@tt-a1i/openpi 0.3.1 → 0.4.0

package/extensions/subagents/src/agent-types.ts

@@ -24,18 +24,18 @@ import * as fs from "node:fs";
 import * as path from "node:path";
 import type { Model } from "@earendil-works/pi-ai";
 import {
-  parseFrontmatter,
   type ModelRegistry,
+  parseFrontmatter,
 } from "@earendil-works/pi-coding-agent";
 import {
   CHILD_EXCLUDED_TOOL_NAMES,
   CHILD_SAFE_PACKAGE_TOOL_NAMES,
 } from "../../shared/child-session.ts";
-import { sanitizeTerminalText } from "../../shared/terminal-text.ts";
 import {
   isSubagentRoleName,
   type SubagentRoleModel,
 } from "../../shared/subagent-roles.ts";
+import { sanitizeTerminalText } from "../../shared/terminal-text.ts";
 import { REASONING_EFFORTS, type ReasoningEffort } from "./domain.ts";
 
 /** Directory name scanned under both the agent dir and a project's `.pi`. */
@@ -101,7 +101,17 @@ export interface AgentType {
   readonly source: string;
 }
 
-const READ_ONLY_TOOLS = ["read", "grep", "find", "ls", "fd", "rg"];
+export const READ_ONLY_AGENT_TOOLS = [
+  "read",
+  "grep",
+  "find",
+  "ls",
+  "fd",
+  "rg",
+  "git_show",
+  "git_diff",
+  "git_log",
+];
 
 /**
  * Built-in role definitions are deliberately provider-free: model selection is
@@ -112,33 +122,45 @@ export const BUILT_IN_AGENT_TYPES: readonly AgentType[] = [
   {
     name: "explorer",
     description:
-      "Read-only codebase exploration. Use high for routine, local, direct tracing; xhigh for interacting state transitions, concurrency or trust boundaries, or subtle multi-path lifecycle/control-flow; max only for exceptionally difficult broad unfamiliar architecture with unresolved competing flows.",
-    tools: READ_ONLY_TOOLS,
-    reasoningEffort: "high",
+      "Read-only codebase exploration. Usually use moderate reasoning, increasing it for harder tasks.",
+    tools: READ_ONLY_AGENT_TOOLS,
     body: "Explore the codebase read-only. Trace the real flow, inspect related callers, and report concise evidence with file paths and line references.",
     source: "built-in:explorer",
   },
   {
     name: "implementer",
-    description: "Focused implementation with repository checks.",
-    tools: ["read", "bash", "edit", "write", "grep", "find", "ls", "fd", "rg"],
-    reasoningEffort: "high",
+    description:
+      "Focused implementation with repository checks. Usually use medium-high reasoning, adjusted for scope, risk, and task difficulty.",
+    tools: [
+      "read",
+      "bash",
+      "edit",
+      "write",
+      "grep",
+      "find",
+      "ls",
+      "fd",
+      "rg",
+      "git_show",
+      "git_diff",
+      "git_log",
+    ],
     body: "Implement the requested change carefully. Trace the affected flow first, make the smallest correct edit, and run relevant checks before reporting results.",
     source: "built-in:implementer",
   },
   {
     name: "reviewer",
-    description: "Read-only review for correctness, safety, and regressions.",
-    tools: READ_ONLY_TOOLS,
-    reasoningEffort: "medium",
+    description:
+      "Read-only review for correctness, safety, and regressions. Usually use high reasoning, adjusted for task difficulty.",
+    tools: READ_ONLY_AGENT_TOOLS,
     body: "Review the requested code or change read-only. Identify concrete correctness, security, and regression risks with evidence; do not modify files.",
     source: "built-in:reviewer",
   },
   {
     name: "advisor",
-    description: "Deep read-only analysis and technical advice.",
-    tools: READ_ONLY_TOOLS,
-    reasoningEffort: "xhigh",
+    description:
+      "Deep read-only analysis and technical advice. Usually use high reasoning, adjusted for task difficulty.",
+    tools: READ_ONLY_AGENT_TOOLS,
     body: "Analyze the problem deeply without modifying files. Explain the relevant tradeoffs, risks, and recommended next step using repository evidence.",
     source: "built-in:advisor",
   },

