@oh-my-pi/pi-coding-agent 14.9.2 → 14.9.5

package/CHANGELOG.md

@@ -2,6 +2,95 @@
 
 ## [Unreleased]
 
+## [14.9.5] - 2026-05-12
+### Breaking Changes
+
+- Removed the `jobs://` internal URL protocol; inspect background jobs via the `job` tool's `list: true` operation instead
+
+### Added
+
+- Added `since` and `until` date-range filters to `search_issues`, `search_prs`, `search_commits`, and `search_repos`, accepting relative durations (`m`/`h`/`d`/`w`/`mo`/`y`), ISO dates, and ISO datetimes
+- Added `dateField` support for date filtering (`created` or `updated`) so search results can be constrained by creation, update, pushed (for repos), or committer date (for commits)
+- Added owner-based scoping to async job registration and queries so background jobs can be registered with an `ownerId` and filtered per agent in `getRunningJobs`, `getRecentJobs`, `getAllJobs`, and `cancelAll`
+- Added agent ownership metadata to async jobs started by `task` and `bash` tools so their lifecycle and cancellation is attributed to the creating agent
+- Added `list: true` operation to the `job` tool, returning an immediate snapshot of every job spawned by the calling agent without waiting (replaces the deleted `jobs://` URL)
+- Added per-agent visibility scoping to the `job` tool so `list`, `poll`, and `cancel` only see and act on jobs owned by the calling agent; cross-agent operations now return `not_found`
+
+### Changed
+
+- Changed `search_issues`, `search_prs`, `search_commits`, and `search_repos` to allow date-only queries where `query` is omitted if `since`/`until` is provided
+- Changed `search_code` to return a validation error when `since`/`until` is supplied because GitHub code search does not support date qualifiers
+- Changed async job manager ownership so subagents inherit the parent session’s global `AsyncJobManager` instead of creating and owning separate instances
+- Changed session lifecycle cleanup so the global async-job manager is disposed only by the owning top-level session
+- Changed subagent session switches and handoff paths to stop global async-job cancellation and cancel only jobs owned by that session
+- Changed `agent://` and `artifact://` URL resolution to search artifact outputs across all active sessions instead of only the current session, allowing parent and subagent sessions to read each other’s generated outputs by ID
+- Changed `memory://` URL resolution to walk all active sessions’ memory roots and return the first matching file, so worktree-based subagents can access their own memory views as well as shared roots
+- Changed internal URL routing to use a shared process-global `InternalUrlRouter` and protocol handlers, so built-in tools resolve `agent://`, `artifact://`, `memory://`, `skill://`, `rule://`, `mcp://`, and `local://` URLs without requiring session-specific router wiring
+- Changed `mcp://` handler to use the globally registered MCP manager so MCP resource links work for agents sharing session context
+
+### Changed
+
+- Changed the `ask.timeout` default from `30` (seconds) to `0` (wait indefinitely). Auto-selecting the recommended option after a fixed delay was surprising users mid-deliberation; the timer is now strictly opt-in. The legacy auto-select behavior is preserved when `ask.timeout` is set to a non-zero value, and the `ask` tool's prompt has been updated so the model expects unlimited reply time by default.
+
+### Fixed
+
+- Added `ModelRegistry.hasConfiguredAuth(model)` to mirror the upstream `@mariozechner/pi-coding-agent` API surface; external plugins and downstream wrappers that pre-flight auth before launching a subagent no longer crash with `this._modelRegistry.hasConfiguredAuth is not a function` on the direct agent-launch path. ([#993](https://github.com/can1357/oh-my-pi/issues/993))
+- Fixed an ESM circular-import TDZ that crashed test suites when modules from the `task/` and `tools/` graphs were evaluated together (e.g. `executor-warnings.test.ts` + `task-simple-mode.test.ts`) by deferring `BUILTIN_TOOLS.task`'s `TaskTool.create` dereference to factory-call time and sourcing `truncateTail` from `session/streaming-output` instead of the `tools/` barrel
+- Treat keyless-by-design providers (llama.cpp, ollama, lm-studio) as authenticated in subagent model resolution; fixes silent fallback to parent remote model when a local model is configured. ([#1008](https://github.com/can1357/oh-my-pi/issues/1008))
+- Fixed subagent disposal and session transitions that previously canceled all running async jobs, preventing inadvertent termination of a parent agent’s background work
+- Fixed multi-entry edits silently rendering a fake success when every entry failed (e.g. all hit the auto-generated guard), by surfacing `isError: true` from the single-path edit orchestrator so the renderer takes the error branch instead of falling through to the streaming-preview fallback that displays the *proposed* diff
+- Fixed the auto-generated streaming guard being gated behind `edit.streamingAbort` (default false), so it now pre-empts streaming edit tool calls targeting auto-generated files regardless of that setting
+- Fixed subagents launched in the same parallel batch not seeing each other in their initial `# IRC Peers` system-prompt block by pre-registering the agent in the global `AgentRegistry` before `rebuildSystemPrompt` runs and attaching the live session afterwards
+- Fixed plugin manifest extensions whose entry points at a directory (e.g. `pi-goal`'s `"pi": { "extensions": [".pi/extensions/pi-goal"] }`) failing to load with `Failed to load extension: Directories cannot be read like files`. The plugin path resolver now resolves directory entries to their `index.{ts,js,mjs,cjs}` file, matching the behavior of native auto-discovery via `resolveExtensionEntries`.
