@oh-my-pi/pi-coding-agent 15.1.2 → 15.1.3

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

@@ -6,8 +6,32 @@ import type { RenderResultOptions } from "../extensibility/custom-tools/types";
 import { type Theme } from "../modes/theme/theme";
 import { type ToolSession } from ".";
 export declare const EVAL_DEFAULT_PREVIEW_LINES = 10;
+/**
+ * Per-cell input. Each cell runs in order; state persists within a language
+ * across cells and across tool calls.
+ */
+declare const evalCellSchema: z.ZodObject<{
+    language: z.ZodEnum<{
+        js: "js";
+        py: "py";
+    }>;
+    code: z.ZodString;
+    title: z.ZodOptional<z.ZodString>;
+    timeout: z.ZodOptional<z.ZodNumber>;
+    reset: z.ZodOptional<z.ZodBoolean>;
+}, z.core.$strip>;
+export type EvalCellInput = z.infer<typeof evalCellSchema>;
 export declare const evalSchema: z.ZodObject<{
-    input: z.ZodString;
+    cells: z.ZodArray<z.ZodObject<{
+        language: z.ZodEnum<{
+            js: "js";
+            py: "py";
+        }>;
+        code: z.ZodString;
+        title: z.ZodOptional<z.ZodString>;
+        timeout: z.ZodOptional<z.ZodNumber>;
+        reset: z.ZodOptional<z.ZodBoolean>;
+    }, z.core.$strip>>;
 }, z.core.$strip>;
 export type EvalToolParams = z.infer<typeof evalSchema>;
 export type EvalToolResult = {
@@ -35,20 +59,30 @@ export declare class EvalTool implements AgentTool<typeof evalSchema> {
     readonly label = "Eval";
     get description(): string;
     readonly parameters: z.ZodObject<{
-        input: z.ZodString;
+        cells: z.ZodArray<z.ZodObject<{
+            language: z.ZodEnum<{
+                js: "js";
+                py: "py";
+            }>;
+            code: z.ZodString;
+            title: z.ZodOptional<z.ZodString>;
+            timeout: z.ZodOptional<z.ZodNumber>;
+            reset: z.ZodOptional<z.ZodBoolean>;
+        }, z.core.$strip>>;
     }, z.core.$strip>;
     readonly concurrency = "exclusive";
     readonly strict = true;
     readonly intent: (args: Partial<z.infer<typeof evalSchema>>) => string | undefined;
-    get customFormat(): {
-        syntax: "lark";
-        definition: string;
-    };
     constructor(session: ToolSession | null, options?: EvalToolOptions);
     execute(_toolCallId: string, params: z.infer<typeof evalSchema>, signal?: AbortSignal, onUpdate?: AgentToolUpdateCallback, _ctx?: AgentToolContext): Promise<AgentToolResult<EvalToolDetails | undefined>>;
 }
+interface EvalRenderCellArg {
+    language?: string;
+    code?: string;
+    title?: string;
+}
 interface EvalRenderArgs {
-    input?: string;
+    cells?: EvalRenderCellArg[];
     __partialJson?: string;
 }
 interface EvalRenderContext {

package/dist/types/tools/irc.d.ts

@@ -21,7 +21,10 @@ import type { AgentTool, AgentToolContext, AgentToolResult, AgentToolUpdateCallb
 import * as z from "zod/v4";
 import type { ToolSession } from ".";
 declare const ircSchema: z.ZodObject<{
-    op: z.ZodUnion<readonly [z.ZodLiteral<"send">, z.ZodLiteral<"list">]>;
+    op: z.ZodEnum<{
+        list: "list";
+        send: "send";
+    }>;
     to: z.ZodOptional<z.ZodString>;
     message: z.ZodOptional<z.ZodString>;
     awaitReply: z.ZodOptional<z.ZodBoolean>;
@@ -59,7 +62,10 @@ export declare class IrcTool implements AgentTool<typeof ircSchema, IrcDetails>
     readonly summary = "Send and receive messages between agents over IRC-like channels";
     readonly description: string;
     readonly parameters: z.ZodObject<{
-        op: z.ZodUnion<readonly [z.ZodLiteral<"send">, z.ZodLiteral<"list">]>;
+        op: z.ZodEnum<{
+            list: "list";
+            send: "send";
+        }>;
         to: z.ZodOptional<z.ZodString>;
         message: z.ZodOptional<z.ZodString>;
         awaitReply: z.ZodOptional<z.ZodBoolean>;

