@oh-my-pi/pi-coding-agent 17.0.9 → 17.1.0

package/dist/types/async/job-manager.d.ts

@@ -30,8 +30,18 @@ export interface AsyncJob {
      */
     queued?: boolean;
 }
+/** Delivery callback for a settled job's result text. */
+export type AsyncJobDeliverySink = (jobId: string, text: string, job?: AsyncJob) => void | Promise<void>;
 export interface AsyncJobManagerOptions {
-    onJobComplete: (jobId: string, text: string, job?: AsyncJob) => void | Promise<void>;
+    /**
+     * Delivery sink for UNOWNED completions (jobs registered without an
+     * `ownerId`). Owned deliveries route exclusively through
+     * {@link AsyncJobManager.registerDeliverySink}; when the owner has no live
+     * sink they are dead-lettered (dropped with a warning; the job row keeps
+     * the result text until retention eviction) — never routed here, which
+     * would leak one agent's result into another session.
+     */
+    onJobComplete?: AsyncJobDeliverySink;
     maxRunningJobs?: number;
     retentionMs?: number;
 }
@@ -120,6 +130,32 @@ export declare class AsyncJobManager {
      */
     cancelAll(filter?: AsyncJobFilter): void;
     waitForAll(): Promise<void>;
+    /**
+     * Route completions for jobs owned by `ownerId` to `sink`. Sessions register
+     * their own sink at construction and unregister on dispose. Owned deliveries
+     * with no live sink are dead-lettered — `onJobComplete` serves only unowned
+     * deliveries.
+     *
+     * Last registration wins for an owner id; the returned unregister clears the
+     * mapping only while it still points at `sink`, so a revived session's fresh
+     * registration survives its parked predecessor's late cleanup.
+     */
+    registerDeliverySink(ownerId: string, sink: AsyncJobDeliverySink): () => void;
+    /**
+     * Wait until every job owned by `ownerId` has settled — its run promise
+     * resolved, which for cancelled jobs means the underlying process actually
+     * exited. Jobs registered while waiting (e.g. by a follow-up turn) are
+     * awaited too. Returns false when `timeoutMs` elapses first.
+     *
+     * `excludeSuppressed` skips jobs whose delivery is suppressed (acknowledged
+     * or `hub`-watched): those can never re-wake a run, so quiescence barriers
+     * pass it to share one contract with the pending-async-wake predicate.
+     * Teardown reaps omit it — worktree safety concerns every owner process.
+     */
+    waitForOwnerJobs(ownerId: string, options?: {
+        timeoutMs?: number;
+        excludeSuppressed?: boolean;
+    }): Promise<boolean>;
     drainDeliveries(options?: {
         timeoutMs?: number;
         filter?: AsyncJobFilter;

package/dist/types/cli/args.d.ts

@@ -2,6 +2,8 @@ import { type ConfiguredThinkingLevel } from "../thinking.js";
 export type Mode = "text" | "json" | "rpc" | "acp" | "rpc-ui";
 export interface Args {
     cwd?: string;
+    /** Workspace directories beyond cwd for this session (repeatable `--add-dir`). */
+    addDir?: string[];
     profile?: string;
     alias?: string;
     allowHome?: boolean;

