@code-yeongyu/senpi-codemode 2026.8.3-3 → 2026.8.4

package/CHANGELOG.md

@@ -12,6 +12,22 @@
 
 ### Removed
 
+## [2026.8.4] - 2026-08-04
+
+### Breaking Changes
+
+- Replaced the eval tool's optional presentation `title` with a required user-language `summary`: every eval call must now describe the cell's purpose in the user's language, callers using `title` must migrate to `summary`, and the generated tool schema, prompt contract, README examples, bridge fixtures, and test corpus all enforce the new argument ([#695](https://github.com/code-yeongyu/senpi/pull/695)).
+
+### Added
+
+- Rendered each eval summary inside its transcript cell frame and used the same summary to label detached cells and their completion notices, so concurrent or long-running JavaScript and Python work remains identifiable after detachment and when results arrive asynchronously ([#695](https://github.com/code-yeongyu/senpi/pull/695)).
+
+### Changed
+
+### Fixed
+
+### Removed
+
 ## [2026.8.3-3] - 2026-08-03
 
 ### Breaking Changes

package/README.md

@@ -14,7 +14,7 @@ task-tool names are known.
 	handle and continue in their existing kernel. Completion is injected with the
 	final value/error and buffered output; use `eval({ action: "peek"|"stop",
 	cell_id })` to inspect or terminate a detached cell. A running peek preserves
-	the original code and title together with current output, phase, status
+	the original code and summary together with current output, phase, status
 	events, tool-call summaries, elapsed duration, and structured display state;
 	a terminal peek preserves the exact final result.
 - Loopback, bearer-authenticated kernel bridge with bounded JSONL frames.
@@ -127,6 +127,16 @@ updates, and transcripts remain owned by that engine.
 `isolated`, `apply`, and `merge` are accepted for compatibility but emit a
 warning because this task-engine integration has no isolation model.
 
+## Required summary
+
+Every `eval` run call MUST include a `summary` — one line in the user's
+conversational language stating what the cell does and for what purpose (e.g.
+a Korean conversation produces a Korean summary such as "src 전체에서
+legacyClient 사용처 집계"). The summary is shown in the TUI while the cell
+runs and in the finished result, so you can always tell what is running and
+why. Values longer than 80 characters are force-truncated. A run request
+without a `summary` fails with a teaching error.
+
 ## Detached cells
 
 `eval` accepts `on_timeout: "detach"|"error"`. The default is `"detach"` in
@@ -137,8 +147,8 @@ a busy error with its cell id and output tail; calls in other languages continue
 normally. Do not re-run the cell.
 
 While any cell is detached, the interactive footer shows a highlighted
-`↗ <language> · <title>` status on the extension status line (the cell id when
-the call had no title), clearing as soon as the last detached cell settles.
+`↗ <language> · <summary>` status on the extension status line (the cell id
+when the call had no summary), clearing as soon as the last detached cell settles.
 
 Use `eval({ action: "peek", cell_id })` for its state and buffered output, or
 `eval({ action: "stop", cell_id })` to cancel it. Python stop interrupts the

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@code-yeongyu/senpi-codemode",
-	"version": "2026.8.3-3",
+	"version": "2026.8.4",
 	"description": "Source-only senpi extension package for codemode evaluation tools",
 	"type": "module",
 	"main": "./src/index.ts",
@@ -30,14 +30,14 @@
 	},
 	"dependencies": {
 		"@babel/parser": "8.0.4",
-		"@earendil-works/pi-ai": "npm:@code-yeongyu/senpi-ai@2026.8.3-3",
+		"@earendil-works/pi-ai": "npm:@code-yeongyu/senpi-ai@2026.8.4",
 		"typebox": "1.3.8"
 	},
 	"peerDependencies": {
-		"@code-yeongyu/senpi": "2026.8.3-3"
+		"@code-yeongyu/senpi": "2026.8.4"
 	},
 	"devDependencies": {
-		"@code-yeongyu/senpi": "2026.8.3-3"
+		"@code-yeongyu/senpi": "2026.8.4"
 	},
 	"keywords": [
 		"senpi",

package/src/extension/eval-status.ts

@@ -28,7 +28,7 @@ function packLabels(labels: readonly string[], budget: number): string {
 }
 
 function labelOf(entry: EvalDetachedCellStatusEntry): string {
-	return entry.title === undefined || entry.title.length === 0 ? entry.cellId : entry.title;
+	return entry.summary === undefined || entry.summary.length === 0 ? entry.cellId : entry.summary;
 }
 
