 seconds)
Stacked Bar Chart: Overview bars show usage breakdown by model color

Visual Improvements

Responsive Model Column: Expands based on terminal width instead of clipping
Sort Indicators: ▲/▼ arrows on active sort column
Subtle Zebra Striping: Dark gray (#232328) alternating rows for readability
Cleaner UI: Removed ugly borders from Models/Daily tabs, removed duplicate labels

Bug Fixes

Fixed Overview Tab Selection: Blue highlight now properly follows arrow key navigation
- Root cause: Keyboard handler only updated scrollOffset, never selectedIndex
- Selection check scrollOffset + i === selectedIndex always failed
Fixed Row Selection Reactivity: Blue highlight updates when navigating with ↑↓
- Root cause: <text> elements use bg prop, not backgroundColor
- Used createMemo inside <For> loop for proper Solid.js reactivity

Code Quality

Removed unused ModelListItem component (inlined due to Solid.js reactivity)
Dead code removal, edge case handling, performance optimizations
Added keyboard modifier types to opentui.d.ts

Technical Details

OpenTUI's <text> uses bg for background color, <box> uses backgroundColor
Solid.js <For> requires createMemo inside callback for external signal reactivity
Component boundaries can break reactivity when passing accessor props

Files Changed

App.tsx - Keyboard handling, copy/export with status messages
ModelView.tsx - Responsive columns, striping, sort cues, selection fix
DailyView.tsx - Striping, selection fix
OverviewView.tsx - Inline rendering for selection reactivity
BarChart.tsx - Stacked bars with model colors
Footer.tsx - Status message display, updated hints
StatsView.tsx - Remove duplicate label
index.ts - Removed unused ModelListItem export

- Remove dead code (redundant header functions in ModelView) - Fix hex colors to terminal-compatible named colors - Pre-sort bar chart models for performance (O(n) vs O(n*m)) - Handle empty models edge case in bar chart - Remove misleading 'Shift+Drag to copy' hint - Remove unnecessary accessor wrapper - Clean up trailing/double blank lines

- Add status message feedback for copy (y) and export (e) operations - Show success/error messages in footer for 2 seconds - Add zebra striping to DailyView for consistency with ModelView - Add sort indicators (▲/▼) to DailyView column headers - Add 'e export' hint to footer

Pass selectedIndex and scrollOffset as Accessor<number> instead of number to child components. This ensures Solid.js properly tracks changes and updates the blue selection highlight when navigating with arrow keys.

<Index> in Solid.js tracks by index (number) rather than by reference, which properly reacts to selectedIndex changes without requiring the array items to change.

Following OpenCode's pattern - use createMemo inside the For callback to properly track selectedIndex changes. This ensures the comparison happens within a reactive context that Solid.js can track.

OpenTUI's <text> element uses 'bg' for background color, not 'backgroundColor'. This fixes the row selection highlight not appearing when navigating with arrow keys.

- Change ModelListItem.isSelected to Accessor<boolean> for reactivity - Pass accessor from OverviewView instead of computed boolean value - Add bg prop to text elements for proper background rendering

Solid.js component boundaries break reactivity when passing accessor props. Inline the rendering logic with createMemo inside the For loop, same pattern that works in ModelView and DailyView.

The bug: In Overview tab, up/down keys only changed scrollOffset, never selectedIndex. So the selection check always failed. Fix: - OverviewView: Check i() === selectedIndex() (within visible list) - App.tsx: Properly update selectedIndex for Overview tab, and only scroll when selection reaches edge of visible area

Use #232328 (very dark gray) for alternating rows instead of brightBlack which was too visible. Only odd rows get the stripe.

Add patterns to ignore: - *.log and ffi_otui_debug_*.log (debug logs) - opencodetmp/ directory (temporary files) - *.zip (binary archives)

Remove files that should be in a separate PR: - FRONTEND_THEMING_RESEARCH.md - RESEARCH_SUMMARY.md - THEMING_QUICK_REFERENCE.md - THEMING_RESEARCH_INDEX.md This PR should focus only on TUI improvements.

The component was inlined in OverviewView due to Solid.js component boundary reactivity issues, making this file dead code.

