feat: TUI polish (truncation, expanded/collapsed, in-flight paths, render cache)

[code-yeongyu]

Summary

Atomic TDD commits implementing 7 TUI improvements:

1. truncatePreview helper with head/tail strategy + PATCH_PREVIEW_MAX_* constants
2. displayPath cwd-relative formatting
3. formatPatchPreview collapsed (header-only) vs expanded (with diff) modes — backward compat preserved via optional defaults
4. formatInFlightCallText for in-flight renderCall labels (path extraction + count)
5. argsComplete fallback to "Patching" placeholder during streaming
6. per-toolCallId render state cache + clearApplyPatchRenderState
7. Pi renderDiff integration with safe try/catch fallback to themed manual rendering

Backward compatibility

• All previously-existing function signatures preserved via optional parameters with defaults
• Tool wire format unchanged (name, description, parameters schema, freeform Lark grammar)
• Existing 18 tests pass without modification

Tests

• Existing 18 tests pass unchanged
• New test/render.test.ts (13 tests) covering: truncatePreview, displayPath, formatPatchPreview collapsed/expanded, formatInFlightCallText, render state cache, renderResult collapsed/multi-file/truncation, renderCall argsComplete/paths
• Total: 31 passing

Verification

• npm run check: clean
• npm test: 31/31 green

Commits (atomic TDD)

(list will be auto-shown in PR)

[cubic-dev-ai]

2 issues found across 3 files

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="src/index.ts">

<violation number="1" location="src/index.ts:101">
P2: Render-state caching is unbounded and retains full patch strings per tool call, which can cause memory growth during long sessions.</violation>

<violation number="2" location="src/index.ts:331">
P2: `truncatePreview` can duplicate lines for short-but-very-long diffs because head/tail slices overlap. This produces confusing previews and inaccurate truncation output.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}

[cubic-dev-ai]

P2: truncatePreview can duplicate lines for short-but-very-long diffs because head/tail slices overlap. This produces confusing previews and inaccurate truncation output.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At src/index.ts, line 331:

<comment>`truncatePreview` can duplicate lines for short-but-very-long diffs because head/tail slices overlap. This produces confusing previews and inaccurate truncation output.</comment>

<file context>
@@ -275,16 +309,48 @@ function formatPatchOperation(operation: ApplyPatchOperation): string {
+	}
+
+	const lines = text.split("\n");
+	const head = lines.slice(0, PATCH_PREVIEW_HEAD_LINES);
+	const tail = lines.slice(-PATCH_PREVIEW_TAIL_LINES);
+	let preview = [...head, "…", ...tail].join("\n");
</file context>

Suggested change

	const head = lines.slice(0, PATCH_PREVIEW_HEAD_LINES);
	const tail = lines.slice(-PATCH_PREVIEW_TAIL_LINES);
	const headCount = Math.min(PATCH_PREVIEW_HEAD_LINES, Math.ceil(lines.length / 2));
	const tailStart = Math.max(headCount, lines.length - PATCH_PREVIEW_TAIL_LINES);
	const head = lines.slice(0, headCount);
	const tail = lines.slice(tailStart);

[cubic-dev-ai]

P2: Render-state caching is unbounded and retains full patch strings per tool call, which can cause memory growth during long sessions.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At src/index.ts, line 101:

<comment>Render-state caching is unbounded and retains full patch strings per tool call, which can cause memory growth during long sessions.</comment>

<file context>
@@ -86,6 +94,11 @@ type ApplyPatchTheme = {
+export const PATCH_PREVIEW_MAX_CHARS = 4000;
+const PATCH_PREVIEW_HEAD_LINES = 8;
+const PATCH_PREVIEW_TAIL_LINES = 8;
+const applyPatchRenderStates = new Map<string, ApplyPatchRenderState>();
 
 function normalizeApplyPatchArguments(args: unknown): ApplyPatchParams {
</file context>