fix(pi-claude-cli): unblock parameterless MCP tool calls in triage

Triage with claude-sonnet-4-6 via pi-claude-cli kept looping on
fn_review_spec calls that were rejected by pi's validator with
"root: must be object". Parameterless MCP tools (schema
{type:"object", properties:{}}) emit zero input_json_delta events,
so partialJson stayed "" and the catch fell through to
finalArgs = "" — a string, which TypeBox's Type.Object({}) rightly
refuses. Default empty partialJson to {} so the call lands.

Also:
- Add a 2-step reminder loop in triage before swapping to the
  fallback planning model — primary models that wrote PROMPT.md
  but forgot fn_review_spec recover from a nudge, no need to pay
  the cold-start tax of a new triage on a different model.
- Inject @runfusion/fusion's own pi extension into dashboard/
  daemon/serve sessions and propagate the path to createFnAgent
  via setHostExtensionPaths so fn_* tools register globally
  without requiring `pi install npm:@runfusion/fusion`.
- Drop the "historical" qualifier from replayed tool labels —
  Claude was reading "TOOL RESULT (historical Read):" as
  "previous session, ignore" and looping on verification.
- Remove subprocess-lifecycle stderr debug logs that landed for
  hang diagnosis — root cause is fixed, the noise can go.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
gsxdsm
2026-04-26 00:00:49 -07:00
parent 15149905fd
commit c1a2b6edd7
13 changed files with 269 additions and 89 deletions

View File

@@ -25,7 +25,7 @@ import {
} from "@fusion/core"; } from "@fusion/core";
import type { AutomationRunResult, ScheduledTask } from "@fusion/core"; import type { AutomationRunResult, ScheduledTask } from "@fusion/core";
import { createServer, GitHubClient, createSkillsAdapter, getProjectSettingsPath, loadTlsCredentialsFromEnv } from "@fusion/dashboard"; import { createServer, GitHubClient, createSkillsAdapter, getProjectSettingsPath, loadTlsCredentialsFromEnv } from "@fusion/dashboard";
import { ProjectEngineManager, PeerExchangeService } from "@fusion/engine"; import { ProjectEngineManager, PeerExchangeService, setHostExtensionPaths } from "@fusion/engine";
import { import {
AuthStorage, AuthStorage,
DefaultPackageManager, DefaultPackageManager,
@@ -49,6 +49,7 @@ import {
resolveClaudeCliExtensionPaths, resolveClaudeCliExtensionPaths,
setCachedClaudeCliResolution, setCachedClaudeCliResolution,
} from "./claude-cli-extension.js"; } from "./claude-cli-extension.js";
import { resolveSelfExtension } from "./self-extension.js";
import { createReadOnlyAuthFileStorage, mergeAuthStorageReads, wrapAuthStorageWithApiKeyProviders } from "./provider-auth.js"; import { createReadOnlyAuthFileStorage, mergeAuthStorageReads, wrapAuthStorageWithApiKeyProviders } from "./provider-auth.js";
import { getFusionAuthPath, getLegacyAuthPaths, getModelRegistryModelsPath, getPackageManagerAgentDir } from "./auth-paths.js"; import { getFusionAuthPath, getLegacyAuthPaths, getModelRegistryModelsPath, getPackageManagerAgentDir } from "./auth-paths.js";
import { resolveProject } from "../project-context.js"; import { resolveProject } from "../project-context.js";
@@ -416,8 +417,17 @@ export async function runDaemon(opts: DaemonOptions = {}) {
// external `pi-claude-cli` install. Drops shadowing externals (e.g. a // external `pi-claude-cli` install. Drops shadowing externals (e.g. a
// global `npm install -g pi-claude-cli`) so the upstream's once-and-lock // global `npm install -g pi-claude-cli`) so the upstream's once-and-lock
// MCP-config bug can't poison sessions. // MCP-config bug can't poison sessions.
// Inject the cli's own extension (@runfusion/fusion) so fn_* tools
// register globally without requiring `pi install npm:@runfusion/fusion`.
const selfExtension = resolveSelfExtension();
const selfExtensionPaths = selfExtension.status === "ok" ? [selfExtension.path] : [];
if (selfExtension.status !== "ok") {
console.warn(`[extensions] self: ${selfExtension.reason}`);
}
setHostExtensionPaths(selfExtensionPaths);
const reconciledExtensionPaths = reconcileClaudeCliPaths( const reconciledExtensionPaths = reconcileClaudeCliPaths(
[...getEnabledPiExtensionPaths(cwd), ...packageExtensionPaths, ...claudeCliPaths], [...selfExtensionPaths, ...getEnabledPiExtensionPaths(cwd), ...packageExtensionPaths, ...claudeCliPaths],
claudeCliPaths[0] ?? null, claudeCliPaths[0] ?? null,
); );