package/dist/types/tools/report-tool-issue.d.ts

@@ -1,6 +1,123 @@
+/**
+ * report_tool_issue — automated QA tool for tracking unexpected tool behavior.
+ *
+ * Enabled by default; gated behind PI_AUTO_QA=1 / `dev.autoqa` so a user
+ * who flips the setting off short-circuits injection entirely.
+ * Always injected into every agent (including subagents) regardless of tool selection.
+ * Records grievances to a local SQLite database; never throws.
+ *
+ * Before the first record lands, the user's consent is checked. If they've
+ * never been asked (`dev.autoqa.consent === "unset"`) the process-global
+ * consent handler — wired by `InteractiveMode` to a Yes/No popup — is
+ * invoked exactly once and the decision is persisted. Subsequent calls
+ * (including from subagents) read the cached decision without prompting.
+ *
+ * When the user grants consent, push is automatically active against the
+ * bundled endpoint (`dev.autoqaPush.endpoint`, default `qa.omp.sh`). Each
+ * insert schedules a background flush that POSTs pending rows and deletes
+ * them on HTTP 2xx. `PI_AUTO_QA_PUSH=1` forces push in non-interactive
+ * environments where the consent dialog never fires. Tool execution is
+ * never blocked on the network and never throws.
+ */
+import { Database } from "bun:sqlite";
 import type { AgentTool } from "@oh-my-pi/pi-agent-core";
 import type { Settings } from "..";
 import type { ToolSession } from "./index";
 export declare function isAutoQaEnabled(settings?: Settings): boolean;
+/**
+ * Resolver for the user's "share grievances?" consent.
+ *
+ * Return values:
+ *   - `true`  — user agreed; record + ship for this run and persist.
+ *   - `false` — user declined; suppress for this run and persist.
+ *   - `null`  — user dismissed the dialog (ESC, click-away, …) without
+ *               picking an option. The decision is NOT cached or persisted,
+ *               so the next `report_tool_issue` invocation re-prompts.
+ *
+ * Persistence is the tool's job (so subagent invocations can persist into
+ * the disk-backed `Settings` instance the host registered alongside the
+ * handler), not the handler's. Implementations live in hosts that have UI
+ * affordances — today only `InteractiveMode`. When no handler is
+ * registered (CLI subcommands, tests, non-interactive runs) consent
+ * defaults to `false` — the explicit "don't collect by default" stance.
+ */
+export type AutoQaConsentHandler = () => Promise<boolean | null>;
+/**
+ * Register the consent handler and the persistent {@link Settings} instance
+ * the decision should be written to. Passing `null` clears the handler
+ * (e.g. on `InteractiveMode` teardown). Re-registration is authoritative.
+ */
+export declare function setAutoQaConsentHandler(handler: AutoQaConsentHandler | null, persistentSettings?: Settings | null): void;
+/** Test-only: clear consent cache + handler. Never call from production code. */
+export declare function __resetAutoQaConsentForTests(): void;
+/**
+ * Resolve the user's consent for `report_tool_issue` grievances.
+ *
+ * Precedence (highest first):
+ *   1. Process-global cache (set on first successful resolution).
+ *   2. Persistent setting (`dev.autoqa.consent` on the supplied `Settings`).
