@oh-my-pi/pi-coding-agent 16.2.7 → 16.2.9

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

@@ -1813,6 +1813,34 @@ export declare const SETTINGS_SCHEMA: {
             }[];
         };
     };
+    readonly "stt.submitTrigger": {
+        readonly type: "enum";
+        readonly values: readonly ["never", "release", "release-complete", "say-submit"];
+        readonly default: "never";
+        readonly ui: {
+            readonly tab: "interaction";
+            readonly group: "Speech";
+            readonly label: "Speech-to-Text Submit Trigger";
+            readonly description: "Choose when speech dictation automatically submits: Never, Release (2+ words), Release with complete sentence, or When I Say Submit.";
+            readonly options: ({
+                value: "never";
+                label: string;
+                description: string;
+            } | {
+                value: "release";
+                label: string;
+                description: string;
+            } | {
+                value: "release-complete";
+                label: string;
+                description: string;
+            } | {
+                value: "say-submit";
+                label: string;
+                description: string;
+            })[];
+        };
+    };
     readonly "contextPromotion.enabled": {
         readonly type: "boolean";
         readonly default: false;
@@ -4228,7 +4256,7 @@ export declare const SETTINGS_SCHEMA: {
             readonly tab: "tasks";
             readonly group: "Subagents";
             readonly label: "Soft Subagent Request Budget";
-            readonly description: "Soft per-subagent request budget (assistant requests per run). Crossing it injects one steering notice asking the subagent to wrap up; at 1.5x the budget the run is aborted gracefully, salvaging partial output. 0 disables the guard. Bundled explore/quick_task agents use a lower built-in budget.";
+            readonly description: "Soft per-subagent request budget (assistant requests per run). Crossing it injects one steering notice asking the subagent to wrap up; at 1.5x the budget the run is aborted gracefully, salvaging partial output. 0 disables the guard. Bundled explore/sonic agents use a lower built-in budget.";
             readonly options: readonly [{
                 readonly value: "0";
                 readonly label: "Disabled";
@@ -4847,7 +4875,7 @@ export declare const SETTINGS_SCHEMA: {
     };
     readonly "providers.memoryModel": {
         readonly type: "enum";
-        readonly values: readonly ["online", "qwen3-1.7b", "gemma-3-1b", "qwen2.5-1.5b", "lfm2-1.2b"];
+        readonly values: readonly ["online", "qwen3-1.7b", "llama3.2:3b", "gemma-3-1b", "qwen2.5-1.5b", "lfm2-1.2b"];
         readonly default: "online";
         readonly ui: {
             readonly tab: "memory";
@@ -4860,15 +4888,15 @@ export declare const SETTINGS_SCHEMA: {
                 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." | "Disabled for local inference: onnxruntime-node cannot run this ONNX export's RotaryEmbedding cache updates." | "Fastest load; solid all-rounder, slightly noisier extraction labels.";
+                value: "gemma-3-1b" | "lfm2-1.2b" | "llama3.2:3b" | "qwen2.5-1.5b" | "qwen3-1.7b";
+                label: "Gemma 3 1B" | "LFM2 1.2B" | "Llama 3.2 3B" | "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." | "Disabled for local inference: onnxruntime-node cannot run this ONNX export's RotaryEmbedding cache updates." | "Fastest load; solid all-rounder, slightly noisier extraction labels." | "Larger Llama 3.2 option for local memory/classifier tasks; higher quality potential at higher disk/RAM/latency cost.";
             })[];
         };
     };
     readonly "providers.autoThinkingModel": {
         readonly type: "enum";
-        readonly values: readonly ["online", "qwen3-1.7b", "gemma-3-1b", "qwen2.5-1.5b", "lfm2-1.2b"];
+        readonly values: readonly ["online", "qwen3-1.7b", "llama3.2:3b", "gemma-3-1b", "qwen2.5-1.5b", "lfm2-1.2b"];
         readonly default: "online";
         readonly ui: {
             readonly tab: "model";
@@ -4881,9 +4909,9 @@ export declare const SETTINGS_SCHEMA: {
                 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." | "Disabled for local inference: onnxruntime-node cannot run this ONNX export's RotaryEmbedding cache updates." | "Fastest load; solid all-rounder, slightly noisier extraction labels.";