View File

@@ -26,7 +26,7 @@ import {
stopAllDevServers, stopAllDevServers,
type RuntimeLogger, type RuntimeLogger,
} from "@fusion/dashboard"; } from "@fusion/dashboard";
import { aiMergeTask, MissionAutopilot, MissionExecutionLoop, HeartbeatMonitor, HeartbeatTriggerScheduler, type WakeContext, ProjectEngineManager, PeerExchangeService } from "@fusion/engine"; import { aiMergeTask, MissionAutopilot, MissionExecutionLoop, HeartbeatMonitor, HeartbeatTriggerScheduler, type WakeContext, ProjectEngineManager, PeerExchangeService, setHostExtensionPaths } from "@fusion/engine";
import { AuthStorage, DefaultPackageManager, ModelRegistry, SettingsManager, discoverAndLoadExtensions, createExtensionRuntime } from "@mariozechner/pi-coding-agent"; import { AuthStorage, DefaultPackageManager, ModelRegistry, SettingsManager, discoverAndLoadExtensions, createExtensionRuntime } from "@mariozechner/pi-coding-agent";
import { import {
getMergeStrategy, getMergeStrategy,
@@ -46,6 +46,7 @@ import {
resolveClaudeCliExtensionPaths, resolveClaudeCliExtensionPaths,
setCachedClaudeCliResolution, setCachedClaudeCliResolution,
} from "./claude-cli-extension.js"; } from "./claude-cli-extension.js";
import { resolveSelfExtension } from "./self-extension.js";
import { DashboardTUI, DashboardLogSink, isTTYAvailable, type SystemInfo, type GitStatus, type GitCommit, type GitCommitDetail, type GitBranch, type GitWorktree, type FileEntry, type FileReadResult, type TaskStep as TUITaskStep, type TaskLogEntry as TUITaskLogEntry, type TaskDetailData, type TaskEvent } from "./dashboard-tui/index.js"; import { DashboardTUI, DashboardLogSink, isTTYAvailable, type SystemInfo, type GitStatus, type GitCommit, type GitCommitDetail, type GitBranch, type GitWorktree, type FileEntry, type FileReadResult, type TaskStep as TUITaskStep, type TaskLogEntry as TUITaskLogEntry, type TaskDetailData, type TaskEvent } from "./dashboard-tui/index.js";
// Re-export for backward compatibility with tests // Re-export for backward compatibility with tests
@@ -1131,9 +1132,24 @@ export async function runDashboard(port: number, opts: { paused?: boolean; dev?:
} }
})(); })();
// Always inject the cli's own extension (`@runfusion/fusion`) so its
// `fn_*` tools register globally even when the user hasn't run
// `pi install npm:@runfusion/fusion`. Without this, agent chat with
// pi-claude-cli has no fn_* tools at all.
const selfExtension = resolveSelfExtension();
const selfExtensionPaths = selfExtension.status === "ok" ? [selfExtension.path] : [];
if (selfExtension.status !== "ok") {
logSink.warn(`[extensions] self: ${selfExtension.reason}`, "extensions");
}
// Propagate self-extension path to engine so createFnAgent sessions
// (chat, refine, mission, etc.) also load fn_* tools, not just the
// dashboard's extension runtime.
setHostExtensionPaths(selfExtensionPaths);
// Load all enabled extensions: Fusion/Pi filesystem-discovered + package-resolved. // Load all enabled extensions: Fusion/Pi filesystem-discovered + package-resolved.
const extensionsResult = await discoverAndLoadExtensions( const extensionsResult = await discoverAndLoadExtensions(
[ [
...selfExtensionPaths,
...getEnabledPiExtensionPaths(cwd), ...getEnabledPiExtensionPaths(cwd),
...packageExtensionPaths, ...packageExtensionPaths,
...claudeCliPaths, ...claudeCliPaths,