+- Fixed the SSH tool on native Windows by avoiding OpenSSH ControlMaster multiplexing, which Win32-OpenSSH does not support and reports as `getsockname failed` ([#154](https://github.com/can1357/oh-my-pi/issues/154)).
+- Fixed `/export` and `/tree` not showing developer-role messages (including the plan content injected after `/plan` approval) so the HTML export and TUI session tree now render developer messages dimmed with their actual content instead of hiding them entirely ([#753](https://github.com/can1357/oh-my-pi/issues/753))
+- Fixed `Timed out initializing browser tab worker` on prebuilt binaries by rewriting `spawnTabWorker` to import the worker entry with `with { type: "file" }` so Bun's `--compile` bundler statically discovers and embeds `tab-worker-entry.ts` in the single-file binary ([#1011](https://github.com/can1357/oh-my-pi/issues/1011))
+
+## [14.9.3] - 2026-05-10
+### Breaking Changes
+
+- Changed the `eval` tool input format to canonical `*** Begin <LANG>` ... `*** End <LANG>` cells with `*** Title`, `*** Timeout`, and `*** Reset` directives, so legacy `===== ... =====` eval inputs are no longer accepted for execution
+- Removed the `sectionSeparator` re-export from `config/prompt-templates`, so existing imports from `@oh-my-pi/pi-coding-agent/config/prompt-templates` now need to resolve `sectionSeparator` from its utility package
+
+### Added
+
+- Added support for the `*** Abort` recovery marker in eval and hashline parsing to terminate processing safely when stream corruption is detected
+- Added support for wrapping hashline edits in `*** Begin Patch` and `*** End Patch` markers so patch input with these envelopes is parsed and applied
+- Added support in the HTML export renderer for the new `*** Begin`/`*** End` eval cell format
+- Added a dedicated `[now]` prompt block to `buildSystemPrompt` output containing current date, current working directory, and required end-of-turn continuation/verification guidance
+- Added a new `[project]` prompt block wrapper around workstation and workspace context and ensured it is emitted as a separate system prompt segment
+- Added dedicated HTML rendering for `eval` tool calls, including cell-by-cell parsing of `===== ... =====` blocks with inferred Python/JS/TypeScript highlighting
+- Added dedicated rendering support for `search`, `recipe`, and `irc` tool calls in transcript exports
+- Added a collapsible `Available Tools` section with a tool count and chip-style compact tool names
+- Added macOS power assertion settings `power.preventIdleSleep`, `power.preventSystemSleep`, `power.declareUserActive`, and `power.preventDisplaySleep` so users can control what types of sleep are blocked during sessions
+
+### Changed
+
+- Kept legacy `===== ... =====` eval transcripts renderable in HTML while adding parsing for the new `*** Begin` format for newer transcripts
+- Changed the system prompt’s Bash usage guidance to explicitly forbid specific anti-patterns (`sed`/`awk` line-range reads, stderr redirects, and `| head|tail` pagination) and require using dedicated tools for those operations
+- Changed delegated subagent prompts so shared task context is now rendered only in the system-level `[context]` block, while the user-facing task message now contains only the assignment prompt text
+- Changed system prompt rendering to use block markers such as `[env]`, `[contract]`, `[role]`, `[coop]`, and `[closure]` for more explicit structural instructions
+- Changed the working-directory value in rendered prompts to use `shortenPath` before interpolation
+- Updated subagent prompt assembly to compose prompt blocks and place the `[now]` block after the subagent-specific instructions
+- Changed GitHub (`gh`) tool cards to include operation, PR, branch, and truncated query/title/body details
+- Changed tool-call output to display internal `_i` intent separately and hide it from rendered argument JSON
+- Changed `ast_edit` and `find`/`search` rendering to show resolved path values and option flags such as `limit`, `no-hidden`, and `no-reply`
+- Changed power assertion behavior to take effect only while a prompt is in flight, replacing session-level persistent assertions
+
+### Removed
+
+- Removed the unused `head` and `tail` parameters from the `bash` tool schema, along with the dead `normalizeBashCommand` / `applyHeadTail` post-processing module — output truncation is already handled by the harness's streaming tail buffer and artifact spillover, so the agent should rely on `read` (or the artifact link) instead of inline truncation pipes.
+
+### Fixed
+
+- Fixed eval tool outputs to append a truncation warning and ask users to re-issue remaining work when parsing is aborted by `*** Abort`
+- Fixed hashline parsing and input splitting to stop at `*** Abort` and ignore trailing edits after the marker
+- Fixed subagent task prompt construction so a trailing `[now]` block in the base prompts is preserved and not swallowed when rendering `subagent-system-prompt`
+- Fixed edit rendering so provided `input` text is shown in the export even without a file path
+- Fixed `args.paths` handling in `ast_edit` and `find` so multiple paths are shown as a comma-separated list
+- Fixed power assertion state handling so subsequent prompts are no longer blocked after an aborted or canceled prompt
+- Fixed IRC background exchange poll loop leaking after session disposal: `#scheduleBackgroundExchangeFlush` now stops immediately when `dispose()` is called, preventing stale `setTimeout` callbacks from firing against a torn-down agent
+
 ## [14.9.2] - 2026-05-10
 ### Added
 

