@oh-my-pi/pi-coding-agent 15.7.1 → 15.7.3

package/dist/types/config/settings-schema.d.ts

@@ -704,13 +704,13 @@ export declare const SETTINGS_SCHEMA: {
     };
     readonly defaultThinkingLevel: {
         readonly type: "enum";
-        readonly values: readonly import("@oh-my-pi/pi-ai").Effort[];
+        readonly values: readonly [...import("@oh-my-pi/pi-ai").Effort[], "auto"];
         readonly default: "high";
         readonly ui: {
             readonly tab: "model";
             readonly label: "Thinking Level";
             readonly description: "Reasoning depth for thinking-capable models";
-            readonly options: readonly import("../thinking").ThinkingLevelMetadata[];
+            readonly options: readonly [import("../thinking").ConfiguredThinkingLevelMetadata, ...import("../thinking").ThinkingLevelMetadata[]];
         };
     };
     readonly hideThinkingBlock: {
@@ -1298,12 +1298,12 @@ export declare const SETTINGS_SCHEMA: {
     };
     readonly "compaction.strategy": {
         readonly type: "enum";
-        readonly values: readonly ["context-full", "handoff", "off"];
+        readonly values: readonly ["context-full", "handoff", "shake", "shake-summary", "off"];
         readonly default: "context-full";
         readonly ui: {
             readonly tab: "context";
             readonly label: "Compaction Strategy";
-            readonly description: "Choose in-place context-full maintenance, auto-handoff, or disable auto maintenance (off)";
+            readonly description: "Choose in-place context-full maintenance, auto-handoff, surgical shake (drop heavy content), shake with local-model summaries, or disable auto maintenance (off)";
             readonly options: readonly [{
                 readonly value: "context-full";
                 readonly label: "Context-full";
@@ -1312,6 +1312,14 @@ export declare const SETTINGS_SCHEMA: {
                 readonly value: "handoff";
                 readonly label: "Handoff";
                 readonly description: "Generate handoff and continue in a new session";
+            }, {
+                readonly value: "shake";
+                readonly label: "Shake";
+                readonly description: "Drop heavy content (tool results + large blocks) in place; recover via artifact";
+            }, {
+                readonly value: "shake-summary";
+                readonly label: "Shake (summary)";
+                readonly description: "Shake, but compress heavy regions with a local on-device model instead of dropping";
             }, {
                 readonly value: "off";
                 readonly label: "Off";
@@ -1610,12 +1618,12 @@ export declare const SETTINGS_SCHEMA: {
     };
     readonly "memory.backend": {
         readonly type: "enum";
-        readonly values: readonly ["off", "local", "hindsight", "mnemosyne"];
+        readonly values: readonly ["off", "local", "hindsight", "mnemopi"];
         readonly default: "off";
         readonly ui: {
             readonly tab: "memory";
             readonly label: "Memory Backend";
-            readonly description: "Off, local summary pipeline, Mnemosyne SQLite, or Hindsight remote memory";
+            readonly description: "Off, local summary pipeline, Mnemopi SQLite, or Hindsight remote memory";
             readonly options: readonly [{
                 readonly value: "off";
                 readonly label: "Off";
@@ -1629,129 +1637,129 @@ export declare const SETTINGS_SCHEMA: {
                 readonly label: "Hindsight";
                 readonly description: "Vectorize Hindsight remote memory service";
             }, {
-                readonly value: "mnemosyne";
-                readonly label: "Mnemosyne";
+                readonly value: "mnemopi";
+                readonly label: "Mnemopi";
                 readonly description: "Local SQLite recall/retain backend with optional embeddings";
             }];
         };
     };
-    readonly "mnemosyne.dbPath": {
+    readonly "mnemopi.dbPath": {
         readonly type: "string";
         readonly default: undefined;
         readonly ui: {
             readonly tab: "memory";
-            readonly label: "Mnemosyne DB Path";
+            readonly label: "Mnemopi DB Path";
             readonly description: "Optional SQLite DB path. Defaults to the agent memories directory.";
-            readonly condition: "mnemosyneActive";
+            readonly condition: "mnemopiActive";
         };
     };
-    readonly "mnemosyne.bank": {
+    readonly "mnemopi.bank": {
         readonly type: "string";
         readonly default: undefined;
         readonly ui: {
             readonly tab: "memory";
-            readonly label: "Mnemosyne Bank";
+            readonly label: "Mnemopi Bank";