+                value: "gemma-3-1b" | "lfm2-1.2b" | "llama3.2:3b" | "qwen2.5-1.5b" | "qwen3-1.7b";
+                label: "Gemma 3 1B" | "LFM2 1.2B" | "Llama 3.2 3B" | "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." | "Disabled for local inference: onnxruntime-node cannot run this ONNX export's RotaryEmbedding cache updates." | "Fastest load; solid all-rounder, slightly noisier extraction labels." | "Larger Llama 3.2 option for local memory/classifier tasks; higher quality potential at higher disk/RAM/latency cost.";
             })[];
         };
     };
@@ -4899,7 +4927,7 @@ export declare const SETTINGS_SCHEMA: {
     };
     readonly "providers.unexpectedStopModel": {
         readonly type: "enum";
-        readonly values: readonly ["online", "qwen3-1.7b", "gemma-3-1b", "qwen2.5-1.5b", "lfm2-1.2b"];
+        readonly values: readonly ["online", "qwen3-1.7b", "llama3.2:3b", "gemma-3-1b", "qwen2.5-1.5b", "lfm2-1.2b"];
         readonly default: "online";
         readonly ui: {
             readonly tab: "providers";
@@ -4912,9 +4940,9 @@ export declare const SETTINGS_SCHEMA: {
                 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." | "Disabled for local inference: onnxruntime-node cannot run this ONNX export's RotaryEmbedding cache updates." | "Fastest load; solid all-rounder, slightly noisier extraction labels.";
+                value: "gemma-3-1b" | "lfm2-1.2b" | "llama3.2:3b" | "qwen2.5-1.5b" | "qwen3-1.7b";
+                label: "Gemma 3 1B" | "LFM2 1.2B" | "Llama 3.2 3B" | "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." | "Disabled for local inference: onnxruntime-node cannot run this ONNX export's RotaryEmbedding cache updates." | "Fastest load; solid all-rounder, slightly noisier extraction labels." | "Larger Llama 3.2 option for local memory/classifier tasks; higher quality potential at higher disk/RAM/latency cost.";
             })[];
         };
     };
@@ -4962,6 +4990,62 @@ export declare const SETTINGS_SCHEMA: {
             }];
         };
     };
+    readonly "providers.streamFirstEventTimeoutSeconds": {
+        readonly type: "number";
+        readonly default: -1;
+        readonly ui: {
+            readonly tab: "providers";
+            readonly group: "Timeouts";
+            readonly label: "Stream First Event Timeout";
+            readonly description: "Seconds to wait for the first model stream event; -1 uses provider/env defaults, 0 disables the watchdog";
+            readonly options: readonly [{
+                readonly value: "-1";
+                readonly label: "Auto";
+                readonly description: "Use provider defaults and PI_* timeout env vars";
+            }, {
+                readonly value: "0";
+                readonly label: "Off";
+                readonly description: "Disable first-event timeout";
+            }, {
+                readonly value: "300";
+                readonly label: "5 minutes";
+            }, {
+                readonly value: "600";
+                readonly label: "10 minutes";
+            }, {
+                readonly value: "1800";
+                readonly label: "30 minutes";
+            }];
+        };
+    };
+    readonly "providers.streamIdleTimeoutSeconds": {
+        readonly type: "number";
+        readonly default: -1;
+        readonly ui: {
+            readonly tab: "providers";
+            readonly group: "Timeouts";
+            readonly label: "Stream Idle Timeout";
+            readonly description: "Seconds a model stream may stay silent between events; -1 uses provider/env defaults, 0 disables the watchdog";
+            readonly options: readonly [{
+                readonly value: "-1";
+                readonly label: "Auto";
+                readonly description: "Use provider defaults and PI_* timeout env vars";
+            }, {
+                readonly value: "0";
+                readonly label: "Off";
+                readonly description: "Disable idle timeout";
+            }, {
+                readonly value: "300";
+                readonly label: "5 minutes";
+            }, {
+                readonly value: "600";
+                readonly label: "10 minutes";
+            }, {
+                readonly value: "1800";
+                readonly label: "30 minutes";
+            }];
+        };
+    };
     readonly "providers.openrouterVariant": {
         readonly type: "enum";
         readonly values: readonly ["default", "nitro", "floor", "online", "exacto"];