View File

@@ -0,0 +1,88 @@
/**
* Resolver for the CLI's own pi extension (`@runfusion/fusion`).
*
* `packages/cli/src/extension.ts` registers all `fn_*` tools (`fn_task_create`,
* `fn_mission_create`, etc.). For these to appear in pi sessions launched by
* the dashboard/daemon (e.g. agent chat using pi-claude-cli), the extension
* must be loaded by pi's `discoverAndLoadExtensions`.
*
* Pi normally only loads extensions that are registered in
* `~/.pi/agent/settings.json` packages or symlinked into an extensions
* directory. To avoid requiring users to `pi install npm:@runfusion/fusion`
* before fn_* tools work, we resolve the bundled extension path at runtime
* and inject it into the load list — same pattern as
* `claude-cli-extension.ts` does for `@fusion/pi-claude-cli`.
*
* In dev (`pnpm dev dashboard`), this resolves to `packages/cli/src/extension.ts`.
* In a published install, it resolves to `<install>/dist/extension.js`.
*/
import { existsSync, readFileSync } from "node:fs";
import { dirname, resolve } from "node:path";
import { fileURLToPath } from "node:url";
export type SelfExtensionResolution =
| { status: "ok"; path: string; packageVersion: string }
| { status: "missing"; reason: string };
/**
* Resolve the absolute path to the cli's own pi extension entry file.
*
* Walk up from this module to the @runfusion/fusion package.json and read
* `pi.extensions[0]`. Prefer `src/extension.ts` over `dist/extension.js` when
* both exist so dev iterations don't require a rebuild.
*/
export function resolveSelfExtension(): SelfExtensionResolution {
const here = dirname(fileURLToPath(import.meta.url));
let pkgDir: string | undefined;
let cur = here;
for (let i = 0; i < 5; i++) {
if (existsSync(resolve(cur, "package.json"))) {
try {
const parsed = JSON.parse(readFileSync(resolve(cur, "package.json"), "utf-8")) as { name?: string };
if (parsed.name === "@runfusion/fusion") {
pkgDir = cur;
break;
}
} catch {
// ignore and keep walking
}
}
const parent = resolve(cur, "..");
if (parent === cur) break;
cur = parent;
}
if (!pkgDir) {
return { status: "missing", reason: "Could not locate @runfusion/fusion package.json from CLI module" };
}
let pkgJson: { pi?: { extensions?: unknown }; version?: string };
try {
pkgJson = JSON.parse(readFileSync(resolve(pkgDir, "package.json"), "utf-8")) as typeof pkgJson;
} catch (err) {
return { status: "missing", reason: `Failed to read @runfusion/fusion package.json: ${err instanceof Error ? err.message : String(err)}` };
}
// Prefer src/extension.ts in dev (always fresh); fall back to pi.extensions[0]
// (dist/extension.js in production).
const srcEntry = resolve(pkgDir, "src", "extension.ts");
if (existsSync(srcEntry)) {
return { status: "ok", path: srcEntry, packageVersion: pkgJson.version ?? "unknown" };
}
const extensions = pkgJson.pi?.extensions;
if (!Array.isArray(extensions) || extensions.length === 0) {
return { status: "missing", reason: "@runfusion/fusion package.json has no pi.extensions array" };
}
const rawEntry = extensions[0];
if (typeof rawEntry !== "string" || rawEntry.length === 0) {
return { status: "missing", reason: "@runfusion/fusion pi.extensions[0] is not a valid path string" };
}
const entryPath = resolve(pkgDir, rawEntry);
if (!existsSync(entryPath)) {
return { status: "missing", reason: `@runfusion/fusion extension file not found at ${entryPath}` };
}
return { status: "ok", path: entryPath, packageVersion: pkgJson.version ?? "unknown" };
}