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-coding-agent",
-	"version": "14.9.2",
+	"version": "14.9.5",
 	"description": "Coding agent CLI with read, bash, edit, write tools and session management",
 	"homepage": "https://github.com/can1357/oh-my-pi",
 	"author": "Can Boluk",
@@ -47,12 +47,12 @@
 		"@agentclientprotocol/sdk": "0.21.0",
 		"@babel/parser": "^7.29.3",
 		"@mozilla/readability": "^0.6.0",
-		"@oh-my-pi/omp-stats": "14.9.2",
-		"@oh-my-pi/pi-agent-core": "14.9.2",
-		"@oh-my-pi/pi-ai": "14.9.2",
-		"@oh-my-pi/pi-natives": "14.9.2",
-		"@oh-my-pi/pi-tui": "14.9.2",
-		"@oh-my-pi/pi-utils": "14.9.2",
+		"@oh-my-pi/omp-stats": "14.9.5",
+		"@oh-my-pi/pi-agent-core": "14.9.5",
+		"@oh-my-pi/pi-ai": "14.9.5",
+		"@oh-my-pi/pi-natives": "14.9.5",
+		"@oh-my-pi/pi-tui": "14.9.5",
+		"@oh-my-pi/pi-utils": "14.9.5",
 		"@puppeteer/browsers": "^2.13.0",
 		"@sinclair/typebox": "^0.34.49",
 		"@types/turndown": "5.0.6",

package/scripts/format-prompts.ts

@@ -16,9 +16,9 @@ import { prompt } from "@oh-my-pi/pi-utils";
  */
 import { Glob } from "bun";
 