…737) * fix(antigravity-cli): date each turn by its per-generation timestamp parse_antigravity_cli_file stamped every gen_metadata row with the single session-created time from trajectory_metadata_blob.#2, and the module doc-comment claimed the per-turn proto carries only relative timings. It doesn't: each row carries an absolute {seconds, nanos} Timestamp at chatModel.#9.#4, the same wire shape session_created_ms already decodes. Without this, turns written into a conversation older than the live-tail window are stamped in the past and dropped from the live trace, and long conversations mis-bucket across midnight. Read chatModel.#9.#4 per row (via a shared proto_timestamp_ms) and fall back to the session-created time only when the field is absent or zero. Added per_generation_timestamp_overrides_session_fallback, which fails before the change. * fix(antigravity-cli): guard proto_timestamp_ms against i64 overflow A malformed protobuf can carry a Timestamp whose `seconds` varint is large enough that `seconds * 1000` overflows i64. With overflow-checks on (debug builds), the unchecked multiply panics instead of degrading to None, which breaks the module's malformed-data contract (every other decode path returns None on bad input). Switch to checked_mul/checked_add and return None on overflow. Regression test feeds a seconds varint of i64::MAX and asserts None without a panic, checks the in-range boundary (i64::MAX/1000) still decodes, and keeps a normal seconds+nanos stamp decoding to ms. * fix(antigravity-cli): range-validate Timestamp nanos in proto_timestamp_ms Reject `nanos` outside the protobuf Timestamp spec (0..=999_999_999), including values large enough to be negative once cast to i64. An out-of-range nanos now marks the whole stamp malformed (None) so the caller's `ms > 0` filter and session-timestamp fallback take over instead of producing a skewed per-turn time. Valid-input behavior is unchanged. Confidence: high Scope-risk: narrow --------- Co-authored-by: Junho Yeo <i@junho.io>

…unhoyeo#737) * fix(antigravity-cli): date each turn by its per-generation timestamp parse_antigravity_cli_file stamped every gen_metadata row with the single session-created time from trajectory_metadata_blob.junhoyeo#2, and the module doc-comment claimed the per-turn proto carries only relative timings. It doesn't: each row carries an absolute {seconds, nanos} Timestamp at chatModel.junhoyeo#9.junhoyeo#4, the same wire shape session_created_ms already decodes. Without this, turns written into a conversation older than the live-tail window are stamped in the past and dropped from the live trace, and long conversations mis-bucket across midnight. Read chatModel.junhoyeo#9.junhoyeo#4 per row (via a shared proto_timestamp_ms) and fall back to the session-created time only when the field is absent or zero. Added per_generation_timestamp_overrides_session_fallback, which fails before the change. * fix(antigravity-cli): guard proto_timestamp_ms against i64 overflow A malformed protobuf can carry a Timestamp whose `seconds` varint is large enough that `seconds * 1000` overflows i64. With overflow-checks on (debug builds), the unchecked multiply panics instead of degrading to None, which breaks the module's malformed-data contract (every other decode path returns None on bad input). Switch to checked_mul/checked_add and return None on overflow. Regression test feeds a seconds varint of i64::MAX and asserts None without a panic, checks the in-range boundary (i64::MAX/1000) still decodes, and keeps a normal seconds+nanos stamp decoding to ms. * fix(antigravity-cli): range-validate Timestamp nanos in proto_timestamp_ms Reject `nanos` outside the protobuf Timestamp spec (0..=999_999_999), including values large enough to be negative once cast to i64. An out-of-range nanos now marks the whole stamp malformed (None) so the caller's `ms > 0` filter and session-timestamp fallback take over instead of producing a skewed per-turn time. Valid-input behavior is unchanged. Confidence: high Scope-risk: narrow --------- Co-authored-by: Junho Yeo <i@junho.io>

…unhoyeo#737) * fix(antigravity-cli): date each turn by its per-generation timestamp parse_antigravity_cli_file stamped every gen_metadata row with the single session-created time from trajectory_metadata_blob.#2, and the module doc-comment claimed the per-turn proto carries only relative timings. It doesn't: each row carries an absolute {seconds, nanos} Timestamp at chatModel.#9.#4, the same wire shape session_created_ms already decodes. Without this, turns written into a conversation older than the live-tail window are stamped in the past and dropped from the live trace, and long conversations mis-bucket across midnight. Read chatModel.#9.#4 per row (via a shared proto_timestamp_ms) and fall back to the session-created time only when the field is absent or zero. Added per_generation_timestamp_overrides_session_fallback, which fails before the change. * fix(antigravity-cli): guard proto_timestamp_ms against i64 overflow A malformed protobuf can carry a Timestamp whose `seconds` varint is large enough that `seconds * 1000` overflows i64. With overflow-checks on (debug builds), the unchecked multiply panics instead of degrading to None, which breaks the module's malformed-data contract (every other decode path returns None on bad input). Switch to checked_mul/checked_add and return None on overflow. Regression test feeds a seconds varint of i64::MAX and asserts None without a panic, checks the in-range boundary (i64::MAX/1000) still decodes, and keeps a normal seconds+nanos stamp decoding to ms. * fix(antigravity-cli): range-validate Timestamp nanos in proto_timestamp_ms Reject `nanos` outside the protobuf Timestamp spec (0..=999_999_999), including values large enough to be negative once cast to i64. An out-of-range nanos now marks the whole stamp malformed (None) so the caller's `ms > 0` filter and session-timestamp fallback take over instead of producing a skewed per-turn time. Valid-input behavior is unchanged. Confidence: high Scope-risk: narrow --------- Co-authored-by: Junho Yeo <i@junho.io>