 /**

package/src/prompt/eval-prompt.ts

@@ -59,7 +59,7 @@ type Context = Readonly<Record<string, ContextValue>>;
 type EvalPromptExample = {
 	readonly caption: string;
 	readonly language: keyof EnabledLanguages;
-	readonly title: string;
+	readonly summary: string;
 	readonly code: string;
 };
 
@@ -71,19 +71,19 @@ const REUSE_CHAIN_EXAMPLES = [
 	{
 		caption: "First call — set up once",
 		language: "py",
-		title: "collect targets",
+		summary: "Count all TypeScript source files under src/ excluding tests",
 		code: "from pathlib import Path\nfrom collections import Counter\nfiles = [p for p in Path('src').rglob('*.ts') if 'test' not in p.parts]\nprint(len(files))",
 	},
 	{
 		caption: "Second call — reuse `files`, batch-read in one cell",
 		language: "py",
-		title: "scan usages",
+		summary: "Find which files reference legacyClient so we know what to migrate",
 		code: "hits = Counter()\nfor p in files:\n    hits[p.name] = read(p).count('legacyClient')\ndisplay({k: v for k, v in hits.items() if v})",
 	},
 	{
 		caption: "Third call — reuse results, fan out session tools in parallel",
 		language: "py",
-		title: "confirm callsites",
+		summary: "Confirm exact callsite lines in each directory to plan the refactor",
 		code: "dirs = ['src/core', 'src/tools']\ndisplay(parallel([lambda d=d: tool.grep({'pattern': 'legacyClient', 'path': d}) for d in dirs]))",
 	},
 ] as const satisfies readonly EvalPromptExample[];
@@ -122,7 +122,7 @@ Fields:
 
 - \`language\` — {{#if py}}\`"py"\` IPython kernel{{/if}}{{#ifAll py js}}, {{/ifAll}}{{#if js}}\`"js"\` persistent JavaScript VM{{/if}}{{#if rb}}{{#ifAny py js}}, {{/ifAny}}\`"rb"\` persistent Ruby kernel{{/if}}{{#if jl}}{{#ifAny py js rb}}, {{/ifAny}}\`"jl"\` persistent Julia kernel{{/if}}.
 - \`code\` — cell body, verbatim. Newlines/quotes JSON-encoded; no fences, no headers.
-- \`title\` (optional) — short transcript label (e.g. \`"imports"\`).
+- \`summary\` (REQUIRED for run) — ONE line in the USER'S conversational language stating WHAT this cell does and FOR WHAT PURPOSE (e.g. Korean conversation -> "src 전체에서 legacyClient 사용처 집계"); shown in the TUI while the cell runs; >80 chars is force-truncated.
 - \`timeout\` (optional) — seconds. Raise only for heavy compute or long{{#if spawns}} non-agent{{/if}} tool calls.
 - \`on_timeout\` (optional) — \`"detach"\` keeps pure computation running in interactive sessions (the default); \`"error"\` interrupts for deadline-sensitive work and is the print/json default.
 - \`reset\` (optional) — wipe this language's kernel first.{{#ifAll py js}} Per-language: a \`py\` reset never touches the JS VM.{{/ifAll}}
@@ -214,7 +214,7 @@ export function buildEvalPrompt(
 	};
 	const examples = REUSE_CHAIN_EXAMPLES.filter((example) => enabled[example.language])
 		.map((example) => {
-			const call = { language: example.language, title: example.title, code: example.code };
+			const call = { language: example.language, summary: example.summary, code: example.code };
 			return `### ${example.caption}\n\`\`\`json\n${JSON.stringify(call, null, 2)}\n\`\`\``;
 		})
 		.join("\n\n");