+ *   3. Persistent setting on the registered host `Settings`.
+ *   4. Consent handler popup (single-flight; persists the answer).
+ *   5. Default-deny when no handler is registered.
+ *
+ * Never throws — handler errors degrade to "denied for this call" without
+ * caching, so a subsequent invocation can re-prompt instead of being
+ * permanently locked into the false branch.
+ */
+export declare function resolveAutoQaConsent(settings: Settings | undefined): Promise<boolean>;
 export declare function getAutoQaDbPath(): string;
-export declare function createReportToolIssueTool(session: ToolSession): AgentTool;
+/**
+ * Open (or return the cached handle for) the auto-QA SQLite database at
+ * `~/.omp/agent/autoqa.db`. Idempotently runs schema creation, the
+ * `pushed`-column migration, and index setup so every consumer — tool
+ * execute path, manual `omp grievances push`, future debug scripts —
+ * sees the same prepared schema. Returns `null` only on a hard open
+ * failure (filesystem permissions, etc.); a missing file is created.
+ *
+ * Exported because the `omp grievances` CLI handlers need the migrated
+ * handle too — having a second `openDb` in the CLI led to the column
+ * never being added on the manual-push path.
+ */
+export declare function openAutoQaDb(): Database | null;
+export interface FlushResult {
+    pushed: number;
+    ok: boolean;
+    skipped?: boolean;
+}
+/**
+ * Optional per-flush controls. Used by `omp grievances push` to surface
+ * progress to a TTY and to skip the user-facing consent gate (manual
+ * pushes are the user's explicit intent, not a side effect of a tool call).
+ */
+export interface FlushOptions {
+    /**
+     * Skip the `dev.autoqa.consent === "granted"` gate in
+     * {@link resolvePushConfig}. Endpoint configuration is still required.
+     * Reserved for explicit user-driven pushes (CLI `grievances push`,
+     * future debug recipes); never set from the tool's auto-flush path.
+     */
+    bypassConsent?: boolean;
+    /**
+     * Fires once at the start of the loop with the snapshot count of
+     * unpushed rows. Subsequent inserts won't be reflected (the count is
+     * a planning hint for progress reporters, not a live total).
+     */
+    onStart?: (totalUnpushed: number) => void;
+    /**
+     * Fires after every successfully shipped batch with the running pushed
+     * count. Reporters compare against the `totalUnpushed` they saw in
+     * `onStart` to advance their bar.
+     */
+    onProgress?: (pushedSoFar: number) => void;
+}
+/** Test-only: clear single-flight + cooldown state. Never call from production code. */
+export declare function __resetAutoQaFlushStateForTests(): void;
+/**
+ * Flush queued grievances to the configured backend.
+ *
+ * Single-flight: concurrent callers share the in-flight promise. After a
+ * failed push, retries are skipped for {@link FAILURE_COOLDOWN_MS} ms.
+ * Never throws — all errors are caught and routed to the logger.
+ */
+export declare function flushGrievances(db?: Database, settings?: Settings, options?: FlushOptions): Promise<FlushResult>;
+export declare function createReportToolIssueTool(session: ToolSession, activeBuiltinNames?: readonly string[]): AgentTool;