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

@@ -34,6 +34,7 @@ export declare class HashlineFilesystem extends Filesystem {
     canonicalPath(relativePath: string): string;
     allowTagPathRecovery(authoredPath: string, resolvedPath: string): boolean;
     readText(relativePath: string): Promise<string>;
+    readBinary(relativePath: string): Promise<Uint8Array | undefined>;
     preflightWrite(relativePath: string, options?: PreflightWriteOptions): Promise<void>;
     delete(relativePath: string): Promise<void>;
     move(fromRelative: string, toRelative: string, content?: string): Promise<void>;

package/dist/types/modes/components/mcp-add-wizard.d.ts

@@ -21,6 +21,14 @@ export interface MCPAddWizardOAuthResult {
 interface MCPAddWizardOAuthOptions {
     serverUrl?: string;
     resource?: string;
+    /**
+     * External cancellation source. Aborting it tears down the in-flight OAuth
+     * flow and surfaces a neutral cancellation error. The wizard wires its own
+     * controller here so Esc cancels the OAuth wait instead of stepping back
+     * through the form (the wizard is focused, so the editor's Esc hook does
+     * not fire).
+     */
+    abortSignal?: AbortSignal;
 }
 export declare class MCPAddWizard extends Container {
     #private;

package/dist/types/modes/controllers/mcp-command-controller.d.ts

@@ -7,6 +7,15 @@ export declare class MCPAuthorizationLinkPrompt implements Component {
     invalidate(): void;
     render(_width: number): readonly string[];
 }
+/**
+ * Thrown by {@link MCPCommandController}'s OAuth handler when the user (or a
+ * caller-supplied {@link AbortSignal}) cancels the in-flight flow. Distinct
+ * from network/timeout failures so callers can surface a neutral
+ * "cancelled" status instead of an error banner.
+ */
+export declare class MCPOAuthCancelledError extends Error {
+    constructor(message?: string);
+}
 export declare class MCPCommandController {
     #private;
     private ctx;

package/dist/types/session/agent-session.d.ts

@@ -871,7 +871,7 @@ export declare class AgentSession {
      * log but NOT to settings.
      * @throws Error if no API key available for the model
      */
-    setModelTemporary(model: Model, thinkingLevel?: ThinkingLevel, options?: {
+    setModelTemporary(model: Model, thinkingLevel?: ConfiguredThinkingLevel, options?: {
         ephemeral?: boolean;
     }): Promise<void>;
     /**

package/dist/types/session/session-context.d.ts

@@ -41,9 +41,4 @@ export interface BuildSessionContextOptions {
     /** In transcript mode, elide entries replaced by the latest compaction. */
     collapseCompactedHistory?: boolean;
 }
-/**
- * Build the session context from entries using tree traversal.
- * If leafId is provided, walks from that entry to root.
- * Handles compaction and branch summaries along the path.
- */
 export declare function buildSessionContext(entries: SessionEntry[], leafId?: string | null, byId?: Map<string, SessionEntry>, options?: BuildSessionContextOptions): SessionContext;

package/dist/types/session/settings-stream-fn.d.ts

@@ -2,8 +2,8 @@
  * Settings-aware stream wrapper shared by the main agent (sdk.ts) and the
  * advisor agent (AgentSession.#buildAdvisorRuntime).
  *
- * Reads OpenRouter / Antigravity routing variants, Responses-family text
- * verbosity, per-provider in-flight caps, and the loop guard out of `Settings`
+ * verbosity, stream watchdog budgets, per-provider in-flight caps, and the loop
+ * guard out of `Settings`
  * per request, layering them onto whatever options the caller passed. Before
  * this helper existed, advisor turns called bare `streamSimple` while the main
  * turn went through an inline closure that read these settings — so an advisor on

package/dist/types/stt/index.d.ts

@@ -3,5 +3,6 @@ export * from "./asr-protocol";
 export * from "./downloader";
 export * from "./models";
 export * from "./stt-controller";
+export * from "./submit-trigger";
 export * from "./transcriber";
 export * from "./wav";

