With a real codex login now available, record the one shape the fake-key
lab could never produce: a genuine model reply streaming above the pinned
composer, pushing lines into history (baseY grows) while keystrokes land
mid-stream. The recorder gains an opt-in CODEX_RECORD_REAL=1 scenario
using the user's own ~/.codex (fixture secret-scanned for key/JWT
material before writing; scanned clean). The replay test pins: baseY > 0,
mid-stream predictions painted, exact convergence to the typed text.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Independent post-build review found three gaps, all one family: input that
changes the composer without a prediction leaves the DISPLAYED cursor stale
for one RTT, and anchoring a new run on it painted ghosts one cell off
(blank-neutral, so they lived out the full TTL: "tehh" on
backspace-then-retype, exactly on the links the feature targets).
Fix: the addon now HOLDS new predictions after any such edit (backspace with
nothing outstanding = deleting echoed text, clearPredictions, and now also
IME/plain-paste 'text' commits, which the hook clears like 'clear') until
the next PARSED write releases the hold. The inline predictChar reconcile
deliberately does not count: only the emitter pass or the public
reconcile() is the display-caught-up contract. Worst case is exactly one
unpredicted keystroke, whose own echo releases the hold. Also patched the
one bypass path the PR had missed: _handleCjkInput now clears predictions
like insertTerminalText and the other bypass sends.
Package suite 230, vm gating 85, E2E 10/10 all green after the change.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Mosh-style write-through prediction: the consumer sends every keystroke
unchanged; the addon paints predicted glyphs and reconciles against the
parsed buffer. Confirm = cell match + cursor advance (placeholder-safe,
repaint-safe); two-pass mismatch cascade with neutral blanks (measured:
codex clears its placeholder on first echo); TTL bound; baseY-based line
reads; scroll/resize/off-row clears. Zero edits to zerolag-input-addon.ts.
Tests: 30 addon-law specs + renderer geometry (fake performance clock for
TTL/grace), 6 replay suites running the real algorithm through a real
@xterm/headless parser fed by the recorded codex fixtures, and a
500-iteration seeded fuzz with per-op span/record + grid invariants.
227 total, the pre-existing 175 untouched.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Recorder (scripts/dev/record-codex-frames.mjs) captures real codex 0.147
TUI output through the production pipeline (tmux status-off + the codex-mode
full strip from session.ts) into JSONL fixtures with keystroke injection
points; analyzer replays them through @xterm/headless for the measurements
in docs/predictive-echo-plan.md. Composer signature /^> /-style (U+203A),
modal and wrapped rows correctly rejected, echo is unstyled default-fg,
tmux delivers echo as minimal in-place deltas.
Package: types.ts gains optional cursorX/cursorY, getCell, onWriteParsed,
onResize (all additive); prediction-renderer.ts renders per-glyph spans
keyed by prediction seq; @xterm/headless@^6.0.0 devDep for replay tests.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Covers all 7 test plan items: Chinese overlap prevention, double-width
spacing, Japanese/Korean input, cursor positioning, line wrapping at
column boundaries, ASCII regression, and teammate terminal panels.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add charCellWidth/stringCellWidth helpers for Unicode-aware width detection,
fix makeLine to use for...of iteration with visual column positioning, and
fix line splitting in _render to use visual column widths instead of string
length. CJK/fullwidth characters now correctly occupy 2 cell widths.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Pitfall 3 (REAL BUG): setFlushed() called _render() unconditionally.
During tab switch, rendering against a stale buffer found the old
session's prompt and locked the column position via the flushed-text
column lock. After the new buffer loaded, rerender() kept the stale
column. Fix: add render parameter (default true), pass false during
tab-switch restore to defer rendering until buffer is loaded.
Pitfall 2: Tab completion "undo" required calling clearFlushed() +
resetBufferDetection() separately — easy to get wrong. Add
undoDetection() convenience method that atomically clears flushed
state and re-enables detection.
Pitfall 1: Backspace Map sync is app-specific (can't be fixed in lib),
but removeChar() return value + getFlushed() make the pattern clean.
Updated tab-switch test to demonstrate the correct pattern.
Tests: 78 total (+4 new)
- setFlushed render=false prevents stale column lock
- setFlushed render=false keeps overlay hidden
- undoDetection clears flushed + re-enables detection
- undoDetection preserves pending text
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Second deep audit comparing all 43 integration points in Claudeman's
app.js against the library API found one critical gap:
- Add suppressBufferDetection() — needed when switching to sessions
with UI framework text (Ink) after the prompt that would be falsely
detected. Claudeman sets _bufferDetectDone=true externally; this
method provides the public equivalent.
Tests: 74 total (+13 new)
- Tab-switch save/restore pattern (flushed state roundtrip)
- suppressBufferDetection blocks explicit, implicit (addChar), and
cascade (removeChar) detection paths
- clear() resets suppression
- Methods safe before activate() and after dispose()
- addChar implicit buffer detection on first keystroke
- refreshFont with flushed-only text
README: rewritten from 207 to 459 lines with:
- removeChar cascade explanation with return value table
- Flushed text concept explained
- Integration patterns: buffered, char-at-a-time, tab switching,
tab completion, Ink/TUI frameworks, SSE reconnect, resize, font
- Architecture diagram (keypress → DOM overlay → PTY echo flow)
- Prompt column locking, text wrapping, render cache, scroll awareness
- All known limitations documented
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
1. Regex global flag safety — strip `g` flag before matching to prevent
lastIndex mutation and missing .index on match results
2. state.visible false before activate — null overlay no longer reports
visible: true
3. removeChar() returns 'pending' | 'flushed' | false instead of boolean
so consumers can distinguish whether to send backspace to PTY
4. removeChar() implements buffer detection cascade (step 3) — detects
existing prompt text when both pending and flushed are empty
5. RenderKey includes text content, not just length — prevents stale
renders when setFlushed() called with same count but different text
6. Remove dead `import type { Terminal }` from test file
7. Remove internal `FontStyle` from public exports
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Extract Claudeman's local echo overlay into a reusable xterm.js addon
at packages/xterm-zerolag-input/. Provides instant keystroke feedback
via a DOM overlay, eliminating perceived input latency over high-RTT
connections (SSH, mobile, cloud IDEs).
- Zero dependencies, compatible with xterm v5.x and @xterm/xterm v5.4+
- Configurable prompt detection (character, regex, or custom function)
- Flushed text tracking for tab-switch / deferred echo scenarios
- Per-character grid-aligned rendering matching xterm's canvas output
- 56 tests passing (prompt finder, overlay renderer, full addon lifecycle)
- Dual CJS/ESM build with full TypeScript declarations
Claudeman source is unchanged — migration to consume this lib is a
separate follow-up.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Codeman maintainer