View File

@@ -25,7 +25,7 @@ import {
} from "@fusion/core"; } from "@fusion/core";
import type { AutomationRunResult, ScheduledTask } from "@fusion/core"; import type { AutomationRunResult, ScheduledTask } from "@fusion/core";
import { createServer, GitHubClient, createSkillsAdapter, getProjectSettingsPath, loadTlsCredentialsFromEnv } from "@fusion/dashboard"; import { createServer, GitHubClient, createSkillsAdapter, getProjectSettingsPath, loadTlsCredentialsFromEnv } from "@fusion/dashboard";
import { ProjectEngineManager, PeerExchangeService } from "@fusion/engine"; import { ProjectEngineManager, PeerExchangeService, setHostExtensionPaths } from "@fusion/engine";
import { import {
AuthStorage, AuthStorage,
DefaultPackageManager, DefaultPackageManager,
@@ -52,6 +52,7 @@ import {
resolveClaudeCliExtensionPaths, resolveClaudeCliExtensionPaths,
setCachedClaudeCliResolution, setCachedClaudeCliResolution,
} from "./claude-cli-extension.js"; } from "./claude-cli-extension.js";
import { resolveSelfExtension } from "./self-extension.js";
const DIAGNOSTIC_INTERVAL_MS = 30 * 60 * 1000; // 30 minutes const DIAGNOSTIC_INTERVAL_MS = 30 * 60 * 1000; // 30 minutes
let diagnosticIntervalHandle: ReturnType<typeof setInterval> | null = null; let diagnosticIntervalHandle: ReturnType<typeof setInterval> | null = null;
@@ -468,8 +469,18 @@ export async function runServe(
} }
})(); })();
// Inject the cli's own extension so fn_* tools register globally without
// requiring `pi install npm:@runfusion/fusion`.
const selfExtension = resolveSelfExtension();
const selfExtensionPaths = selfExtension.status === "ok" ? [selfExtension.path] : [];
if (selfExtension.status !== "ok") {
console.warn(`[extensions] self: ${selfExtension.reason}`);
}
setHostExtensionPaths(selfExtensionPaths);
const extensionsResult = await discoverAndLoadExtensions( const extensionsResult = await discoverAndLoadExtensions(
[ [
...selfExtensionPaths,
...getEnabledPiExtensionPaths(cwd), ...getEnabledPiExtensionPaths(cwd),
...packageExtensionPaths, ...packageExtensionPaths,
...claudeCliPaths, ...claudeCliPaths,

View File

@@ -840,11 +840,20 @@ describe("Triage re-pick after restart", () => {
const task = makeTask("FN-062", "triage"); const task = makeTask("FN-062", "triage");
store.getTask.mockResolvedValue(makeTaskDetail("FN-062", "triage")); store.getTask.mockResolvedValue(makeTaskDetail("FN-062", "triage"));
// Slow agent to keep task in processing // Slow agent to keep task in processing — only the first prompt() blocks;
// subsequent calls (e.g. the no-APPROVE reminder loop in triage) resolve
// immediately so cleanup can drain.
let resolvePrompt: (() => void) | undefined; let resolvePrompt: (() => void) | undefined;
let promptCallCount = 0;
mockedCreateFnAgent.mockResolvedValue({ mockedCreateFnAgent.mockResolvedValue({
session: { session: {
prompt: vi.fn().mockImplementation(() => new Promise<void>((r) => { resolvePrompt = r; })), prompt: vi.fn().mockImplementation(() => {
promptCallCount += 1;
if (promptCallCount === 1) {
return new Promise<void>((r) => { resolvePrompt = r; });
}
return Promise.resolve();
}),
dispose: vi.fn(), dispose: vi.fn(),
}, },
} as any); } as any);