             readonly description: "Optional shared bank base name. Per-project modes derive project-local banks from it.";
-            readonly condition: "mnemosyneActive";
+            readonly condition: "mnemopiActive";
         };
     };
-    readonly "mnemosyne.scoping": {
+    readonly "mnemopi.scoping": {
         readonly type: "enum";
         readonly values: readonly ["global", "per-project", "per-project-tagged"];
         readonly default: "per-project";
         readonly ui: {
             readonly tab: "memory";
-            readonly label: "Mnemosyne Scoping";
+            readonly label: "Mnemopi Scoping";
             readonly description: "global = one shared bank; per-project = isolated bank per cwd; per-project-tagged = project-local writes plus global recall visibility";
             readonly options: readonly [{
                 readonly value: "global";
                 readonly label: "Global";
-                readonly description: "One shared Mnemosyne bank for every project";
+                readonly description: "One shared Mnemopi bank for every project";
             }, {
                 readonly value: "per-project";
                 readonly label: "Per project";
-                readonly description: "Project-local Mnemosyne bank per cwd basename";
+                readonly description: "Project-local Mnemopi bank per cwd basename";
             }, {
                 readonly value: "per-project-tagged";
                 readonly label: "Per project (tagged)";
                 readonly description: "Write to a project-local bank but merge project + shared recall results";
             }];
-            readonly condition: "mnemosyneActive";
+            readonly condition: "mnemopiActive";
         };
     };
-    readonly "mnemosyne.autoRecall": {
+    readonly "mnemopi.autoRecall": {
         readonly type: "boolean";
         readonly default: true;
         readonly ui: {
             readonly tab: "memory";
-            readonly label: "Mnemosyne Auto Recall";
+            readonly label: "Mnemopi Auto Recall";
             readonly description: "Recall local memories into the first turn of each session";
-            readonly condition: "mnemosyneActive";
+            readonly condition: "mnemopiActive";
         };
     };
-    readonly "mnemosyne.autoRetain": {
+    readonly "mnemopi.autoRetain": {
         readonly type: "boolean";
         readonly default: true;
         readonly ui: {
             readonly tab: "memory";
-            readonly label: "Mnemosyne Auto Retain";
-            readonly description: "Retain completed conversation turns into local Mnemosyne memory";
-            readonly condition: "mnemosyneActive";
+            readonly label: "Mnemopi Auto Retain";
+            readonly description: "Retain completed conversation turns into local Mnemopi memory";
+            readonly condition: "mnemopiActive";
         };
     };
-    readonly "mnemosyne.noEmbeddings": {
+    readonly "mnemopi.noEmbeddings": {
         readonly type: "boolean";
         readonly default: false;
         readonly ui: {
             readonly tab: "memory";
-            readonly label: "Mnemosyne Disable Embeddings";
+            readonly label: "Mnemopi Disable Embeddings";
             readonly description: "Force deterministic FTS-only recall instead of vector embeddings";
-            readonly condition: "mnemosyneActive";
+            readonly condition: "mnemopiActive";
         };
     };
-    readonly "mnemosyne.embeddingModel": {
+    readonly "mnemopi.embeddingModel": {
         readonly type: "string";
         readonly default: undefined;
         readonly ui: {
             readonly tab: "memory";
-            readonly label: "Mnemosyne Embedding Model";
-            readonly description: "Optional embedding model override passed to Mnemosyne";
-            readonly condition: "mnemosyneActive";
+            readonly label: "Mnemopi Embedding Model";
+            readonly description: "Optional embedding model override passed to Mnemopi";
+            readonly condition: "mnemopiActive";
         };
     };
-    readonly "mnemosyne.embeddingApiUrl": {
+    readonly "mnemopi.embeddingApiUrl": {
         readonly type: "string";
         readonly default: undefined;
         readonly ui: {
             readonly tab: "memory";
-            readonly label: "Mnemosyne Embedding API URL";
-            readonly description: "Optional OpenAI-compatible embedding endpoint passed to Mnemosyne";
-            readonly condition: "mnemosyneActive";
+            readonly label: "Mnemopi Embedding API URL";
+            readonly description: "Optional OpenAI-compatible embedding endpoint passed to Mnemopi";
+            readonly condition: "mnemopiActive";
         };
     };