-const PROMPTS_DIR = new URL("../src/prompts/", import.meta.url).pathname;
-const COMMIT_PROMPTS_DIR = new URL("../src/commit/prompts/", import.meta.url).pathname;
-const AGENTIC_PROMPTS_DIR = new URL("../src/commit/agentic/prompts/", import.meta.url).pathname;
+const PROMPTS_DIR = `${import.meta.dir}/../src/prompts/`;
+const COMMIT_PROMPTS_DIR = `${import.meta.dir}/../src/commit/prompts/`;
+const AGENTIC_PROMPTS_DIR = `${import.meta.dir}/../src/commit/agentic/prompts/`;
 
 const PROMPT_DIRS = [PROMPTS_DIR, COMMIT_PROMPTS_DIR, AGENTIC_PROMPTS_DIR];
 

package/src/async/job-manager.ts

@@ -16,6 +16,13 @@ export interface AsyncJob {
 	promise: Promise<void>;
 	resultText?: string;
 	errorText?: string;
+	/**
+	 * Registry id of the agent that registered the job (e.g. "0-Main",
+	 * "3-AuthLoader"). Used by scoped cancel/list APIs so a subagent's teardown
+	 * does not cancel its parent's jobs. Undefined for callers that don't
+	 * supply an id (e.g. legacy tests, SDK consumers without an agent context).
+	 */
+	ownerId?: string;
 }
 
 export interface AsyncJobManagerOptions {
@@ -41,10 +48,38 @@ export interface AsyncJobDeliveryState {
 
 export interface AsyncJobRegisterOptions {
 	id?: string;
+	/** Registry id of the agent that owns this job; used to scope cancelAll. */
+	ownerId?: string;
 	onProgress?: (text: string, details?: Record<string, unknown>) => void | Promise<void>;
 }
 
+/**
+ * Filter applied to job query/cancel APIs. With `ownerId`, results are
+ * restricted to jobs registered by that agent (registry id from
+ * `AgentRegistry`, e.g. "0-Main", "3-AuthLoader").
+ */
+export interface AsyncJobFilter {
+	ownerId?: string;
+}
+
 export class AsyncJobManager {
+	static #instance: AsyncJobManager | undefined;
+
+	/** Process-global instance shared by internal URL protocol handlers and tools. */
+	static instance(): AsyncJobManager | undefined {
+		return AsyncJobManager.#instance;
+	}
+
+	/** Install or clear the process-global instance. */
+	static setInstance(value: AsyncJobManager | undefined): void {
+		AsyncJobManager.#instance = value;
+	}
+
+	/** Reset the process-global instance. Test-only. */
+	static resetForTests(): void {
+		AsyncJobManager.#instance = undefined;
+	}
+
 	readonly #jobs = new Map<string, AsyncJob>();
 	readonly #deliveries: AsyncJobDelivery[] = [];
 	readonly #suppressedDeliveries = new Set<string>();
@@ -56,6 +91,16 @@ export class AsyncJobManager {
 	#deliveryLoop: Promise<void> | undefined;
 	#disposed = false;
 
+	#filterJobs(jobs: Iterable<AsyncJob>, filter?: AsyncJobFilter): AsyncJob[] {
+		const ownerId = filter?.ownerId;
+		if (!ownerId) return Array.from(jobs);
+		const out: AsyncJob[] = [];
+		for (const job of jobs) {
+			if (job.ownerId === ownerId) out.push(job);
+		}
+		return out;
+	}
+
 	constructor(options: AsyncJobManagerOptions) {
 		this.#onJobComplete = options.onJobComplete;
 		this.#maxRunningJobs = Math.max(1, Math.floor(options.maxRunningJobs ?? DEFAULT_MAX_RUNNING_JOBS));
@@ -95,6 +140,7 @@ export class AsyncJobManager {
 			label,
 			abortController,
 			promise: Promise.resolve(),
+			ownerId: options?.ownerId,
 		};
 
 		const reportProgress = async (text: string, details?: Record<string, unknown>): Promise<void> => {
@@ -138,9 +184,15 @@ export class AsyncJobManager {
 		return id;
 	}
 
-	cancel(id: string): boolean {
+	/**
+	 * Cancel a single job by id. When `filter.ownerId` is set and does not
+	 * match the job's owner, the call is treated as not-found (returns false)
+	 * so cross-agent cancellation is rejected at the manager level.
+	 */
+	cancel(id: string, filter?: AsyncJobFilter): boolean {
 		const job = this.#jobs.get(id);
 		if (!job) return false;
+		if (filter?.ownerId && job.ownerId !== filter.ownerId) return false;
 		if (job.status !== "running") return false;
 		job.status = "cancelled";
 		job.abortController.abort();
@@ -152,19 +204,19 @@ export class AsyncJobManager {
 		return this.#jobs.get(id);
 	}
 