package/dist/types/cli/bench-cli.d.ts

@@ -12,6 +12,11 @@ export interface BenchCommandArgs {
         serviceTier?: string;
         json?: boolean;
         par?: number;
+        cache?: boolean;
+        cachePrefixFile?: string;
+        cachePrefixBytes?: number;
+        cachePairs?: number;
+        cacheConcurrency?: number;
     };
 }
 export interface BenchModelRegistry {
@@ -38,6 +43,36 @@ export interface BenchRunFailure {
     error: string;
 }
 export type BenchRunResult = BenchRunSuccess | BenchRunFailure;
+export type CacheObservation = "prompt_cache_read_observed" | "prompt_cache_write_observed" | "response_cache_hit_observed" | "no_provider_proof";
+export interface BenchCacheUsage {
+    inputTokens: number;
+    outputTokens: number;
+    cacheReadTokens: number;
+    cacheWriteTokens: number;
+    totalTokens: number;
+    cost: number;
+}
+export interface BenchCacheRunReport {
+    phase: "cold" | "warm";
+    result: BenchRunResult;
+    usage?: BenchCacheUsage;
+    requestIdObserved: boolean;
+    observations: CacheObservation[];
+}
+export interface BenchCachePairReport {
+    cold: BenchCacheRunReport;
+    warm: BenchCacheRunReport;
+    /** The nominal cold request showed cache reuse, so it is not a true cold baseline. */
+    coldAlreadyWarm: boolean;
+    /** Structural-only comparisons: prompt text and cache keys are never emitted. */
+    stablePrefix: true;
+    suffixChanged: true;
+    promptCacheKeyStable: true;
+    statefulResponsesDisabled: true;
+    freshProviderSessionState: true;
+    /** "unavailable" when a transport does not expose the provider payload locally. */
+    payloadStructureStable: boolean | "unavailable";
+}
 export interface BenchAverages {
     ttftMs: number;
     durationMs: number;
@@ -54,6 +89,7 @@ export interface BenchModelReport {
     results: BenchRunResult[];
     /** Averages over successful runs; null when every run failed. */
     average: BenchAverages | null;
+    cachePairs?: BenchCachePairReport[];
 }
 export interface BenchSummary {
     runs: number;
@@ -62,6 +98,10 @@ export interface BenchSummary {
     failures: number;
     /** Requested per-family service tiers, resolved per model before reaching the wire. */
     serviceTierByFamily?: ServiceTierByFamily;
+    cache?: {
+        pairs: number;
+        concurrency: number;
+    };
 }
 type BenchStreamSimple = (model: Model<Api>, context: Context, options?: SimpleStreamOptions) => AssistantMessageEventStream;
 export interface BenchDependencies {
@@ -72,6 +112,7 @@ export interface BenchDependencies {
     setExitCode?: (code: number) => void;
     streamSimple?: BenchStreamSimple;
     now?: () => number;
+    readTextFile?: (path: string, maxBytes: number) => Promise<string>;
     stdoutIsTTY?: boolean;
 }
 export declare function formatBenchTable(summary: BenchSummary): string;

package/dist/types/commands/bench.d.ts

@@ -11,11 +11,9 @@ export default class Bench extends Command {
     static flags: {
         runs: import("@oh-my-pi/pi-utils/cli").FlagDescriptor<"integer"> & {
             description: string;
-            default: number;
         };
         "max-tokens": import("@oh-my-pi/pi-utils/cli").FlagDescriptor<"integer"> & {
             description: string;
-            default: number;
         };
         prompt: import("@oh-my-pi/pi-utils/cli").FlagDescriptor<"string"> & {
             description: string;
@@ -29,7 +27,21 @@ export default class Bench extends Command {
         };
         par: import("@oh-my-pi/pi-utils/cli").FlagDescriptor<"integer"> & {
             description: string;
-            default: number;
+        };
+        cache: import("@oh-my-pi/pi-utils/cli").FlagDescriptor<"boolean"> & {
+            description: string;
+        };
+        "cache-prefix-file": import("@oh-my-pi/pi-utils/cli").FlagDescriptor<"string"> & {
+            description: string;
+        };
+        "cache-prefix-bytes": import("@oh-my-pi/pi-utils/cli").FlagDescriptor<"integer"> & {
+            description: string;
+        };
+        "cache-pairs": import("@oh-my-pi/pi-utils/cli").FlagDescriptor<"integer"> & {
+            description: string;
+        };
+        "cache-concurrency": import("@oh-my-pi/pi-utils/cli").FlagDescriptor<"integer"> & {
+            description: string;
         };
     };
     static examples: string[];

package/dist/types/commands/launch.d.ts

@@ -72,6 +72,10 @@ export default class Index extends Command {
             description: string;
             multiple: true;
         };
+        "add-dir": import("@oh-my-pi/pi-utils/cli").FlagDescriptor<"string"> & {
+            description: string;
+            multiple: true;
+        };
         print: import("@oh-my-pi/pi-utils/cli").FlagDescriptor<"boolean"> & {
             char: string;
             description: string;

package/dist/types/config/model-registry.d.ts

@@ -175,7 +175,9 @@ export declare class ModelRegistry {
     /**
      * Get API key for a model.
      */
-    getApiKey(model: Model<Api>, sessionId?: string): Promise<string | undefined>;
+    getApiKey(model: Model<Api>, sessionId?: string, options?: {
+        signal?: AbortSignal;
+    }): Promise<string | undefined>;
     /**
      * Get API key for a provider (e.g., "openai").
      *