package/dist/types/tools/resolve.d.ts

@@ -5,7 +5,10 @@ import type { RenderResultOptions } from "../extensibility/custom-tools/types";
 import type { Theme } from "../modes/theme/theme";
 import type { ToolSession } from ".";
 declare const resolveSchema: z.ZodObject<{
-    action: z.ZodUnion<readonly [z.ZodLiteral<"apply">, z.ZodLiteral<"discard">]>;
+    action: z.ZodEnum<{
+        apply: "apply";
+        discard: "discard";
+    }>;
     reason: z.ZodString;
     extra: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
 }, z.core.$strip>;
@@ -54,7 +57,10 @@ export declare class ResolveTool implements AgentTool<typeof resolveSchema, Reso
     readonly hidden = true;
     readonly description: string;
     readonly parameters: z.ZodObject<{
-        action: z.ZodUnion<readonly [z.ZodLiteral<"apply">, z.ZodLiteral<"discard">]>;
+        action: z.ZodEnum<{
+            apply: "apply";
+            discard: "discard";
+        }>;
         reason: z.ZodString;
         extra: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
     }, z.core.$strip>;

package/examples/custom-tools/README.md

@@ -47,7 +47,6 @@ See [docs/custom-tools.md](../../docs/custom-tools.md) for full documentation.
 **Factory pattern:**
 
 ```typescript
-import { StringEnum } from "@oh-my-pi/pi-ai";
 import { Text } from "@oh-my-pi/pi-tui";
 import type { CustomToolFactory } from "@oh-my-pi/pi-coding-agent";
 
@@ -56,7 +55,7 @@ const factory: CustomToolFactory = (pi) => ({
 	label: "My Tool",
 	description: "Tool description for LLM",
 	parameters: pi.zod.object({
-		action: StringEnum(["list", "add"] as const),
+		action: pi.zod.enum(["list", "add"]),
 	}),
 
 	// Called on session start/switch/branch/clear
@@ -76,9 +75,6 @@ const factory: CustomToolFactory = (pi) => ({
 
 export default factory;
 ```
-
-**Legacy:** `parameters: pi.typebox.Type.Object({ ... })` still works; the injected `typebox` is a small Zod-backed shim, and schemas flow through the same Zod pipeline as `pi.zod` schemas.
-
 **Custom rendering:**
 
 ```typescript
@@ -97,17 +93,12 @@ renderResult(result, { expanded, isPartial }, theme) {
 },
 ```
 
-**Use `StringEnum` for discriminated string tool args** (required for Google API compatibility):
+**Use `z.enum` for discriminated string tool args:**
 
 ```typescript
-import { StringEnum } from "@oh-my-pi/pi-ai";
-
 const { z } = pi.zod;
 
-// Good — Google-safe enum wiring
 parameters: z.object({
-	action: StringEnum(["list", "add"] as const),
+	action: z.enum(["list", "add"]),
 });
-
-// Avoid raw union-of-literals patterns that don't degrade well for strict JSON Schema providers
 ```

package/examples/extensions/README.md

@@ -108,29 +108,16 @@ export default function (pi: ExtensionAPI) {
 	});
 }
 ```
-
-**Legacy TypeBox-style schemas** (`pi.typebox`) remain available for older extensions and are backed by a tiny Zod-shim — prefer `pi.zod` directly for new code.
-
-```typescript
-const { Type } = pi.typebox;
-parameters: Type.Object({ name: Type.String() });
-```
-
 ## Key Patterns
 
-**Use `StringEnum` for discriminated string tool args** (required for Google API compatibility):
+**Use `z.enum` for discriminated string tool args:**
 
 ```typescript
-import { StringEnum } from "@oh-my-pi/pi-ai";
-
 const { z } = pi.zod;
 