-	getRunningJobs(): AsyncJob[] {
-		return Array.from(this.#jobs.values()).filter(job => job.status === "running");
+	getRunningJobs(filter?: AsyncJobFilter): AsyncJob[] {
+		return this.#filterJobs(this.#jobs.values(), filter).filter(job => job.status === "running");
 	}
 
-	getRecentJobs(limit = 10): AsyncJob[] {
-		return Array.from(this.#jobs.values())
+	getRecentJobs(limit = 10, filter?: AsyncJobFilter): AsyncJob[] {
+		return this.#filterJobs(this.#jobs.values(), filter)
 			.filter(job => job.status !== "running")
 			.sort((a, b) => b.startTime - a.startTime)
 			.slice(0, limit);
 	}
 
-	getAllJobs(): AsyncJob[] {
-		return Array.from(this.#jobs.values());
+	getAllJobs(filter?: AsyncJobFilter): AsyncJob[] {
+		return this.#filterJobs(this.#jobs.values(), filter);
 	}
 
 	getDeliveryState(): AsyncJobDeliveryState {
@@ -221,8 +273,13 @@ export class AsyncJobManager {
 		return before - this.#deliveries.length;
 	}
 
-	cancelAll(): void {
-		for (const job of this.getRunningJobs()) {
+	/**
+	 * Cancel running jobs. With `filter.ownerId` set, cancels only jobs the
+	 * matching agent registered; with no filter, cancels every running job
+	 * (used by `dispose()` to nuke the manager's state).
+	 */
+	cancelAll(filter?: AsyncJobFilter): void {
+		for (const job of this.getRunningJobs(filter)) {
 			job.status = "cancelled";
 			job.abortController.abort();
 			this.#scheduleEviction(job.id);

package/src/capability/rule.ts

@@ -209,6 +209,26 @@ export function parseRuleConditionAndScope(frontmatter: RuleFrontmatter): Pick<R
 	};
 }
 
+let activeRules: readonly Rule[] = [];
+
+/**
+ * Process-global snapshot of rules the active session loaded.
+ * Read by internal URL protocol handlers (rule://).
+ */
+export function getActiveRules(): readonly Rule[] {
+	return activeRules;
+}
+
+/** Replace the active rule snapshot. Called once per top-level session. */
+export function setActiveRules(value: readonly Rule[]): void {
+	activeRules = value;
+}
+
+/** Reset the active rule snapshot. Test-only. */
+export function resetActiveRulesForTests(): void {
+	activeRules = [];
+}
+
 export const ruleCapability = defineCapability<Rule>({
 	id: "rules",
 	displayName: "Rules",

package/src/config/model-registry.ts

@@ -2072,6 +2072,19 @@ export class ModelRegistry {
 		return this.#models.filter(model => this.#isModelAvailable(model));
 	}
 
+	/**
+	 * Check whether auth is configured for a model's provider.
+	 *
+	 * Mirrors the upstream `@mariozechner/pi-coding-agent` API surface so that
+	 * external plugins/extensions and downstream wrappers (e.g. subagent launch
+	 * paths that pre-flight auth before model resolution) can probe a model
+	 * without resolving an API key. Returns true for keyless providers as well
+	 * as providers with stored credentials. See issue #993.
+	 */
+	hasConfiguredAuth(model: Model<Api>): boolean {
+		return this.#keylessProviders.has(model.provider) || this.authStorage.hasAuth(model.provider);
+	}
+
 	getDiscoverableProviders(): string[] {
 		const disabledProviders = getDisabledProviderIdsFromSettings();
 		return this.#discoverableProviders

package/src/config/model-resolver.ts

@@ -16,7 +16,7 @@ import chalk from "chalk";
 import MODEL_PRIO from "../priority.json" with { type: "json" };
 import { parseThinkingLevel, resolveThinkingLevelForModel } from "../thinking";
 import { fuzzyMatch } from "../utils/fuzzy";
-import { isAuthenticated, MODEL_ROLE_IDS, type ModelRegistry, type ModelRole } from "./model-registry";
+import { isAuthenticated, kNoAuth, MODEL_ROLE_IDS, type ModelRegistry, type ModelRole } from "./model-registry";
 import type { Settings } from "./settings";
 