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

@@ -250,12 +250,15 @@ export declare const ModelOverrideSchema: import("arktype/internal/variants/obje
             requiresToolResultId?: boolean | undefined;
             replayUnsignedThinking?: boolean | undefined;
         } | undefined;
+        promptCacheMode?: "automatic" | "explicit" | "none" | undefined;
+        promptCacheMinimumTokens?: number | undefined;
+        promptCacheMaximumCheckpoints?: number | undefined;
     } | undefined;
     contextPromotionTarget?: string | undefined;
     compactionModel?: string | undefined;
     remoteCompaction?: {
         enabled?: boolean | undefined;
-        api?: "anthropic-messages" | "azure-openai-responses" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
+        api?: "anthropic-messages" | "azure-openai-responses" | "bedrock-converse-stream" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
         endpoint?: string | undefined;
         model?: string | undefined;
         v2StreamingEnabled?: boolean | undefined;
@@ -275,7 +278,7 @@ export declare const ModelsConfigSchema: import("arktype/internal/variants/objec
         [x: string]: {
             baseUrl?: string | undefined;
             apiKey?: string | undefined;
-            api?: "anthropic-messages" | "azure-openai-responses" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
+            api?: "anthropic-messages" | "azure-openai-responses" | "bedrock-converse-stream" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
             headers?: {
                 [x: string]: string;
             } | undefined;
@@ -384,10 +387,13 @@ export declare const ModelsConfigSchema: import("arktype/internal/variants/objec
                     requiresToolResultId?: boolean | undefined;
                     replayUnsignedThinking?: boolean | undefined;
                 } | undefined;
+                promptCacheMode?: "automatic" | "explicit" | "none" | undefined;
+                promptCacheMinimumTokens?: number | undefined;
+                promptCacheMaximumCheckpoints?: number | undefined;
             } | undefined;
             remoteCompaction?: {
                 enabled?: boolean | undefined;
-                api?: "anthropic-messages" | "azure-openai-responses" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
+                api?: "anthropic-messages" | "azure-openai-responses" | "bedrock-converse-stream" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
                 endpoint?: string | undefined;
                 model?: string | undefined;
                 v2StreamingEnabled?: boolean | undefined;
@@ -402,7 +408,7 @@ export declare const ModelsConfigSchema: import("arktype/internal/variants/objec
             models?: {
                 id: string;
                 name?: string | undefined;
-                api?: "anthropic-messages" | "azure-openai-responses" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
+                api?: "anthropic-messages" | "azure-openai-responses" | "bedrock-converse-stream" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
                 baseUrl?: string | undefined;
                 reasoning?: boolean | undefined;
                 thinking?: ((In: {
@@ -548,12 +554,15 @@ export declare const ModelsConfigSchema: import("arktype/internal/variants/objec
                         requiresToolResultId?: boolean | undefined;
                         replayUnsignedThinking?: boolean | undefined;
                     } | undefined;
+                    promptCacheMode?: "automatic" | "explicit" | "none" | undefined;
+                    promptCacheMinimumTokens?: number | undefined;
+                    promptCacheMaximumCheckpoints?: number | undefined;
                 } | undefined;
                 contextPromotionTarget?: string | undefined;
                 compactionModel?: string | undefined;
                 remoteCompaction?: {
                     enabled?: boolean | undefined;
-                    api?: "anthropic-messages" | "azure-openai-responses" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
+                    api?: "anthropic-messages" | "azure-openai-responses" | "bedrock-converse-stream" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
                     endpoint?: string | undefined;
                     model?: string | undefined;
                     v2StreamingEnabled?: boolean | undefined;
@@ -708,12 +717,15 @@ export declare const ModelsConfigSchema: import("arktype/internal/variants/objec
                             requiresToolResultId?: boolean | undefined;
                             replayUnsignedThinking?: boolean | undefined;
                         } | undefined;
+                        promptCacheMode?: "automatic" | "explicit" | "none" | undefined;
+                        promptCacheMinimumTokens?: number | undefined;
+                        promptCacheMaximumCheckpoints?: number | undefined;
                     } | undefined;
                     contextPromotionTarget?: string | undefined;
                     compactionModel?: string | undefined;
                     remoteCompaction?: {
                         enabled?: boolean | undefined;
-                        api?: "anthropic-messages" | "azure-openai-responses" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
+                        api?: "anthropic-messages" | "azure-openai-responses" | "bedrock-converse-stream" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
                         endpoint?: string | undefined;
                         model?: string | undefined;
                         v2StreamingEnabled?: boolean | undefined;