package/dist/types/stt/stt-controller.d.ts

@@ -12,6 +12,8 @@ interface Editor {
     setVolatileText(text: string): void;
     clearVolatileText(): void;
     commitVolatileText(text: string): void;
+    submit(): void;
+    deleteBeforeCursor(count: number): void;
 }
 export declare class STTController {
     #private;

package/dist/types/stt/submit-trigger.d.ts

@@ -0,0 +1,30 @@
+/**
+ * TTS/STT Submit Trigger options and evaluation logic.
+ */
+export declare const STT_SUBMIT_TRIGGER_VALUES: readonly ["never", "release", "release-complete", "say-submit"];
+export type SttSubmitTrigger = (typeof STT_SUBMIT_TRIGGER_VALUES)[number];
+export declare const STT_SUBMIT_TRIGGER_OPTIONS: ({
+    value: "never";
+    label: string;
+    description: string;
+} | {
+    value: "release";
+    label: string;
+    description: string;
+} | {
+    value: "release-complete";
+    label: string;
+    description: string;
+} | {
+    value: "say-submit";
+    label: string;
+    description: string;
+})[];
+/**
+ * Evaluate the submit trigger against a transcribed utterance.
+ * Returns whether to submit, and the number of characters to trim from the end of the utterance.
+ */
+export declare function evaluateSubmitTrigger(utterance: string, trigger: SttSubmitTrigger): {
+    submit: boolean;
+    trimTrailing: number;
+};

package/dist/types/task/index.d.ts

@@ -22,7 +22,7 @@ export declare function formatResultOutputFallback(result: Pick<SingleResult, "o
  * Advisory — never a rejection — nudging the spawner toward tailored
  * specialists when it spawns generic role-less workers and still holds spawn
  * capacity (DepthCapacity: it currently has the `task` tool). Fires when a
- * generic `task`/`quick_task` spawn carries no `role`, or when one call clones
+ * generic `task`/`sonic` spawn carries no `role`, or when one call clones
  * the same agent ≥2× all without roles. Returns undefined when no nudge applies.
  */
 export declare function buildSpecializationAdvisory(agentName: string | undefined, items: TaskItem[], depthCapacity: boolean): string | undefined;

package/dist/types/task/types.d.ts

@@ -74,7 +74,7 @@ export interface TaskItem {
     isolated?: boolean;
 }
 export declare const taskSchema: import("arktype/internal/variants/object.ts").ObjectType<{
-    agent: string;
+    agent: import("arktype/internal/attributes.ts").Default<string, "task">;
     id?: string | undefined;
     description?: string | undefined;
     role?: string | undefined;
@@ -82,20 +82,20 @@ export declare const taskSchema: import("arktype/internal/variants/object.ts").O
     isolated?: boolean | undefined;
 }, {}>;
 declare const ALL_TASK_SCHEMAS: readonly [import("arktype/internal/variants/object.ts").ObjectType<{
-    agent: string;
+    agent: import("arktype/internal/attributes.ts").Default<string, "task">;
     id?: string | undefined;
     description?: string | undefined;
     role?: string | undefined;
     assignment: string;
     isolated?: boolean | undefined;
 }, {}>, import("arktype/internal/variants/object.ts").ObjectType<{
-    agent: string;
+    agent: import("arktype/internal/attributes.ts").Default<string, "task">;
     id?: string | undefined;
     description?: string | undefined;
     role?: string | undefined;
     assignment: string;
 }, {}>, import("arktype/internal/variants/object.ts").ObjectType<{
-    agent: string;
+    agent: import("arktype/internal/attributes.ts").Default<string, "task">;
     context: string;
     tasks: {
         id?: string | undefined;
@@ -105,7 +105,7 @@ declare const ALL_TASK_SCHEMAS: readonly [import("arktype/internal/variants/obje
         isolated?: boolean | undefined;
     }[];
 }, {}>, import("arktype/internal/variants/object.ts").ObjectType<{
-    agent: string;
+    agent: import("arktype/internal/attributes.ts").Default<string, "task">;
     context: string;
     tasks: {
         id?: string | undefined;
@@ -129,7 +129,7 @@ export declare function getTaskSchema(options: {
  * transcripts using the flat form keep working under either setting.
  */
 export interface TaskParams {
-    /** Agent type; required. */
+    /** Agent type to spawn; defaults to `"task"` (the general-purpose worker) when omitted. */
     agent?: string;
     /** Stable agent id (flat form); default = generated AdjectiveNoun. */
     id?: string;