 /** Default model IDs for each known provider */
@@ -743,6 +743,12 @@ export function resolveModelOverride(
  * `modelRoles.task` pointing at an unqualified id whose only available
  * provider variant has no configured credentials — see #985).
  *
+ * Keyless-by-design providers (llama.cpp, ollama, lm-studio) advertise the
+ * `kNoAuth` sentinel from `getApiKey` to signal that they do not require
+ * credentials. Those are treated as authenticated here so an explicitly
+ * configured local model is never silently rerouted to the parent's remote
+ * provider (see #1008).
+ *
  * If neither the subagent nor the parent has working auth, returns the
  * primary resolution unchanged so the existing error path still surfaces
  * a meaningful failure downstream.
@@ -764,7 +770,7 @@ export async function resolveModelOverrideWithAuthFallback(
 	}
 
 	const primaryKey = await modelRegistry.getApiKey(primary.model);
-	if (isAuthenticated(primaryKey)) {
+	if (primaryKey === kNoAuth || isAuthenticated(primaryKey)) {
 		return { ...primary, authFallbackUsed: false };
 	}
 

package/src/config/prompt-templates.ts

@@ -30,11 +30,6 @@ prompt.registerHelper("jtdToTypeScript", (schema: unknown): string => {
 	}
 });
 
-// `sectionSeparator` + SECTION_SEPARATOR helper live in pi-utils/prompt so every
-// template consumer gets them registered without a coupling back to this module.
-// Re-exported here for call sites that already reference the coding-agent path.
-export { sectionSeparator } from "@oh-my-pi/pi-utils/prompt";
-
 function formatHashlineRef(lineNum: unknown, content: unknown): { num: number; text: string; ref: string } {
 	const num = typeof lineNum === "number" ? lineNum : Number.parseInt(String(lineNum), 10);
 	const raw = typeof content === "string" ? content : String(content ?? "");

package/src/config/settings-schema.ts

@@ -243,6 +243,44 @@ export const SETTINGS_SCHEMA = {
 			description: "Automatically resume the most recent session in the current directory",
 		},
 	},
+
+	// macOS power assertions (caffeinate flags). No-op on other platforms.
+	"power.preventIdleSleep": {
+		type: "boolean",
+		default: true,
+		ui: {
+			tab: "interaction",
+			label: "Prevent Idle Sleep (macOS)",
+			description: "caffeinate -i: keep the system awake while a session is open",
+		},
+	},
+	"power.preventSystemSleep": {
+		type: "boolean",
+		default: false,
+		ui: {
+			tab: "interaction",
+			label: "Prevent System Sleep on AC (macOS)",
+			description: "caffeinate -s: block all system sleep while on AC power",
+		},
+	},
+	"power.declareUserActive": {
+		type: "boolean",
+		default: false,
+		ui: {
+			tab: "interaction",
+			label: "Declare User Active (macOS)",
+			description: "caffeinate -u: keep the display lit and treat the user as active",
+		},
+	},
+	"power.preventDisplaySleep": {
+		type: "boolean",
+		default: false,
+		ui: {
+			tab: "interaction",
+			label: "Prevent Display Sleep (macOS)",
+			description: "caffeinate -d: keep the display from idle-sleeping while a session is open",
+		},
+	},
 	shellPath: { type: "string", default: undefined },
 
 	extensions: { type: "array", default: EMPTY_STRING_ARRAY },
@@ -867,7 +905,7 @@ export const SETTINGS_SCHEMA = {
 
 	"ask.timeout": {
 		type: "number",
-		default: 30,
+		default: 0,
 		ui: {
 			tab: "interaction",
 			label: "Ask Timeout",

package/src/edit/index.ts

@@ -204,6 +204,7 @@ async function executeSinglePathEntries(
 	const contentTexts: string[] = [];
 	const diffTexts: string[] = [];
 	let firstChangedLine: number | undefined;
+	let errorCount = 0;
 
 	for (let i = 0; i < runs.length; i++) {
 		const isLast = i === runs.length - 1;
@@ -221,6 +222,7 @@ async function executeSinglePathEntries(
 		} catch (err) {
 			const errorText = err instanceof Error ? err.message : String(err);
 			contentTexts.push(`Error editing ${path}: ${errorText}`);
+			errorCount++;
 		}
 