-// Good — Google-safe enum wiring
 parameters: z.object({
-	action: StringEnum(["list", "add"] as const),
+	action: z.enum(["list", "add"]),
 });
-
-// Avoid raw union-of-literals patterns that don't degrade well for strict JSON Schema providers
 ```
 
 **State persistence via details:**

package/examples/extensions/api-demo.ts

@@ -10,9 +10,6 @@ import type { ExtensionAPI } from "@oh-my-pi/pi-coding-agent";
 export default function (pi: ExtensionAPI) {
 	const { z } = pi.zod;
 
-	// Access shared schema helpers from package exports (e.g. StringEnum for Google-safe enums)
-	const { StringEnum } = pi.pi;
-
 	// Access the logger for debugging
 	pi.logger.debug("API demo extension loaded");
 
@@ -22,10 +19,7 @@ export default function (pi: ExtensionAPI) {
 		description: "Demonstrates ExtensionAPI capabilities: logger, zod, and pi module access",
 		parameters: z.object({
 			message: z.string().describe("Test message"),
-			logLevel: StringEnum(["error", "warn", "debug"], {
-				description: "Log level to use",
-				default: "debug",
-			}),
+			logLevel: z.enum(["error", "warn", "debug"]).default("debug").describe("Log level to use"),
 		}),
 
 		async execute(_toolCallId, params, _onUpdate, ctx, _signal) {

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/pi-coding-agent",
-	"version": "15.1.2",
+	"version": "15.1.3",
 	"description": "Coding agent CLI with read, bash, edit, write tools and session management",
 	"homepage": "https://omp.sh",
 	"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": "15.1.2",
-		"@oh-my-pi/pi-agent-core": "15.1.2",
-		"@oh-my-pi/pi-ai": "15.1.2",
-		"@oh-my-pi/pi-natives": "15.1.2",
-		"@oh-my-pi/pi-tui": "15.1.2",
-		"@oh-my-pi/pi-utils": "15.1.2",
+		"@oh-my-pi/omp-stats": "15.1.3",
+		"@oh-my-pi/pi-agent-core": "15.1.3",
+		"@oh-my-pi/pi-ai": "15.1.3",
+		"@oh-my-pi/pi-natives": "15.1.3",
+		"@oh-my-pi/pi-tui": "15.1.3",
+		"@oh-my-pi/pi-utils": "15.1.3",
 		"@puppeteer/browsers": "^2.13.0",
 		"@types/turndown": "5.0.6",
 		"@xterm/headless": "^6.0.0",

package/src/autoresearch/tools/init-experiment.ts

@@ -17,42 +17,20 @@ export const DEFAULT_HARNESS_COMMAND = `bash ${HARNESS_FILENAME}`;
 const HARNESS_COMMIT_TITLE = "autoresearch: harness setup";
 
 const initExperimentSchema = z.object({
-	name: z.string().describe("Human-readable experiment name."),
-	goal: z.string().describe("Free-form description of what this session optimizes.").optional(),
-	primary_metric: z
-		.string()
-		.describe(
-			"Primary metric name shown in the dashboard. Match the `METRIC <name>=<value>` lines printed by the benchmark.",
-		),
-	metric_unit: z.string().describe("Unit for the primary metric (e.g. ms, µs, mb). Empty when unitless.").optional(),
+	name: z.string().describe("experiment name"),
+	goal: z.string().describe("session goal").optional(),
+	primary_metric: z.string().describe("primary metric name"),
+	metric_unit: z.string().describe("metric unit (e.g. ms, µs, mb)").optional(),
 	direction: z
 		.enum(["lower", "higher"] as const)
-		.describe("Whether lower or higher values are better. Defaults to lower.")
-		.optional(),
-	secondary_metrics: z
-		.array(z.string())
-		.describe("Names of secondary metrics tracked alongside the primary metric.")
-		.optional(),
-	scope_paths: z
-		.array(z.string())
-		.describe(
-			"Files or directories the agent expects to modify. Used post-hoc to flag scope deviations on log_experiment; never used to block edits.",
-		)
-		.optional(),
-	off_limits: z
-		.array(z.string())
-		.describe(
-			"Paths the agent SHOULD NOT modify. Used post-hoc to flag scope deviations on log_experiment; never used to block edits.",
-		)
-		.optional(),
-	constraints: z.array(z.string()).describe("Free-form constraints (e.g. 'no api break').").optional(),
-	max_iterations: z.number().describe("Soft cap on iterations per segment. Optional.").optional(),
-	new_segment: z
-		.boolean()
-		.describe(
-			"When true, bump to a new segment even when an active session exists. New baselines and best-metric reset.",
-		)