package/dist/types/tiny/models.d.ts

@@ -84,6 +84,13 @@ export declare const TINY_MEMORY_LOCAL_MODELS: readonly [{
     readonly contextNote: "Blocked before load to avoid the unsupported RotaryEmbedding runtime path.";
     readonly reasoning: true;
     readonly unsupportedReason: "onnxruntime-node does not support Qwen3 RotaryEmbedding cache updates in onnx-community/Qwen3-1.7B-ONNX";
+}, {
+    readonly key: "llama3.2:3b";
+    readonly repo: "onnx-community/Llama-3.2-3B-Instruct-ONNX";
+    readonly dtype: "q4";
+    readonly label: "Llama 3.2 3B";
+    readonly description: "Larger Llama 3.2 option for local memory/classifier tasks; higher quality potential at higher disk/RAM/latency cost.";
+    readonly contextNote: "Use when larger model capacity is preferred over faster load times.";
 }, {
     readonly key: "gemma-3-1b";
     readonly repo: "onnx-community/gemma-3-1b-it-ONNX";
@@ -106,7 +113,7 @@ export declare const TINY_MEMORY_LOCAL_MODELS: readonly [{
     readonly description: "Fastest load; solid all-rounder, slightly noisier extraction labels.";
     readonly contextNote: "Use when local startup cost is the priority.";
 }];
-export declare const TINY_MEMORY_MODEL_VALUES: readonly ["online", "qwen3-1.7b", "gemma-3-1b", "qwen2.5-1.5b", "lfm2-1.2b"];
+export declare const TINY_MEMORY_MODEL_VALUES: readonly ["online", "qwen3-1.7b", "llama3.2:3b", "gemma-3-1b", "qwen2.5-1.5b", "lfm2-1.2b"];
 export type TinyMemoryModelKey = (typeof TINY_MEMORY_MODEL_VALUES)[number];
 export type TinyMemoryLocalModelKey = (typeof TINY_MEMORY_LOCAL_MODELS)[number]["key"];
 export declare const TINY_MEMORY_MODEL_OPTIONS: ({
@@ -114,9 +121,9 @@ export declare const TINY_MEMORY_MODEL_OPTIONS: ({
     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." | "Disabled for local inference: onnxruntime-node cannot run this ONNX export's RotaryEmbedding cache updates." | "Fastest load; solid all-rounder, slightly noisier extraction labels.";
+    value: "gemma-3-1b" | "lfm2-1.2b" | "llama3.2:3b" | "qwen2.5-1.5b" | "qwen3-1.7b";
+    label: "Gemma 3 1B" | "LFM2 1.2B" | "Llama 3.2 3B" | "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." | "Disabled for local inference: onnxruntime-node cannot run this ONNX export's RotaryEmbedding cache updates." | "Fastest load; solid all-rounder, slightly noisier extraction labels." | "Larger Llama 3.2 option for local memory/classifier tasks; higher quality potential at higher disk/RAM/latency cost.";
 })[];
 export declare function isTinyMemoryLocalModelKey(value: string): value is TinyMemoryLocalModelKey;
 export declare function getTinyMemoryModelSpec(key: TinyMemoryLocalModelKey): (typeof TINY_MEMORY_LOCAL_MODELS)[number];
@@ -173,6 +180,13 @@ export declare const TINY_LOCAL_MODELS: readonly [{
     readonly contextNote: "Blocked before load to avoid the unsupported RotaryEmbedding runtime path.";
     readonly reasoning: true;
     readonly unsupportedReason: "onnxruntime-node does not support Qwen3 RotaryEmbedding cache updates in onnx-community/Qwen3-1.7B-ONNX";
+}, {
+    readonly key: "llama3.2:3b";
+    readonly repo: "onnx-community/Llama-3.2-3B-Instruct-ONNX";
+    readonly dtype: "q4";
+    readonly label: "Llama 3.2 3B";
+    readonly description: "Larger Llama 3.2 option for local memory/classifier tasks; higher quality potential at higher disk/RAM/latency cost.";
+    readonly contextNote: "Use when larger model capacity is preferred over faster load times.";
 }, {
     readonly key: "gemma-3-1b";
     readonly repo: "onnx-community/gemma-3-1b-it-ONNX";
@@ -203,14 +217,14 @@ export declare const TINY_LOCAL_MODELS: readonly [{
  * sub-1B title models.
  */
 export declare const ONLINE_AUTO_THINKING_MODEL_KEY = "online";
-export declare const AUTO_THINKING_MODEL_VALUES: readonly ["online", "qwen3-1.7b", "gemma-3-1b", "qwen2.5-1.5b", "lfm2-1.2b"];
+export declare const AUTO_THINKING_MODEL_VALUES: readonly ["online", "qwen3-1.7b", "llama3.2:3b", "gemma-3-1b", "qwen2.5-1.5b", "lfm2-1.2b"];
 export type AutoThinkingModelKey = TinyMemoryModelKey;
 export declare const AUTO_THINKING_MODEL_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." | "Disabled for local inference: onnxruntime-node cannot run this ONNX export's RotaryEmbedding cache updates." | "Fastest load; solid all-rounder, slightly noisier extraction labels.";
+    value: "gemma-3-1b" | "lfm2-1.2b" | "llama3.2:3b" | "qwen2.5-1.5b" | "qwen3-1.7b";
+    label: "Gemma 3 1B" | "LFM2 1.2B" | "Llama 3.2 3B" | "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." | "Disabled for local inference: onnxruntime-node cannot run this ONNX export's RotaryEmbedding cache updates." | "Fastest load; solid all-rounder, slightly noisier extraction labels." | "Larger Llama 3.2 option for local memory/classifier tasks; higher quality potential at higher disk/RAM/latency cost.";
 })[];