-    readonly "mnemosyne.embeddingApiKey": {
+    readonly "mnemopi.embeddingApiKey": {
         readonly type: "string";
         readonly default: undefined;
         readonly ui: {
             readonly tab: "memory";
-            readonly label: "Mnemosyne Embedding API Key";
-            readonly description: "Optional embedding API key passed to Mnemosyne";
-            readonly condition: "mnemosyneActive";
+            readonly label: "Mnemopi Embedding API Key";
+            readonly description: "Optional embedding API key passed to Mnemopi";
+            readonly condition: "mnemopiActive";
         };
     };
-    readonly "mnemosyne.llmMode": {
+    readonly "mnemopi.llmMode": {
         readonly type: "enum";
         readonly values: readonly ["none", "smol", "remote"];
         readonly default: "smol";
         readonly ui: {
             readonly tab: "memory";
-            readonly label: "Mnemosyne LLM Mode";
+            readonly label: "Mnemopi LLM Mode";
             readonly description: "Use no LLM, the configured smol model, or a remote OpenAI-compatible endpoint";
-            readonly condition: "mnemosyneActive";
+            readonly condition: "mnemopiActive";
             readonly options: readonly [{
                 readonly value: "none";
                 readonly label: "None";
-                readonly description: "Disable Mnemosyne LLM-backed extraction";
+                readonly description: "Disable Mnemopi LLM-backed extraction";
             }, {
                 readonly value: "smol";
                 readonly label: "Smol";
@@ -1759,61 +1767,61 @@ export declare const SETTINGS_SCHEMA: {
             }, {
                 readonly value: "remote";
                 readonly label: "Remote";
-                readonly description: "Use the Mnemosyne remote LLM settings below";
+                readonly description: "Use the Mnemopi remote LLM settings below";
             }];
         };
     };
-    readonly "mnemosyne.llmBaseUrl": {
+    readonly "mnemopi.llmBaseUrl": {
         readonly type: "string";
         readonly default: undefined;
         readonly ui: {
             readonly tab: "memory";
-            readonly label: "Mnemosyne LLM Base URL";
-            readonly description: "Optional OpenAI-compatible LLM endpoint for Mnemosyne remote mode";
-            readonly condition: "mnemosyneActive";
+            readonly label: "Mnemopi LLM Base URL";
+            readonly description: "Optional OpenAI-compatible LLM endpoint for Mnemopi remote mode";
+            readonly condition: "mnemopiActive";
         };
     };
-    readonly "mnemosyne.llmApiKey": {
+    readonly "mnemopi.llmApiKey": {
         readonly type: "string";
         readonly default: undefined;
         readonly ui: {
             readonly tab: "memory";
-            readonly label: "Mnemosyne LLM API Key";
-            readonly description: "Optional LLM API key for Mnemosyne remote mode";
-            readonly condition: "mnemosyneActive";
+            readonly label: "Mnemopi LLM API Key";
+            readonly description: "Optional LLM API key for Mnemopi remote mode";
+            readonly condition: "mnemopiActive";
         };
     };
-    readonly "mnemosyne.llmModel": {
+    readonly "mnemopi.llmModel": {
         readonly type: "string";
         readonly default: undefined;
         readonly ui: {
             readonly tab: "memory";
-            readonly label: "Mnemosyne LLM Model";
-            readonly description: "Optional LLM model name for Mnemosyne remote mode";
-            readonly condition: "mnemosyneActive";
+            readonly label: "Mnemopi LLM Model";
+            readonly description: "Optional LLM model name for Mnemopi remote mode";
+            readonly condition: "mnemopiActive";
         };
     };
-    readonly "mnemosyne.retainEveryNTurns": {
+    readonly "mnemopi.retainEveryNTurns": {
         readonly type: "number";
         readonly default: 4;
     };
-    readonly "mnemosyne.recallLimit": {
+    readonly "mnemopi.recallLimit": {
         readonly type: "number";
         readonly default: 8;
     };
-    readonly "mnemosyne.recallContextTurns": {
+    readonly "mnemopi.recallContextTurns": {
         readonly type: "number";
         readonly default: 3;
     };
-    readonly "mnemosyne.recallMaxQueryChars": {
+    readonly "mnemopi.recallMaxQueryChars": {
         readonly type: "number";
         readonly default: 4000;
     };
-    readonly "mnemosyne.injectionTokenLimit": {
+    readonly "mnemopi.injectionTokenLimit": {
         readonly type: "number";
         readonly default: 5000;
     };
-    readonly "mnemosyne.debug": {
+    readonly "mnemopi.debug": {
         readonly type: "boolean";
         readonly default: false;
     };