…-2026-06-v2 feat: update AI metric reports 2026-06

…sion window The agy 1.1.18 `chatModel.#9.#10` payload is 8 bytes whose encoding was never confirmed against a real database, so every candidate reading of it is a guess that has to earn acceptance. The only gate on those guesses was an absolute "is this a believable date" window running from 2020-01-01 to five years out. That is not a meaningful test for a raw integer: read as a nanosecond count the window alone covers ~2% of the u64 range, so trying both byte orders leaves an arbitrary payload — an id, a hash, a duration — a few percent chance of passing as a date. A false accept silently buckets a turn into the wrong day and feeds the server-side monotonic ratchet, which has no correction path, making it strictly worse than the known-wrong session-start dating it replaces. Require every inferred reading to land inside the containing session's own lifetime as well: at or after the session-created stamp less one hour, and at or before now plus one hour. A turn cannot predate its conversation nor happen after we read the file, and that pair of bounds is hours or days wide instead of a decade. When there is no positive anchor to corroborate against, decline inference entirely and let the caller fall back as before. The explicit `#9.#4` Timestamp is untouched: it is a confirmed representation read off real pre-1.1.18 databases, keeps its `ms > 0` filter, takes no session bound, and still outranks the inferred reading. Constraint: `#9.#10`'s encoding is inferred from a field dump, not observed Constraint: mis-dating is uncorrectable downstream; under-dating is not Rejected: tightening only the absolute window | no absolute date range is narrow enough to make a raw 8-byte integer a safe timestamp Rejected: day-wide tolerances | hands back the integer space the session window exists to remove Confidence: high Scope-risk: narrow Directive: the one-hour tolerances are load-bearing and pinned by tests; widening them re-opens the false-accept surface this closes Not-tested: a real agy 1.1.18 database — none was available, which is why the reading is inferred in the first place

…1.1.18 gen_metadata layout (#1196) * fix(antigravity-cli): read the per-generation timestamp from the agy 1.1.18 gen_metadata layout agy 1.1.18 dropped `chatModel.#9.#4`, the `{#1: seconds, #2: nanos}` Timestamp this parser used to date each turn. `#9` now carries `#2` = u64::MAX (an int64 -1 "unset" sentinel) plus a new `#10` holding 8 length-delimited bytes. With `#4` gone every row fell through to the session-created stamp, so on a long-running session every turn was bucketed to the session start date and `--today` reported zero. `#9.#4` is still read first and unchanged, so pre-1.1.18 databases and older installs keep their exact behaviour. When it is absent, `#9.#10` is decoded as a nested Timestamp, a nested message holding the scalar in field 1 (varint or fixed64), or the payload itself as 8 raw fixed64-style bytes in either byte order. Every one of those readings is unit-detected by magnitude and range-checked against 2020-01-01..now+5y before it is accepted; anything outside that window is discarded and the session-created fallback takes over. `#9.#2` is never consulted, and u64::MAX is rejected explicitly so no path can promote the sentinel into a date. Constraint: no agy 1.1.18 install or gen_metadata database available, so `#10`'s encoding is inferred from a field dump in the issue, not observed Rejected: read `#9.#2` as the new timestamp | the only value ever seen there is the u64::MAX unset sentinel Rejected: decode the 8 bytes as an IEEE-754 f64 | any double in the 2^30-ish exponent range reads as a plausible epoch-second count, so it is the one candidate with a non-trivial false-positive rate against a non-timestamp payload Confidence: high that pre-1.1.18 parsing is unchanged; medium that the 1.1.18 reading fires on real data Scope-risk: narrow Directive: keep every inferred reading behind `plausible_epoch_ms` — a wrong date silently corrupts day buckets and the monotonic ratchet, which is worse than the session-start fallback this degrades to Not-tested: a real agy 1.1.18 `gen_metadata` row; if `#10` is neither a timestamp nor decodes in range, behaviour is identical to today's * fix(antigravity-cli): bound inferred generation timestamps to the session window The agy 1.1.18 `chatModel.#9.#10` payload is 8 bytes whose encoding was never confirmed against a real database, so every candidate reading of it is a guess that has to earn acceptance. The only gate on those guesses was an absolute "is this a believable date" window running from 2020-01-01 to five years out. That is not a meaningful test for a raw integer: read as a nanosecond count the window alone covers ~2% of the u64 range, so trying both byte orders leaves an arbitrary payload — an id, a hash, a duration — a few percent chance of passing as a date. A false accept silently buckets a turn into the wrong day and feeds the server-side monotonic ratchet, which has no correction path, making it strictly worse than the known-wrong session-start dating it replaces. Require every inferred reading to land inside the containing session's own lifetime as well: at or after the session-created stamp less one hour, and at or before now plus one hour. A turn cannot predate its conversation nor happen after we read the file, and that pair of bounds is hours or days wide instead of a decade. When there is no positive anchor to corroborate against, decline inference entirely and let the caller fall back as before. The explicit `#9.#4` Timestamp is untouched: it is a confirmed representation read off real pre-1.1.18 databases, keeps its `ms > 0` filter, takes no session bound, and still outranks the inferred reading. Constraint: `#9.#10`'s encoding is inferred from a field dump, not observed Constraint: mis-dating is uncorrectable downstream; under-dating is not Rejected: tightening only the absolute window | no absolute date range is narrow enough to make a raw 8-byte integer a safe timestamp Rejected: day-wide tolerances | hands back the integer space the session window exists to remove Confidence: high Scope-risk: narrow Directive: the one-hour tolerances are load-bearing and pinned by tests; widening them re-opens the false-accept surface this closes Not-tested: a real agy 1.1.18 database — none was available, which is why the reading is inferred in the first place