package/dist/types/tools/output-schema-validator.d.ts

@@ -14,6 +14,16 @@ export interface OutputValidator {
     validate(value: unknown): JsonSchemaValidationResult;
     /** Top-level required property names. Empty if the schema has no `required` array at root. */
     readonly requiredFields: readonly string[];
+    /**
+     * Per-label validators for incremental yields (`type: ["<label>"]`). Each entry validates the
+     * `data` payload of a single section against the matching top-level property's sub-schema —
+     * array-typed properties (e.g. `findings`) use the items schema since each yield contributes
+     * one element, while scalar properties use the property schema directly. Unknown labels (not
+     * top-level properties) have no entry and skip per-call validation. Lets the yield tool give
+     * the model retry feedback on a section as soon as it arrives, instead of deferring every
+     * mismatch to the parent's post-mortem `schema_violation`.
+     */
+    readonly validateSection: ReadonlyMap<string, (value: unknown) => JsonSchemaValidationResult>;
 }
 export interface BuildOutputValidatorResult {
     /** Present when the schema produced a usable validator (i.e. constraining schemas). Absent for missing/unconstrained schemas. */

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-coding-agent",
-	"version": "16.2.7",
+	"version": "16.2.9",
 	"description": "Coding agent CLI with read, bash, edit, write tools and session management",
 	"homepage": "https://omp.sh",
 	"author": "Can Boluk",
@@ -55,17 +55,17 @@
 		"@agentclientprotocol/sdk": "0.25.0",
 		"@babel/parser": "^7.29.7",
 		"@mozilla/readability": "^0.6.0",
-		"@oh-my-pi/hashline": "16.2.7",
-		"@oh-my-pi/omp-stats": "16.2.7",
-		"@oh-my-pi/pi-agent-core": "16.2.7",
-		"@oh-my-pi/pi-ai": "16.2.7",
-		"@oh-my-pi/pi-catalog": "16.2.7",
-		"@oh-my-pi/pi-mnemopi": "16.2.7",
-		"@oh-my-pi/pi-natives": "16.2.7",
-		"@oh-my-pi/pi-tui": "16.2.7",
-		"@oh-my-pi/pi-utils": "16.2.7",
-		"@oh-my-pi/pi-wire": "16.2.7",
-		"@oh-my-pi/snapcompact": "16.2.7",
+		"@oh-my-pi/hashline": "16.2.9",
+		"@oh-my-pi/omp-stats": "16.2.9",
+		"@oh-my-pi/pi-agent-core": "16.2.9",
+		"@oh-my-pi/pi-ai": "16.2.9",
+		"@oh-my-pi/pi-catalog": "16.2.9",
+		"@oh-my-pi/pi-mnemopi": "16.2.9",
+		"@oh-my-pi/pi-natives": "16.2.9",
+		"@oh-my-pi/pi-tui": "16.2.9",
+		"@oh-my-pi/pi-utils": "16.2.9",
+		"@oh-my-pi/pi-wire": "16.2.9",
+		"@oh-my-pi/snapcompact": "16.2.9",
 		"@opentelemetry/api": "^1.9.1",
 		"@opentelemetry/context-async-hooks": "^2.7.1",
 		"@opentelemetry/exporter-trace-otlp-proto": "^0.218.0",

package/src/commit/agentic/agent.ts