package/src/tool/cell-runtime.ts

@@ -119,7 +119,7 @@ export class CellResultBuilder {
 		return {
 			language: this.#state.input.language,
 			languages: [this.#state.input.language],
-			...(this.#state.input.title === undefined ? {} : { title: this.#state.input.title }),
+			...(this.#state.input.summary === undefined ? {} : { summary: this.#state.input.summary }),
 			durationMs: this.#state.durationMs,
 			toolCalls: [...this.#state.toolCalls],
 			truncated: output?.truncated ?? false,
@@ -128,7 +128,7 @@ export class CellResultBuilder {
 			cells: [
 				{
 					index: 0,
-					...(this.#state.input.title === undefined ? {} : { title: this.#state.input.title }),
+					...(this.#state.input.summary === undefined ? {} : { summary: this.#state.input.summary }),
 					code: this.#state.input.code,
 					language: this.#state.input.language,
 					output: this.#state.output,
@@ -146,12 +146,12 @@ export class CellResultBuilder {
 	}
 
 	#liveUpdateText(): string {
-		const title = this.#state.input.title === undefined ? "" : ` ${this.#state.input.title}`;
+		const summary = this.#state.input.summary === undefined ? "" : ` ${this.#state.input.summary}`;
 		const aggregateOutput = this.#output.aggregateText();
 		const outputLines = aggregateOutput.split("\n");
 		const hasTrailingNewline = aggregateOutput.endsWith("\n");
 		if (hasTrailingNewline) outputLines.pop();
 		const output = `${outputLines.slice(-8).join("\n")}${hasTrailingNewline ? "\n" : ""}`;
-		return `1/1 cells ${this.#state.status}\n[1] ${this.#state.input.language}${title} ${this.#state.status}${output.length === 0 ? "" : `\n${output}`}`;
+		return `1/1 cells ${this.#state.status}\n[1] ${this.#state.input.language}${summary} ${this.#state.status}${output.length === 0 ? "" : `\n${output}`}`;
 	}
 }

package/src/tool/detached-cell-manager.ts

@@ -50,7 +50,7 @@ export interface EvalDetachedCellNotifier {
 export interface EvalDetachedCellStatusEntry {
 	readonly cellId: string;
 	readonly language: EvalLanguage;
-	readonly title?: string;
+	readonly summary?: string;
 	readonly startedAtMs: number;
 }
 
@@ -193,7 +193,7 @@ export class EvalDetachedCellManager {
 				cellId: cell.cellId,
 				language: cell.input.language,
 				startedAtMs: cell.startedAtMs,
-				...(cell.input.title === undefined ? {} : { title: cell.input.title }),
+				...(cell.input.summary === undefined ? {} : { summary: cell.input.summary }),
 			})),
 		);
 	}

package/src/tool/detached-cell-snapshot.ts

@@ -52,14 +52,14 @@ function fallbackResult(input: EvalToolInput): AgentToolResult<EvalToolDetails>
 		details: {
 			language: input.language,
 			languages: [input.language],
-			...(input.title === undefined ? {} : { title: input.title }),
+			...(input.summary === undefined ? {} : { summary: input.summary }),
 			durationMs: 0,
 			toolCalls: [],
 			truncated: false,
 			cells: [
 				{
 					index: 0,
-					...(input.title === undefined ? {} : { title: input.title }),
+					...(input.summary === undefined ? {} : { summary: input.summary }),
 					code: input.code,
 					language: input.language,
 					output: "",

package/src/tool/eval-request.ts

@@ -1,8 +1,21 @@
 import type { ExtensionContext } from "@code-yeongyu/senpi";
-import type { EvalControlInput, EvalToolInput, EvalToolRequest } from "./types.ts";
+import { EVAL_SUMMARY_MAX_LENGTH, type EvalControlInput, type EvalToolInput, type EvalToolRequest } from "./types.ts";
 
 const NON_INTERACTIVE_MODES = new Set(["print", "json"]);
 