 		if (!isLast && onUpdate) {
@@ -230,6 +232,7 @@ async function executeSinglePathEntries(
 					diff: diffTexts.join("\n"),
 					firstChangedLine,
 				},
+				...(errorCount > 0 ? { isError: true } : {}),
 			});
 		}
 	}
@@ -240,6 +243,11 @@ async function executeSinglePathEntries(
 			diff: diffTexts.join("\n"),
 			firstChangedLine,
 		},
+		// Any per-entry failure marks the aggregate result as an error so the
+		// renderer takes the error branch instead of falling through to the
+		// streaming-edit preview (which displays the *proposed* diff and looks
+		// indistinguishable from success).
+		...(errorCount > 0 ? { isError: true } : {}),
 	};
 }
 

package/src/edit/renderer.ts

@@ -148,6 +148,8 @@ export interface EditRenderContext {
 	editDiffPreview?: DiffResult | DiffError;
 	/** Multi-file streaming diff preview (edits spanning several files) */
 	perFileDiffPreview?: PerFileDiffPreview[];
+	/** Raw in-flight edit text shown while a computed diff preview is unavailable */
+	editStreamingFallback?: string;
 	/** Function to render diff text with syntax highlighting */
 	renderDiff?: (diffText: string, options?: { filePath?: string }) => string;
 }
@@ -272,7 +274,7 @@ function formatMultiFileStreamingDiff(previews: PerFileDiffPreview[], uiTheme: T
 	const parts: string[] = [];
 	for (const preview of previews) {
 		if (!preview.diff && !preview.error) continue;
-		const header = uiTheme.fg("dim", `\n\n\u2500\u2500 ${shortenPath(preview.path)} \u2500\u2500`);
+		const header = uiTheme.fg("dim", `\n\n── ${shortenPath(preview.path)} ──`);
 		if (preview.error) {
 			parts.push(`${header}\n${uiTheme.fg("error", replaceTabs(preview.error, preview.path))}`);
 			continue;
@@ -306,6 +308,9 @@ function getCallPreview(
 	if (args.newText || args.patch) {
 		return renderPlainTextPreview(args.newText ?? args.patch ?? "", uiTheme, rawPath);
 	}
+	if (renderContext?.editStreamingFallback) {
+		return renderContext.editStreamingFallback;
+	}
 	return "";
 }
 

package/src/edit/streaming.ts

@@ -13,14 +13,19 @@
  * the injected `editMode` rather than probing argument shape.
  */
 
+import { sanitizeText } from "@oh-my-pi/pi-natives";
 import {
+	ABORT_MARKER,
+	BEGIN_PATCH_MARKER,
 	computeHashlineDiff,
 	computeHashlineSectionDiff,
 	containsRecognizableHashlineOperations,
+	END_PATCH_MARKER,
 	type HashlineInputSection,
 	splitHashlineInputs,
 } from "../hashline";
 import type { Theme } from "../modes/theme/theme";
+import { replaceTabs, truncateToWidth } from "../tools/render-utils";
 import { type EditMode, resolveEditMode } from "../utils/edit-mode";
 import { computeEditDiff, type DiffError, type DiffResult } from "./diff";
 import { type ApplyPatchEntry, expandApplyPatchToEntries, expandApplyPatchToPreviewEntries } from "./modes/apply-patch";
@@ -61,6 +66,52 @@ export interface EditStreamingStrategy<Args = unknown> {
 	renderStreamingFallback(args: Args, uiTheme: Theme): string;
 }
 