@@ -42,7 +42,7 @@ export async function runCommitAgentSession(input: CommitAgentInput): Promise<Co
 		types_description: typesDescription,
 	});
 	const state: CommitAgentState = { diffText: input.diffText };
-	const spawns = "quick_task";
+	const spawns = "sonic";
 	const tools = createCommitTools({
 		cwd: input.cwd,
 		authStorage: input.authStorage,

package/src/commit/agentic/prompts/system.md

@@ -27,7 +27,7 @@ Tool guidance:
 - git_file_diff: diff for specific files
 - git_hunk: specific hunks for large diffs
 - recent_commits: recent commit subjects + style stats
-- analyze_files: spawn quick_task subagents in parallel for analysis
+- analyze_files: spawn sonic subagents in parallel for analysis
 - propose_changelog: provide changelog entries for each changelog target
 - propose_commit: submit final commit proposal and run validation
 - split_commit: propose multiple commit groups (no overlapping files; all staged files covered)

package/src/commit/agentic/tools/analyze-file.ts

@@ -63,7 +63,7 @@ export function createAnalyzeFileTool(options: {
 	return {
 		name: "analyze_files",
 		label: "Analyze Files",
-		description: "Spawn quick_task agents to analyze files.",
+		description: "Spawn sonic agents to analyze files.",
 		parameters: analyzeFileSchema,
 		async execute(toolCallId, params, _onUpdate, ctx, signal) {
 			const toolSession = buildToolSession(ctx, options);
@@ -83,7 +83,7 @@ export function createAnalyzeFileTool(options: {
 						related_files: relatedFiles,
 					});
 					const taskParams: TaskParams = {
-						agent: "quick_task",
+						agent: "sonic",
 						id: `AnalyzeFile${index + 1}`,
 						description: `Analyze ${file}`,
 						assignment,

package/src/config/settings-schema.ts

@@ -3,6 +3,7 @@ import { DEFAULT_SHARE_URL } from "@oh-my-pi/pi-wire";
 import { SHAPE_VARIANT_NAMES } from "@oh-my-pi/snapcompact";
 import { DEFAULT_RELAY_URL } from "../collab/protocol";
 import { DEFAULT_STT_MODEL_KEY, STT_MODEL_OPTIONS, STT_MODEL_VALUES } from "../stt/models";
+import { STT_SUBMIT_TRIGGER_OPTIONS, STT_SUBMIT_TRIGGER_VALUES } from "../stt/submit-trigger";
 import { AUTO_THINKING, getConfiguredThinkingLevelMetadata, getThinkingLevelMetadata } from "../thinking";
 import {
 	TINY_MODEL_DEVICE_DEFAULT,
@@ -144,7 +145,7 @@ export const TAB_GROUPS: Record<SettingTab, readonly string[]> = {
 		"Developer",
 	],
 	tasks: ["Modes", "Subagents", "Isolation", "Commands & Skills"],
-	providers: ["Services", "Fireworks", "Tiny Model", "Protocol", "Privacy"],
+	providers: ["Services", "Fireworks", "Tiny Model", "Protocol", "Timeouts", "Privacy"],
 };
 