…ded session anchor (#1202) `read_trajectory_meta` collapsed the decoded `trajectory_metadata_blob` created-at and the file mtime into one `i64`, and that number was used both as the per-row timestamp fallback and as the trust anchor bounding the inferred agy 1.1.18 `chatModel.#9.#10` reading. A file mtime is always positive, so the `session_timestamp <= 0` guard that was supposed to decline inference without an anchor almost never fired: databases with a missing or undecodable metadata blob ran inference against a window built on the last write to the file. An opaque payload decoding near the mtime was accepted, and a genuine older turn was rejected for sitting below it. Report the created-at as `Option<i64>` alongside the effective fallback and thread it through `parse_gen_metadata` into `generation_timestamp_ms` and `inferred_epoch_ms`. With no decoded created-at the `#9.#10` inference is skipped entirely and the row keeps the fallback dating it had before the 1.1.18 layout was handled. The confirmed `#9.#4` Timestamp path returns before any of this and is unchanged. The existing missing-anchor test only passed 0 and -1, neither of which a real database produces. Adds an end-to-end SQLite test covering both shapes that yield no created-at (table absent, and a blob carrying no `#2`) with a payload the old mtime-anchored window would have accepted. Constraint: `#9.#10`'s 8-byte layout is inferred from a field dump, not read off a real database, so an accepted reading must be corroborated Rejected: Keep one i64 and sentinel the mtime case | any in-band sentinel is a valid epoch-ms value Confidence: high Scope-risk: narrow Directive: `session_window_ms` must only ever be handed a decoded created-at; do not reintroduce the mtime as an anchor Not-tested: a real agy 1.1.18 database (none available; `#9.#10` remains inferred)

junhoyeo added 2 commits December 13, 2025 11:13

fix(tui): prevent model names from clipping

094c386

fix(tui): allow model column to expand

767e091

junhoyeo changed the title ~~fix(tui): improve model table width~~ [DONT MERGE] fix(tui): improve model table width Dec 13, 2025

junhoyeo added 2 commits December 13, 2025 11:19

fix(tui): stripe model rows and clamp render

a0c76ab

fix(tui): add sort cues and preformat rows

3f3e6c5

junhoyeo changed the title ~~[DONT MERGE] fix(tui): improve model table width~~ fix(tui): improve model table readability Dec 13, 2025

fix(tui): stack bar graph models and hint copy

2c2c8ca

junhoyeo changed the title ~~fix(tui): improve model table readability~~ fix(tui): stacked bar charts and model table improvements Dec 13, 2025

junhoyeo changed the title ~~fix(tui): stacked bar charts and model table improvements~~ fix(tui): stacked bar charts, model table, and audit fixes Dec 13, 2025

junhoyeo force-pushed the junhoyeo/fix-tables branch from 05a8ed5 to 9f91635 Compare December 13, 2025 03:39

junhoyeo added 15 commits December 13, 2025 12:43

fix(tui): remove ugly border and duplicate Models label

9aaa772

fix(tui): remove border from DailyView

10f52e4

feat(tui): add Cmd+C to copy selected row

23d8ced

fix(tui): use y key for copy (vim-style)