+const STREAMING_FALLBACK_LINES = 12;
+const STREAMING_FALLBACK_WIDTH = 80;
+
+function isHashlineHeaderLine(line: string): boolean {
+	const trimmed = line.trimEnd();
+	return trimmed.startsWith("@") && trimmed.length > 1;
+}
+
+function isHashlineEnvelopeMarkerLine(line: string): boolean {
+	const trimmed = line.trimEnd();
+	return trimmed === BEGIN_PATCH_MARKER || trimmed === END_PATCH_MARKER || trimmed === ABORT_MARKER;
+}
+
+function trimHashlineStreamingSyntax(lines: string[]): string[] {
+	let index = lines.findIndex(line => line.trim().length > 0);
+	if (index === -1) return [];
+
+	if (lines[index].trimEnd() === BEGIN_PATCH_MARKER) {
+		index++;
+		while (index < lines.length && lines[index].trim().length === 0) index++;
+	}
+	if (index < lines.length && isHashlineHeaderLine(lines[index])) {
+		index++;
+	}
+
+	return lines.slice(index).filter(line => !isHashlineEnvelopeMarkerLine(line));
+}
+
+function renderHashlineInputFallback(input: string, uiTheme: Theme): string {
+	const lines = trimHashlineStreamingSyntax(sanitizeText(input).split("\n"));
+	if (!lines.some(line => line.trim().length > 0)) return "";
+
+	const displayLines = lines.slice(-STREAMING_FALLBACK_LINES);
+	const hidden = lines.length - displayLines.length;
+	let text = "\n\n";
+	text += displayLines
+		.map(line => uiTheme.fg("toolOutput", truncateToWidth(replaceTabs(line), STREAMING_FALLBACK_WIDTH)))
+		.join("\n");
+	if (hidden > 0) {
+		text += uiTheme.fg("dim", `\n… (streaming +${hidden} lines)`);
+	} else {
+		text += uiTheme.fg("dim", "\n(streaming)");
+	}
+	return text;
+}
+
 // -----------------------------------------------------------------------------
 // Partial-JSON handling
 // -----------------------------------------------------------------------------
@@ -273,8 +324,8 @@ const hashlineStrategy: EditStreamingStrategy<HashlineArgs> = {
 		}
 		return previews.length > 0 ? previews : null;
 	},
-	renderStreamingFallback() {
-		return "";
+	renderStreamingFallback(args, uiTheme) {
+		return typeof args.input === "string" ? renderHashlineInputFallback(args.input, uiTheme) : "";
 	},
 };
 

package/src/eval/eval.lark

@@ -1,37 +1,16 @@
-%import common.LF
-%import common.WS_INLINE
-
-// Canonical Eval input. Each cell is introduced by a header line:
-//
-//   ===== <info> =====
-//
-// where each side is at least 5 equal signs. The info between the bars is
-// a list of space-separated tokens, all optional, in any order:
-//
-//   py | js | ts                  language for this cell
-//   py:"..." | js:"..." | ts:"..." language plus title shorthand
-//   id:"..."                      cell title (when language unchanged)
-//   t:<digits>(ms|s|m)?           per-cell timeout (default 30s)
-//   rst                           reset this language's kernel before running
-//
-// Everything between one header line and the next (or end of input) is
-// the cell's code, verbatim. The runtime additionally accepts content
-// before the first header as an implicit default-language cell, but that
-// is lenient fallback only and MUST NOT be relied on.
-
 start: cell+
-cell:  header LF code_line*
 
-header: BAR (WS_INLINE attr)+ WS_INLINE BAR
-      | BAR WS_INLINE? BAR
+cell: begin_cell attr* code_line* end_cell
+begin_cell: "*** Begin " LANG LF
+end_cell: "*** End Cell" LF?
 
-attr: LANG_TITLE | LANG | ID_ATTR | T_ATTR | RST_FLAG
+attr: title | timeout | reset
+title:   "*** Title: " /(.+)/ LF
+timeout: "*** Timeout: " /\d+(ms|s|m)?/ LF
+reset:   "*** Reset" LF
 
 code_line: /[^\r\n]*/ LF
 
-BAR:        /={5,}/
-LANG:       "py" | "js" | "ts"
-LANG_TITLE: ("py" | "js" | "ts") ":\"" /[^"\r\n]*/ "\""
-ID_ATTR:    "id:\"" /[^"\r\n]*/ "\""
-T_ATTR:     "t:" /\d+(ms|s|m)?/
-RST_FLAG:   "rst"
+LANG: "JS" | "TS" | "PY"
+
+%import common.LF

package/src/eval/index.ts

@@ -2,4 +2,5 @@ export * from "./backend";
 export { default as jsBackend } from "./js";
 export * from "./parse";
 export { default as pythonBackend } from "./py";
+export * from "./sniff";
 export * from "./types";