@@ -3385,11 +3393,11 @@ export declare const SETTINGS_SCHEMA: {
         readonly ui: {
             readonly tab: "providers";
             readonly label: "Tiny Model Device";
-            readonly description: "ONNX execution provider for local tiny models (titles + memory). Default picks DirectML on Windows, CUDA on Linux x64, CPU elsewhere. The PI_TINY_DEVICE env var overrides this.";
+            readonly description: "ONNX execution provider for local tiny models (titles + memory). Default uses CPU-only inference. The PI_TINY_DEVICE env var overrides this.";
             readonly options: readonly [{
                 readonly value: "default";
                 readonly label: "Default";
-                readonly description: "DirectML on Windows, CUDA on Linux x64, CPU elsewhere";
+                readonly description: "CPU-only inference";
             }, {
                 readonly value: "gpu";
                 readonly label: "GPU";
@@ -3519,8 +3527,8 @@ export declare const SETTINGS_SCHEMA: {
         readonly ui: {
             readonly tab: "memory";
             readonly label: "Memory Model";
-            readonly description: "Mnemosyne LLM for fact extraction + consolidation: online (smol/remote) by default, or a local on-device model";
-            readonly condition: "mnemosyneActive";
+            readonly description: "Mnemopi LLM for fact extraction + consolidation: online (smol/remote) by default, or a local on-device model";
+            readonly condition: "mnemopiActive";
             readonly options: ({
                 value: "online";
                 label: string;
@@ -3532,6 +3540,41 @@ export declare const SETTINGS_SCHEMA: {
             })[];
         };
     };
+    readonly "providers.autoThinkingModel": {
+        readonly type: "enum";
+        readonly values: readonly ["online", "qwen3-1.7b", "gemma-3-1b", "qwen2.5-1.5b", "lfm2-1.2b"];
+        readonly default: "online";
+        readonly ui: {
+            readonly tab: "model";
+            readonly label: "Auto Thinking Model";
+            readonly description: "Difficulty classifier for the `auto` thinking level: online smol by default, or a local on-device model";
+            readonly condition: "autoThinkingActive";
+            readonly options: ({
+                value: "online";
+                label: string;
+                description: string;
+            } | {
+                value: "gemma-3-1b" | "lfm2-1.2b" | "qwen2.5-1.5b" | "qwen3-1.7b";
+                label: "Gemma 3 1B" | "LFM2 1.2B" | "Qwen2.5 1.5B" | "Qwen3 1.7B";
+                description: "Best consolidation/dedup; lighter footprint, but leaks small talk during extraction." | "Best extraction granularity (atomic facts); weaker consolidation." | "Fastest load; solid all-rounder, slightly noisier extraction labels." | "Recommended; most disciplined extraction (ignores chit-chat), good consolidation, about 1.1 GB cached.";
+            })[];
+        };
+    };
+    readonly "providers.shakeSummaryModel": {
+        readonly type: "enum";
+        readonly values: readonly ["qwen3-1.7b", "gemma-3-1b", "qwen2.5-1.5b", "lfm2-1.2b"];
+        readonly default: "gemma-3-1b" | "lfm2-1.2b" | "qwen2.5-1.5b" | "qwen3-1.7b";
+        readonly ui: {
+            readonly tab: "context";
+            readonly label: "Shake Summary Model";
+            readonly description: "Local on-device model used by /shake summary and the shake-summary compaction strategy to compress heavy regions. Runs entirely on-device; downloads on first use. Falls back to plain elide when unavailable.";
+            readonly options: {
+                value: "gemma-3-1b" | "lfm2-1.2b" | "qwen2.5-1.5b" | "qwen3-1.7b";
+                label: "Gemma 3 1B" | "LFM2 1.2B" | "Qwen2.5 1.5B" | "Qwen3 1.7B";
+                description: "Best consolidation/dedup; lighter footprint, but leaks small talk during extraction." | "Best extraction granularity (atomic facts); weaker consolidation." | "Fastest load; solid all-rounder, slightly noisier extraction labels." | "Recommended; most disciplined extraction (ignores chit-chat), good consolidation, about 1.1 GB cached.";
+            }[];
+        };
+    };
     readonly "providers.kimiApiFormat": {
         readonly type: "enum";
         readonly values: readonly ["openai", "anthropic"];
@@ -3825,7 +3868,7 @@ export type StatusLineSeparatorStyle = SettingValue<"statusLine.separator">;
 export type TreeFilterMode = SettingValue<"treeFilterMode">;
 export interface CompactionSettings {
     enabled: boolean;
-    strategy: "context-full" | "handoff" | "off";
+    strategy: "context-full" | "handoff" | "shake" | "shake-summary" | "off";
     thresholdPercent: number;
     thresholdTokens: number;
     reserveTokens: number;