package/extensions/subagents/src/backends/stub.ts

@@ -10,6 +10,8 @@
  * - supports interrupt (RunSettled Interrupted -> status "error", matching v1);
  * - fails the run when the prompt starts with "FAIL:", and refuses to spawn
  *   at all when it starts with "SPAWNFAIL:" (error-path testing);
+ * - hangs the run after RunStarted without any assistant event when the
+ * prompt starts with "HANG:" (first-response watchdog testing);
  * - appends every event to a JSONL "session file" in tmpdir so the
  *   "full transcript in session file" pointers resolve.
  */
@@ -114,6 +116,11 @@ const makeStubSession = (
       Effect.gen(function* () {
         yield* emit({ _tag: "RunStarted" });
         const failing = userText.trimStart().startsWith("FAIL:");
+        // Stand in for a provider that accepts the request but never emits
+        // its first assistant event.
+        if (userText.trimStart().startsWith("HANG:")) {
+          return yield* Effect.never;
+        }
 
         const thinking = "Looking at the task and planning an approach...";
         for (const delta of chunked(thinking, 16)) {

package/extensions/subagents/src/id-sequence.ts

@@ -0,0 +1,84 @@
+import type { SessionEntry } from "@earendil-works/pi-coding-agent";
+
+export const SUBAGENT_ID_WATERMARK_ENTRY_TYPE = "subagent-id-watermark";
+
+export interface SubagentIdCounters {
+  readonly modelCounter: number;
+  readonly btwCounter: number;
+}
+
+interface IdWatermarkData {
+  readonly version: 1;
+  readonly id: string;
+}
+
+function record(value: unknown): Record<string, unknown> | undefined {
+  return typeof value === "object" && value !== null
+    ? (value as Record<string, unknown>)
+    : undefined;
+}
+
+function idFrom(value: unknown) {
+  const candidate = record(value)?.id;
+  return typeof candidate === "string" ? candidate : undefined;
+}
+
+function idFromEntry(entry: SessionEntry) {
+  if (entry.type === "custom") {
+    const data = record(entry.data);
+    if (
+      entry.customType === SUBAGENT_ID_WATERMARK_ENTRY_TYPE &&
+      data?.version === 1
+    ) {
+      return idFrom(data);
+    }
+    if (
+      entry.customType === "subagent-finished" ||
+      entry.customType === "btw-result"
+    ) {
+      return idFrom(data);
+    }
+    if (entry.customType === "subagent-result") {
+      return idFrom(data?.details) ?? idFrom(data);
+    }
+    return undefined;
+  }
+
+  if (entry.type === "custom_message") {
+    return entry.customType === "subagent-result"
+      ? idFrom(entry.details)
+      : undefined;
+  }
+
+  if (entry.type !== "message") return undefined;
+  const message = record(entry.message);
+  return message?.role === "toolResult" && message.toolName === "subagent_spawn"
+    ? idFrom(message.details)
+    : undefined;
+}
+
+function sequence(id: string | undefined, prefix: "sa" | "btw") {
+  if (!id) return undefined;
+  const match = new RegExp(`^${prefix}-(\\d+)$`, "u").exec(id);
+  if (!match) return undefined;
+  const value = Number(match[1]);
+  return Number.isSafeInteger(value) && value > 0 ? value : undefined;
+}
+
+/** Restore branch-local id high-water marks, including sessions from before watermarks existed. */
+export function restoreSubagentIdCounters(
+  entries: readonly SessionEntry[],
+): SubagentIdCounters {
+  let modelCounter = 0;
+  let btwCounter = 0;
+  for (const entry of entries) {
+    const id = idFromEntry(entry);
+    modelCounter = Math.max(modelCounter, sequence(id, "sa") ?? 0);
+    btwCounter = Math.max(btwCounter, sequence(id, "btw") ?? 0);
+  }
+  return { modelCounter, btwCounter };
+}
+
+export function subagentIdWatermark(id: string): IdWatermarkData {
+  return { version: 1, id };
+}