+		.describe("better direction (default lower)")
 		.optional(),
+	secondary_metrics: z.array(z.string()).describe("secondary metric names").optional(),
+	scope_paths: z.array(z.string()).describe("expected-to-modify paths").optional(),
+	off_limits: z.array(z.string()).describe("off-limits paths").optional(),
+	constraints: z.array(z.string()).describe("free-form constraints").optional(),
+	max_iterations: z.number().describe("soft iteration cap per segment").optional(),
+	new_segment: z.boolean().describe("bump to a new segment in existing session").optional(),
 });
 
 interface InitExperimentDetails {

package/src/autoresearch/tools/log-experiment.ts

@@ -37,35 +37,21 @@ import type {
 const EXPERIMENT_TOOL_NAMES = ["init_experiment", "run_experiment", "log_experiment", "update_notes"];
 
 const logExperimentSchema = z.object({
-	metric: z
-		.number()
-		.describe("Primary metric value for this run. May differ from the parsed value; deviation is recorded."),
-	status: z.enum(["keep", "discard", "crash", "checks_failed"] as const).describe("Outcome for this run."),
-	description: z.string().describe("Short description of the experiment."),
-	metrics: z.record(z.string(), z.number()).describe("Secondary metrics for this run.").optional(),
-	asi: z
-		.object({})
-		.passthrough()
-		.describe("Free-form structured metadata captured for this run (hypothesis, learnings, etc.).")
-		.optional(),
-	commit: z
-		.string()
-		.describe("Override the commit hash recorded for this run. Defaults to the current HEAD.")
-		.optional(),
-	justification: z
-		.string()
-		.describe(
-			"Required when the run modifies paths outside scope or inside off-limits and you still want it kept. Free-form explanation.",
-		)
-		.optional(),
+	metric: z.number().describe("primary metric value"),
+	status: z.enum(["keep", "discard", "crash", "checks_failed"] as const).describe("run outcome"),
+	description: z.string().describe("short run description"),
+	metrics: z.record(z.string(), z.number()).describe("secondary metrics").optional(),
+	asi: z.object({}).passthrough().describe("free-form structured metadata").optional(),
+	commit: z.string().describe("override recorded commit hash").optional(),
+	justification: z.string().describe("required when keeping a scope-deviating run").optional(),
 	flag_runs: z
 		.array(
 			z.object({
-				run_id: z.number().describe("Run id (#) of a previously logged run to flag as suspect."),
-				reason: z.string().describe("Why this earlier run is suspect (e.g. reward-hacked, broken metric)."),
+				run_id: z.number().describe("run id to flag"),
+				reason: z.string().describe("why this run is suspect"),
 			}),
 		)
-		.describe("Mark earlier runs as flagged. Flagged runs are excluded from baseline and best-metric math.")
+		.describe("flag earlier runs as suspect")
 		.optional(),
 });
 

package/src/autoresearch/tools/run-experiment.ts

@@ -27,7 +27,7 @@ import type { AutoresearchToolFactoryOptions, RunDetails, RunExperimentProgressD
 import { DEFAULT_HARNESS_COMMAND } from "./init-experiment";
 
 const runExperimentSchema = z.object({
-	timeout_seconds: z.number().describe("Timeout in seconds. Defaults to 600.").optional(),
+	timeout_seconds: z.number().describe("timeout in seconds (default 600)").optional(),
 });
 
 interface ProcessExecutionResult {

package/src/autoresearch/tools/update-notes.ts

@@ -9,15 +9,8 @@ import { openAutoresearchStorageIfExists } from "../storage";
 import type { AutoresearchToolFactoryOptions } from "../types";
 
 const updateNotesSchema = z.object({
-	body: z
-		.string()
-		.describe("Replacement markdown body for the active autoresearch session's notes (your durable playbook)."),
-	append_idea: z
-		.string()
-		.describe(
-			"When set, append this string as a new bullet under an Ideas section instead of replacing the body. `body` is ignored.",
-		)
-		.optional(),
+	body: z.string().describe("replacement notes body"),
+	append_idea: z.string().describe("append as bullet under Ideas instead of replacing body").optional(),
 });
 
 interface UpdateNotesDetails {