package/dist/types/edit/hashline/diff.d.ts

@@ -5,9 +5,9 @@
  * pair to {@link generateDiffString} so the renderer can show the diff
  * while the tool call is still streaming.
  *
- * Validation is intentionally light: only the section snapshot tag is checked
- * (so the preview goes red when anchors are stale), no plan-mode guards
- * and no auto-generated-file refusal — those belong on the write path.
+ * Uses the same snapshot-tag semantics as the apply path: a live content-hash
+ * match is accepted even when the tag was minted by a source that did not keep
+ * history, and stale tags recover through the session snapshot store when possible.
  */
 import { type PatchSection, type SnapshotStore } from "@oh-my-pi/hashline";
 export interface HashlineDiffOptions {
@@ -17,6 +17,12 @@ export interface HashlineDiffOptions {
      * preview path only.
      */
     streaming?: boolean;
+    /**
+     * Skip snapshot-tag validation. Streaming previews use this so transient
+     * stale/missing tags do not flash re-read errors while the model is still
+     * authoring input; the final apply path still validates through Patcher.
+     */
+    skipHashValidation?: boolean;
 }
 export declare function computeHashlineSectionDiff(section: PatchSection, cwd: string, snapshots: SnapshotStore, options?: HashlineDiffOptions): Promise<{
     diff: string;

package/dist/types/eval/__tests__/agent-bridge.test.d.ts

@@ -0,0 +1 @@
+export {};

package/dist/types/eval/__tests__/budget-bridge.test.d.ts

@@ -0,0 +1 @@
+export {};

package/dist/types/eval/__tests__/idle-timeout.test.d.ts

@@ -0,0 +1 @@
+export {};

package/dist/types/eval/agent-bridge.d.ts

@@ -0,0 +1,25 @@
+import type { ToolSession } from "../tools";
+import type { JsStatusEvent } from "./js/shared/types";
+import "../tools/review";
+/** Synthetic bridge name reserved for the `agent()` helper across both runtimes. */
+export declare const EVAL_AGENT_BRIDGE_NAME = "__agent__";
+/** Hard recursion limit for eval-driven subagents. */
+export declare const EVAL_AGENT_MAX_DEPTH = 3;
+export interface EvalAgentBridgeOptions {
+    session: ToolSession;
+    signal?: AbortSignal;
+    emitStatus?: (event: JsStatusEvent) => void;
+}
+export interface EvalAgentResult {
+    text: string;
+    details: {
+        agent: string;
+        id: string;
+        model?: string | string[];
+        structured: boolean;
+    };
+}
+/**
+ * Run a single subagent on behalf of an eval cell's `agent()` call.
+ */
+export declare function runEvalAgent(args: unknown, options: EvalAgentBridgeOptions): Promise<EvalAgentResult>;