+const ELLIPSIS = "...";
+
+// Harness-side enforcement of the schema maxLength: the tool advertises the limit, but an
+// over-limit value is force-truncated here (prepareArguments runs before schema validation)
+// instead of failing the call.
+export function clampEvalSummary(value: unknown): string | undefined {
+	if (typeof value !== "string") return undefined;
+	const normalized = value.trim().replace(/\s+/gu, " ");
+	if (normalized.length === 0) return undefined;
+	if (normalized.length <= EVAL_SUMMARY_MAX_LENGTH) return normalized;
+	return `${normalized.slice(0, EVAL_SUMMARY_MAX_LENGTH - ELLIPSIS.length)}${ELLIPSIS}`;
+}
+
 export function parseEvalRequest(params: unknown): EvalToolRequest {
 	if (!isRecord(params)) throw new TypeError("eval parameters must be an object");
 	if (params.action === "peek" || params.action === "stop") {
@@ -14,13 +27,18 @@ export function parseEvalRequest(params: unknown): EvalToolRequest {
 		throw new TypeError(`Unknown eval action "${String(params.action)}"`);
 	if (!isEvalLanguage(params.language)) throw new TypeError("eval run requires language");
 	if (typeof params.code !== "string") throw new TypeError("eval run requires code");
+	const summary = clampEvalSummary(params.summary);
+	if (summary === undefined)
+		throw new TypeError(
+			"eval run requires summary — one line in the user's language: what this cell does and for what purpose",
+		);
 	if (params.on_timeout !== undefined && params.on_timeout !== "detach" && params.on_timeout !== "error")
 		throw new TypeError(`Unknown eval on_timeout value "${String(params.on_timeout)}"`);
 	return {
 		language: params.language,
 		code: params.code,
+		summary,
 		...(params.action === "run" ? { action: "run" as const } : {}),
-		...(typeof params.title === "string" ? { title: params.title } : {}),
 		...(typeof params.timeout === "number" ? { timeout: params.timeout } : {}),
 		...(params.on_timeout === "detach" || params.on_timeout === "error" ? { on_timeout: params.on_timeout } : {}),
 		...(typeof params.reset === "boolean" ? { reset: params.reset } : {}),

package/src/tool/eval-tool.ts

@@ -8,10 +8,16 @@ import { abortError, CellExecution, defaultTimeoutFactory } from "./cell-executi
 import { CellHandler, type CellState } from "./cell-handler.ts";
 import { EvalDetachedCellManager } from "./detached-cell-manager.ts";
 import { detachedKernelBusyError, executeEvalControl, resultAfterDetach } from "./detached-eval-result.ts";
-import { evalTimeoutBehavior, isEvalControlRequest, parseEvalRequest } from "./eval-request.ts";
+import { clampEvalSummary, evalTimeoutBehavior, isEvalControlRequest, parseEvalRequest } from "./eval-request.ts";
 import type { CreateEvalToolOptions, EvalCellInvocation } from "./eval-tool-options.ts";
 import { describeTimeoutState } from "./interrupt-note.ts";
-import { createEvalInputSchema, type EvalInputSchema, type EvalToolDetails, enabledLanguageList } from "./types.ts";
+import {
+	createEvalInputSchema,
+	type EvalInputSchema,
+	type EvalToolDetails,
+	type EvalToolRequest,
+	enabledLanguageList,
+} from "./types.ts";
 