package/dist/types/config/models-config.d.ts

@@ -32,7 +32,7 @@ export declare const ModelsConfigFile: ConfigFile<{
         [x: string]: {
             baseUrl?: string | undefined;
             apiKey?: string | undefined;
-            api?: "anthropic-messages" | "azure-openai-responses" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
+            api?: "anthropic-messages" | "azure-openai-responses" | "bedrock-converse-stream" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
             headers?: {
                 [x: string]: string;
             } | undefined;
@@ -141,10 +141,13 @@ export declare const ModelsConfigFile: ConfigFile<{
                     requiresToolResultId?: boolean | undefined;
                     replayUnsignedThinking?: boolean | undefined;
                 } | undefined;
+                promptCacheMode?: "automatic" | "explicit" | "none" | undefined;
+                promptCacheMinimumTokens?: number | undefined;
+                promptCacheMaximumCheckpoints?: number | undefined;
             } | undefined;
             remoteCompaction?: {
                 enabled?: boolean | undefined;
-                api?: "anthropic-messages" | "azure-openai-responses" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
+                api?: "anthropic-messages" | "azure-openai-responses" | "bedrock-converse-stream" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
                 endpoint?: string | undefined;
                 model?: string | undefined;
                 v2StreamingEnabled?: boolean | undefined;
@@ -159,7 +162,7 @@ export declare const ModelsConfigFile: ConfigFile<{
             models?: {
                 id: string;
                 name?: string | undefined;
-                api?: "anthropic-messages" | "azure-openai-responses" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
+                api?: "anthropic-messages" | "azure-openai-responses" | "bedrock-converse-stream" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
                 baseUrl?: string | undefined;
                 reasoning?: boolean | undefined;
                 thinking?: {
@@ -289,12 +292,15 @@ export declare const ModelsConfigFile: ConfigFile<{
                         requiresToolResultId?: boolean | undefined;
                         replayUnsignedThinking?: boolean | undefined;
                     } | undefined;
+                    promptCacheMode?: "automatic" | "explicit" | "none" | undefined;
+                    promptCacheMinimumTokens?: number | undefined;
+                    promptCacheMaximumCheckpoints?: number | undefined;
                 } | undefined;
                 contextPromotionTarget?: string | undefined;
                 compactionModel?: string | undefined;
                 remoteCompaction?: {
                     enabled?: boolean | undefined;
-                    api?: "anthropic-messages" | "azure-openai-responses" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
+                    api?: "anthropic-messages" | "azure-openai-responses" | "bedrock-converse-stream" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
                     endpoint?: string | undefined;
                     model?: string | undefined;
                     v2StreamingEnabled?: boolean | undefined;
@@ -433,12 +439,15 @@ export declare const ModelsConfigFile: ConfigFile<{
                             requiresToolResultId?: boolean | undefined;
                             replayUnsignedThinking?: boolean | undefined;
                         } | undefined;
+                        promptCacheMode?: "automatic" | "explicit" | "none" | undefined;
+                        promptCacheMinimumTokens?: number | undefined;
+                        promptCacheMaximumCheckpoints?: number | undefined;
                     } | undefined;
                     contextPromotionTarget?: string | undefined;
                     compactionModel?: string | undefined;
                     remoteCompaction?: {
                         enabled?: boolean | undefined;
-                        api?: "anthropic-messages" | "azure-openai-responses" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
+                        api?: "anthropic-messages" | "azure-openai-responses" | "bedrock-converse-stream" | "google-gemini-cli" | "google-generative-ai" | "google-vertex" | "openai-codex-responses" | "openai-completions" | "openai-responses" | undefined;
                         endpoint?: string | undefined;
                         model?: string | undefined;
                         v2StreamingEnabled?: boolean | undefined;

package/dist/types/config/provider-globals.d.ts

@@ -1,7 +1,7 @@
 interface ProviderGlobalSettings {
+    get(path: "providers.webSearchOrder"): unknown;
     get(path: "providers.webSearchExclude"): unknown;
-    get(path: "providers.webSearch"): unknown;
-    get(path: "providers.image"): unknown;
+    get(path: "providers.imageOrder"): unknown;
 }
 export declare function applyProviderGlobalsFromSettings(settings: ProviderGlobalSettings): void;
 export {};

package/dist/types/config/service-tier.d.ts

@@ -1,4 +1,4 @@
-import type { ServiceTier, ServiceTierByFamily } from "@oh-my-pi/pi-ai";
+import type { ServiceTier, ServiceTierByFamily, ServiceTierFamily } from "@oh-my-pi/pi-ai";
 import type { SubmenuOption } from "./settings-schema.js";
 /**
  * Per-family service-tier setting values. `"none"` is the omit-the-parameter
@@ -13,6 +13,10 @@ export declare const SERVICE_TIER_GOOGLE_VALUES: readonly ["none", "flex", "prio
 export type ServiceTierOpenAISettingValue = (typeof SERVICE_TIER_OPENAI_VALUES)[number];
 export type ServiceTierAnthropicSettingValue = (typeof SERVICE_TIER_ANTHROPIC_VALUES)[number];
 export type ServiceTierGoogleSettingValue = (typeof SERVICE_TIER_GOOGLE_VALUES)[number];
+/** Whether a runtime value names a provider family with an independent service-tier knob. */
+export declare function isServiceTierFamily(value: unknown): value is ServiceTierFamily;
+/** Whether a runtime value is a supported service tier for one provider family. */
+export declare function isServiceTierForFamily(family: string, tier: unknown): tier is ServiceTier;
 /**
  * Inherit-capable single value for the subagent/advisor tiers. The chosen tier
  * is broadcast across families and applied to whichever family the spawned