package/dist/types/eval/backend.d.ts

@@ -1,5 +1,5 @@
 import type { ToolSession } from "../tools";
-import type { EvalDisplayOutput, EvalLanguage } from "./types";
+import type { EvalDisplayOutput, EvalLanguage, EvalStatusEvent } from "./types";
 /** Per-cell execute() options. */
 export interface ExecutorBackendExecOptions {
     cwd: string;
@@ -8,11 +8,26 @@ export interface ExecutorBackendExecOptions {
     kernelOwnerId: string | undefined;
     signal?: AbortSignal;
     session: ToolSession;
-    deadlineMs: number;
+    /**
+     * Inactivity budget in milliseconds (the cell's `timeout`). Cancellation is
+     * driven entirely by `signal`, which the eval tool arms as an idle watchdog
+     * that fires a `TimeoutError` reason after this much time with no progress
+     * (status) events. Backends use this value only for timeout-annotation text
+     * and as cold-start headroom; they MUST NOT derive a competing wall-clock
+     * timer from it.
+     */
+    idleTimeoutMs: number;
     reset: boolean;
     artifactPath: string | undefined;
     artifactId: string | undefined;
     onChunk: (chunk: string) => void;
+    /**
+     * Live status events (read/write/agent/…) delivered as they are emitted,
+     * before the cell finishes. The same events are also returned in
+     * `displayOutputs`; this channel exists so callers can stream long-running
+     * progress (e.g. `agent()` subagents) into the UI mid-execution.
+     */
+    onStatus?: (event: EvalStatusEvent) => void;
 }
 /** Result returned by a backend's execute(). */
 export interface ExecutorBackendResult {

package/dist/types/eval/budget-bridge.d.ts

@@ -0,0 +1,29 @@
+/**
+ * Host-side handler for the eval `budget` helper.
+ *
+ * Reports the active token ceiling and amount spent so kernel helpers can
+ * compute remaining budget. Precedence: a `+Nk`/`+Nk!` per-turn directive (the
+ * user's immediate intent) wins; otherwise an active Goal Mode budget; otherwise
+ * no ceiling, with `spent` still reflecting this turn's output where available.
+ */
+import type { ToolSession } from "../tools";
+import type { JsStatusEvent } from "./js/shared/types";
+/** Synthetic bridge name reserved for the `budget` helper across both runtimes. */
+export declare const EVAL_BUDGET_BRIDGE_NAME = "__budget__";
+export interface EvalBudgetBridgeOptions {
+    session: ToolSession;
+    signal?: AbortSignal;
+    emitStatus?: (event: JsStatusEvent) => void;
+}
+export interface EvalBudgetResult {
+    total: number | null;
+    spent: number;
+    /** Whether the ceiling is enforced (eval `agent()` throws past it) vs advisory. */
+    hard: boolean;
+}
+/**
+ * Resolve the current token budget snapshot for an eval cell's `budget` helper.
+ * The returned object is JSON-passed verbatim by the bridge transport; kernel
+ * helpers read `.total`/`.spent`/`.hard` directly.
+ */
+export declare function runEvalBudget(_args: unknown, options: EvalBudgetBridgeOptions): Promise<EvalBudgetResult>;

package/dist/types/eval/idle-timeout.d.ts

@@ -0,0 +1,28 @@
+/**
+ * Inactivity watchdog for eval cells.
+ *
+ * A cell's `timeout` is treated as an *idle* budget rather than a hard
+ * wall-clock deadline: the watchdog aborts {@link signal} (with a
+ * `TimeoutError` reason, matching `AbortSignal.timeout`) only once `idleMs`
+ * elapses with no {@link bump}. Every progress signal re-arms it, so a
+ * long-running fanout that keeps reporting progress (e.g. `agent()` status
+ * updates, `log()`/`phase()`) never trips the timeout, while a genuinely
+ * stalled cell still gets interrupted.
+ *
+ * The timer self-reschedules instead of being torn down and recreated on every
+ * bump, so a high-frequency stream of bumps (sub-second agent progress) costs
+ * one timestamp write per event rather than churning a timer each time.
+ */
+export declare class IdleTimeout {
+    #private;
+    constructor(idleMs: number);
+    /** Aborts with a `TimeoutError` reason once the inactivity budget is exhausted. */
+    get signal(): AbortSignal;
+    /** Configured inactivity budget in milliseconds. */
+    get idleMs(): number;
+    /** Record activity, pushing the inactivity deadline forward by `idleMs`. */
+    bump(): void;
+    /** Stop the watchdog. Safe to call multiple times. */
+    dispose(): void;
+    [Symbol.dispose](): void;
+}

package/dist/types/eval/js/executor.d.ts

@@ -1,10 +1,18 @@
 import type { ToolSession } from "../../tools";
 import { type JsDisplayOutput } from "./context-manager";
+import type { JsStatusEvent } from "./shared/types";
 export interface JsExecutorOptions {
     cwd?: string;
     timeoutMs?: number;
     deadlineMs?: number;
+    /**
+     * Inactivity budget (ms). Used for worker cold-start headroom and
+     * timeout-annotation text when the caller drives cancellation via an
+     * idle-aware `signal` instead of `deadlineMs`/`timeoutMs`. Never arms a timer.
+     */
+    idleTimeoutMs?: number;
     onChunk?: (chunk: string) => Promise<void> | void;
+    onStatus?: (event: JsStatusEvent) => void;
     signal?: AbortSignal;
     sessionId: string;
     reset?: boolean;