 /** Status line segment identifiers */
@@ -1794,6 +1795,19 @@ export const SETTINGS_SCHEMA = {
 			options: STT_MODEL_OPTIONS,
 		},
 	},
+	"stt.submitTrigger": {
+		type: "enum",
+		values: STT_SUBMIT_TRIGGER_VALUES,
+		default: "never",
+		ui: {
+			tab: "interaction",
+			group: "Speech",
+			label: "Speech-to-Text Submit Trigger",
+			description:
+				"Choose when speech dictation automatically submits: Never, Release (2+ words), Release with complete sentence, or When I Say Submit.",
+			options: STT_SUBMIT_TRIGGER_OPTIONS,
+		},
+	},
 
 	// ────────────────────────────────────────────────────────────────────────
 	// Context
@@ -4081,7 +4095,7 @@ export const SETTINGS_SCHEMA = {
 			group: "Subagents",
 			label: "Soft Subagent Request Budget",
 			description:
-				"Soft per-subagent request budget (assistant requests per run). Crossing it injects one steering notice asking the subagent to wrap up; at 1.5x the budget the run is aborted gracefully, salvaging partial output. 0 disables the guard. Bundled explore/quick_task agents use a lower built-in budget.",
+				"Soft per-subagent request budget (assistant requests per run). Crossing it injects one steering notice asking the subagent to wrap up; at 1.5x the budget the run is aborted gracefully, salvaging partial output. 0 disables the guard. Bundled explore/sonic agents use a lower built-in budget.",
 			options: [
 				{ value: "0", label: "Disabled" },
 				{ value: "40", label: "40 requests" },
@@ -4554,6 +4568,44 @@ export const SETTINGS_SCHEMA = {
 		},
 	},
 
+	"providers.streamFirstEventTimeoutSeconds": {
+		type: "number",
+		default: -1,
+		ui: {
+			tab: "providers",
+			group: "Timeouts",
+			label: "Stream First Event Timeout",
+			description:
+				"Seconds to wait for the first model stream event; -1 uses provider/env defaults, 0 disables the watchdog",
+			options: [
+				{ value: "-1", label: "Auto", description: "Use provider defaults and PI_* timeout env vars" },
+				{ value: "0", label: "Off", description: "Disable first-event timeout" },
+				{ value: "300", label: "5 minutes" },
+				{ value: "600", label: "10 minutes" },
+				{ value: "1800", label: "30 minutes" },
+			],
+		},
+	},
+
+	"providers.streamIdleTimeoutSeconds": {
+		type: "number",
+		default: -1,
+		ui: {
+			tab: "providers",
+			group: "Timeouts",
+			label: "Stream Idle Timeout",
+			description:
+				"Seconds a model stream may stay silent between events; -1 uses provider/env defaults, 0 disables the watchdog",
+			options: [
+				{ value: "-1", label: "Auto", description: "Use provider defaults and PI_* timeout env vars" },
+				{ value: "0", label: "Off", description: "Disable idle timeout" },
+				{ value: "300", label: "5 minutes" },
+				{ value: "600", label: "10 minutes" },
+				{ value: "1800", label: "30 minutes" },
+			],
+		},
+	},
+
 	"providers.openrouterVariant": {
 		type: "enum",
 		values: ["default", "nitro", "floor", "online", "exacto"] as const,

package/src/debug/profiler.ts

@@ -114,7 +114,13 @@ function formatProfileAsMarkdown(profileJson: string): string {
  */
 export async function startCpuProfile(): Promise<ProfilerSession> {
 	const v8 = await import("node:v8");
-	v8.setFlagsFromString("--allow-natives-syntax");
+	try {
+		// Enables `%GetOptimizationStatus` and friends when V8 natives are needed
+		// for ad-hoc profiling. Best-effort: Bun does not implement
+		// `setFlagsFromString` (oven-sh/bun#1702) but the CPU profiler itself
+		// works without it, so swallow the error and continue.
+		v8.setFlagsFromString("--allow-natives-syntax");
+	} catch {}
 
 	const { Session } = await import("node:inspector/promises");
 	const session = new Session();