8e34b04

fix(tui): fix selectedIndex reactivity for row selection

830a72b

Pass selectedIndex and scrollOffset as Accessor<number> instead of number to child components. This ensures Solid.js properly tracks changes and updates the blue selection highlight when navigating with arrow keys.

fix(tui): inline selection logic to ensure reactivity in For loop

0a5d30b

fix(tui): use Index instead of For for proper selectedIndex reactivity

0570bf9

<Index> in Solid.js tracks by index (number) rather than by reference, which properly reacts to selectedIndex changes without requiring the array items to change.

fix(tui): use createMemo inside For loop for selection reactivity

82791e6

Following OpenCode's pattern - use createMemo inside the For callback to properly track selectedIndex changes. This ensures the comparison happens within a reactive context that Solid.js can track.

fix(tui): use 'bg' prop instead of 'backgroundColor' for text elements

9853a6e

OpenTUI's <text> element uses 'bg' for background color, not 'backgroundColor'. This fixes the row selection highlight not appearing when navigating with arrow keys.

fix(tui): fix selection reactivity in OverviewView

f3372c9

- Change ModelListItem.isSelected to Accessor<boolean> for reactivity - Pass accessor from OverviewView instead of computed boolean value - Add bg prop to text elements for proper background rendering

fix(tui): remove zebra striping for cleaner look

5a8a0bc

fix(tui): inline model list rendering in OverviewView for reactivity

777079b

Solid.js component boundaries break reactivity when passing accessor props. Inline the rendering logic with createMemo inside the For loop, same pattern that works in ModelView and DailyView.

feat(tui): add subtle zebra striping with darker color

ee3c95b

Use #232328 (very dark gray) for alternating rows instead of brightBlack which was too visible. Only odd rows get the stripe.

junhoyeo changed the title ~~fix(tui): stacked bar charts, model table, and audit fixes~~ fix(tui): comprehensive TUI improvements and bug fixes Dec 13, 2025

junhoyeo force-pushed the junhoyeo/fix-tables branch from ca95169 to 5190223 Compare December 13, 2025 04:56

junhoyeo added 2 commits December 13, 2025 13:58

chore: update .gitignore to prevent accidental commits

9f9c267

Add patterns to ignore: - *.log and ffi_otui_debug_*.log (debug logs) - opencodetmp/ directory (temporary files) - *.zip (binary archives)

chore: remove unrelated research documentation

efb43d6

Remove files that should be in a separate PR: - FRONTEND_THEMING_RESEARCH.md - RESEARCH_SUMMARY.md - THEMING_QUICK_REFERENCE.md - THEMING_RESEARCH_INDEX.md This PR should focus only on TUI improvements.

junhoyeo force-pushed the junhoyeo/fix-tables branch from 5190223 to efb43d6 Compare December 13, 2025 04:58

chore: remove unused ModelListItem component

761e9bb

The component was inlined in OverviewView due to Solid.js component boundary reactivity issues, making this file dead code.

junhoyeo merged commit ea740a0 into main Dec 13, 2025
2 checks passed

junhoyeo deleted the junhoyeo/fix-tables branch December 13, 2025 05:07

ShivamB25 mentioned this pull request Apr 18, 2026

fix(submit): make replayed history deterministic #444

Closed

4 tasks

Ailcope mentioned this pull request Jun 12, 2026

Support agy CLI (Gemini Code Assist) as data source #711

Closed

cubic-dev-ai Bot mentioned this pull request Jun 21, 2026

fix(antigravity-cli): date each turn by its per-generation timestamp #737

Merged

leecoder pushed a commit to leecoder/tokscale that referenced this pull request Jul 7, 2026

Merge pull request junhoyeo#4 from tmobi-internal/feat/update-reports…

149154e

…-2026-06-v2 feat: update AI metric reports 2026-06

cubic-dev-ai Bot mentioned this pull request Aug 25, 2026

fix(antigravity-cli): read the per-generation timestamp from the agy 1.1.18 gen_metadata layout #1196

Merged

Navigation Menu

fix(tui): comprehensive TUI improvements and bug fixes - #4

fix(tui): comprehensive TUI improvements and bug fixes#4
junhoyeo merged 24 commits into
mainfrom
junhoyeo/fix-tables

junhoyeo commented Dec 13, 2025 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

junhoyeo commented Dec 13, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

New Features

Visual Improvements

Bug Fixes

Code Quality

Technical Details

Files Changed

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

junhoyeo commented Dec 13, 2025 •

edited

Loading