package/dist/types/eval/js/tool-bridge.d.ts

@@ -1,4 +1,5 @@
 import type { ToolSession } from "../../tools";
+import { type EvalBudgetResult } from "../budget-bridge";
 import type { JsStatusEvent } from "./shared/types";
 export type { JsStatusEvent } from "./shared/types";
 interface ToolBridgeOptions {
@@ -6,7 +7,7 @@ interface ToolBridgeOptions {
     signal?: AbortSignal;
     emitStatus?: (event: JsStatusEvent) => void;
 }
-type ToolValue = string | {
+type ToolValue = string | EvalBudgetResult | {
     text: string;
     details?: unknown;
     images?: Array<{

package/dist/types/eval/py/executor.d.ts

@@ -9,6 +9,12 @@ export interface PythonExecutorOptions {
     timeoutMs?: number;
     /** Absolute wall-clock deadline in milliseconds since epoch */
     deadlineMs?: number;
+    /**
+     * Inactivity budget (ms). Used only for timeout-annotation text when the
+     * caller drives cancellation via an idle-aware `signal` instead of a
+     * wall-clock `deadlineMs`/`timeoutMs`. Does not arm a timer.
+     */
+    idleTimeoutMs?: number;
     /** Callback for streaming output chunks (already sanitized) */
     onChunk?: (chunk: string) => Promise<void> | void;
     /** AbortSignal for cancellation */
@@ -41,6 +47,13 @@ export interface PythonExecutorOptions {
     toolSession?: ToolSession;
     /** Callback for status events emitted by tool bridge invocations. */
     emitStatus?: (event: JsStatusEvent) => void;
+    /**
+     * Live status events streamed as they are emitted (both host-side bridge
+     * helpers like `agent()` and kernel-side `display`/`log`/`phase`). Mirrors
+     * what lands in `displayOutputs` so callers can render progress before the
+     * cell finishes.
+     */
+    onStatus?: (event: JsStatusEvent) => void;
     /** @internal Bridge session id, set by `executePython` before delegating. */
     bridgeSessionId?: string;
     /** @internal Bridge endpoint info, set by `executePython` before delegating. */

package/dist/types/exec/bash-executor.d.ts

@@ -4,6 +4,7 @@ export interface BashExecutorOptions {
     cwd?: string;
     timeout?: number;
     onChunk?: (chunk: string) => void;
+    chunkThrottleMs?: number;
     signal?: AbortSignal;
     /** Session key suffix to isolate shell sessions per agent */
     sessionKey?: string;

package/dist/types/extensibility/custom-tools/types.d.ts

@@ -86,10 +86,10 @@ export type CustomToolSessionEvent = {
 } | {
     reason: "auto_compaction_start";
     trigger: "threshold" | "overflow" | "idle" | "incomplete";
-    action: "context-full" | "handoff";
+    action: "context-full" | "handoff" | "shake" | "shake-summary";
 } | {
     reason: "auto_compaction_end";
-    action: "context-full" | "handoff";
+    action: "context-full" | "handoff" | "shake" | "shake-summary";
     result: CompactionResult | undefined;
     aborted: boolean;
     willRetry: boolean;

package/dist/types/extensibility/extensions/runner.d.ts

@@ -83,6 +83,13 @@ export declare class ExtensionRunner {
     getExtensionPaths(): string[];
     /** Get all registered tools from all extensions. */
     getAllRegisteredTools(): RegisteredTool[];
+    /**
+     * Aggregate the registered CLI flags across a set of extensions (last write
+     * wins on name collision). Static so callers that need the flag set before a
+     * runner exists — e.g. the CLI resolving `@file`/flag args before session
+     * creation — share this exact logic instead of duplicating it.
+     */
+    static aggregateFlags(extensions: readonly Extension[]): Map<string, ExtensionFlag>;
     getFlags(): Map<string, ExtensionFlag>;
     getFlagValues(): Map<string, boolean | string>;
     setFlagValue(name: string, value: boolean | string): void;