 export type { EvalTimeoutFactory } from "./cell-execution.ts";
 export type { CreateEvalToolOptions } from "./eval-tool-options.ts";
@@ -35,6 +41,15 @@ export function createEvalTool(options: CreateEvalToolOptions): ToolDefinition<E
 		promptGuidelines: [...prompt.promptGuidelines],
 		parameters,
 		executionMode: "sequential",
+		prepareArguments: (args) => {
+			if (typeof args !== "object" || args === null) return args as EvalToolRequest;
+			const record = args as Record<string, unknown>;
+			if (record.action === "peek" || record.action === "stop") return args as EvalToolRequest;
+			const summary = clampEvalSummary(record.summary);
+			if (summary === undefined) delete record.summary;
+			else record.summary = summary;
+			return args as EvalToolRequest;
+		},
 		...(options.renderers?.renderCall === undefined ? {} : { renderCall: options.renderers.renderCall }),
 		...(options.renderers?.renderResult === undefined ? {} : { renderResult: options.renderers.renderResult }),
 		async execute(toolCallId, params, signal, onUpdate, ctx) {

package/src/tool/render.ts

@@ -255,8 +255,7 @@ function renderPrefixed(text: string, environment: RenderEnvironment, prefixStyl
 
 function cellHeader(cell: EvalCellResult, environment: RenderEnvironment, badges: CellBadges): string {
 	const presentation = cellPresentation(cell.status, environment.spinnerFrame);
-	const title = cell.title === undefined ? "" : ` ${cell.title}`;
-	let header = `eval ${cell.language}${title} ${presentation.label} ${presentation.icon}`;
+	let header = `eval ${cell.language} ${presentation.label} ${presentation.icon}`;
 	if (cell.durationMs !== undefined) header += ` · ${formatDuration(cell.durationMs)}`;
 	if (badges.reset) header += " · reset";
 	if (badges.timeout !== undefined) header += ` · timeout ${badges.timeout}s`;
@@ -511,6 +510,16 @@ function renderCell(cell: EvalCellResult, environment: RenderEnvironment, badges
 		continuation: "│  ",
 		color: "borderAccent",
 	});
+	if (cell.summary !== undefined) {
+		appendLines(
+			lines,
+			renderPrefixed(style(environment.theme, "muted", cell.summary), environment, {
+				prefix: "│ ",
+				continuation: "│ ",
+				color: "muted",
+			}),
+		);
+	}
 	const innerWidth = Math.max(1, environment.width - 2);
 	const codePreview = previewText(
 		highlightedCode(cell.code, cell.language, environment.theme),
@@ -726,7 +735,6 @@ function resultHeader(
 	status: "running" | "done" | "error",
 	theme: Theme | undefined,
 ): string {
-	const title = details?.title === undefined ? "" : ` ${details.title}`;
 	let color: ThemeColor;
 	switch (status) {
 		case "running":
@@ -739,7 +747,7 @@ function resultHeader(
 			color = "error";
 			break;
 	}
-	return style(theme, color, `eval ${details?.language ?? "?"}${title} ${status}`);
+	return style(theme, color, `eval ${details?.language ?? "?"} ${status}`);
 }
 
 function resultMetadata(
@@ -771,11 +779,11 @@ export function renderEvalCall(
 		return component;
 	}
 	if (theme === undefined && context.spinnerFrame === undefined) {
-		const title = args.title === undefined ? "" : ` ${args.title}`;
 		const reset = args.reset === true ? " reset" : "";
 		const timeout = args.timeout === undefined ? "" : ` timeout ${args.timeout}s`;
 		component.setBlocks([
-			{ kind: "text", text: style(theme, "toolTitle", `eval ${args.language}${title}${reset}${timeout}`) },
+			{ kind: "text", text: style(theme, "toolTitle", `eval ${args.language}${reset}${timeout}`) },
+			...(args.summary === undefined ? [] : [{ kind: "text" as const, text: style(theme, "muted", args.summary) }]),
 			{
 				kind: "text",
 				text: style(theme, "mdCodeBlock", args.code.trim().length > 0 ? args.code : "..."),
@@ -799,7 +807,7 @@ export function renderEvalCall(
 				};
 				const cell: EvalCellResult = {
 					index: 0,
-					...(args.title === undefined ? {} : { title: args.title }),
+					...(args.summary === undefined ? {} : { summary: args.summary }),
 					code: args.code,
 					language: args.language,
 					output: "",
@@ -850,6 +858,9 @@ export function renderEvalResult(
 	const status = resultStatus(details, options, context.isError);
 	const blocks: RenderBlock[] = [
 		{ kind: "text", text: resultHeader(details, status, theme) },
+		...(details?.summary === undefined
+			? []
+			: [{ kind: "text" as const, text: style(theme, "muted", details.summary) }]),
 		...resultMetadata(details, options, theme),
 		{ kind: "blank" },
 	];

package/src/tool/types.ts

@@ -11,11 +11,13 @@ export function enabledLanguageList(enabled: EnabledEvalLanguages): EvalLanguage
 	return evalLanguageOrder.filter((language) => enabled[language]);
 }
 
+export const EVAL_SUMMARY_MAX_LENGTH = 80;
+
 export interface EvalToolInput {
 	readonly language: EvalLanguage;
 	readonly code: string;
 	readonly action?: "run";
-	readonly title?: string;
+	readonly summary: string;
 	readonly timeout?: number;
 	readonly on_timeout?: "detach" | "error";
 	readonly reset?: boolean;
@@ -38,7 +40,13 @@ const fullEvalInputSchema = Type.Object({
 		Type.Union([Type.Literal("py"), Type.Literal("js"), Type.Literal("rb"), Type.Literal("jl")]),
 	),
 	code: Type.Optional(Type.String({ description: "Cell body, verbatim." })),
-	title: Type.Optional(Type.String({ description: "Short transcript label." })),
+	summary: Type.Optional(
+		Type.String({
+			maxLength: EVAL_SUMMARY_MAX_LENGTH,
+			description:
+				"REQUIRED for run. ONE line in the USER'S conversational language (Korean conversation -> Korean summary) stating WHAT this cell does and FOR WHAT PURPOSE; shown in the TUI while the cell runs. Longer values are force-truncated to 80 chars.",
+		}),
+	),
 	timeout: Type.Optional(Type.Number({ minimum: 1, description: "Timeout in seconds." })),
 	on_timeout: Type.Optional(
 		Type.Union([Type.Literal("detach"), Type.Literal("error")], {
@@ -68,7 +76,13 @@ export function createEvalInputSchema(enabled: EnabledEvalLanguages): EvalInputS
 			),
 			language: Type.Optional(languageSchema),
 			code: Type.Optional(Type.String({ description: "Cell body, verbatim." })),
-			title: Type.Optional(Type.String({ description: "Short transcript label." })),
+			summary: Type.Optional(
+				Type.String({
+					maxLength: EVAL_SUMMARY_MAX_LENGTH,
+					description:
+						"REQUIRED for run. ONE line in the USER'S conversational language (Korean conversation -> Korean summary) stating WHAT this cell does and FOR WHAT PURPOSE; shown in the TUI while the cell runs. Longer values are force-truncated to 80 chars.",
+				}),
+			),
 			timeout: Type.Optional(Type.Number({ minimum: 1, description: "Timeout in seconds." })),
 			on_timeout: Type.Optional(
 				Type.Union([Type.Literal("detach"), Type.Literal("error")], {
@@ -134,7 +148,7 @@ export type EvalDisplayOutput =
 
 export type EvalCellResult = {
 	readonly index: number;
-	readonly title?: string;
+	readonly summary?: string;
 	readonly code: string;
 	readonly language: EvalLanguage;
 	readonly output: string;
@@ -148,7 +162,7 @@ export type EvalCellResult = {
 export interface EvalToolDetails {
 	readonly language: EvalLanguage;
 	readonly languages?: readonly EvalLanguage[];
-	readonly title?: string;
+	readonly summary?: string;
 	readonly durationMs: number;
 	readonly toolCalls: readonly EvalToolCallSummary[];
 	readonly truncated: boolean;