View File

@@ -17,7 +17,7 @@ export { MissionAutopilot, type MissionAutopilotOptions } from "./mission-autopi
export { MissionExecutionLoop, type MissionExecutionLoopOptions, type ValidationResult, loopLog } from "./mission-execution-loop.js"; export { MissionExecutionLoop, type MissionExecutionLoopOptions, type ValidationResult, loopLog } from "./mission-execution-loop.js";
export { aiMergeTask, type MergerOptions } from "./merger.js"; export { aiMergeTask, type MergerOptions } from "./merger.js";
export { reviewStep, type ReviewType, type ReviewVerdict, type ReviewResult, type ReviewOptions } from "./reviewer.js"; export { reviewStep, type ReviewType, type ReviewVerdict, type ReviewResult, type ReviewOptions } from "./reviewer.js";
export { createFnAgent, promptWithFallback, describeModel, type AgentOptions, type AgentResult } from "./pi.js"; export { createFnAgent, promptWithFallback, describeModel, setHostExtensionPaths, getHostExtensionPaths, type AgentOptions, type AgentResult } from "./pi.js";
export { export {
resolveSessionSkills, resolveSessionSkills,
createSkillsOverrideFromSelection, createSkillsOverrideFromSelection,

View File

@@ -50,6 +50,24 @@ export interface AgentResult {
sessionFile?: string; sessionFile?: string;
} }
/**
* Process-global list of extension paths to inject into every createFnAgent
* session. Set once at startup by the host (cli's dashboard/daemon/serve)
* before any sessions are created. The paths are passed to pi's
* `DefaultResourceLoader` as `additionalExtensionPaths` so the cli's own
* `@runfusion/fusion` extension (registering `fn_*` tools) is loaded inside
* every agent session — including chat sessions that pass no `customTools`.
*/
let hostExtensionPaths: string[] = [];
export function setHostExtensionPaths(paths: readonly string[]): void {
hostExtensionPaths = [...paths];
}
export function getHostExtensionPaths(): readonly string[] {
return hostExtensionPaths;
}
export interface PromptableSession extends AgentSession { export interface PromptableSession extends AgentSession {
promptWithFallback: (prompt: string, options?: unknown) => Promise<void>; promptWithFallback: (prompt: string, options?: unknown) => Promise<void>;
} }
@@ -1066,6 +1084,10 @@ export async function createFnAgent(options: AgentOptions): Promise<AgentResult>
settingsManager, settingsManager,
systemPromptOverride: () => options.systemPrompt, systemPromptOverride: () => options.systemPrompt,
appendSystemPromptOverride: () => [], appendSystemPromptOverride: () => [],
// Inject host-supplied extension paths (e.g. cli's own `@runfusion/fusion`
// extension that registers `fn_*` tools) so they're loaded inside every
// agent session, including chat sessions that don't pass `customTools`.
...(hostExtensionPaths.length > 0 ? { additionalExtensionPaths: [...hostExtensionPaths] } : {}),
...(skillsOverrideFn ? { skillsOverride: skillsOverrideFn } : {}), ...(skillsOverrideFn ? { skillsOverride: skillsOverrideFn } : {}),
}); });
await resourceLoader.reload(); await resourceLoader.reload();

View File

@@ -1043,6 +1043,44 @@ export class TriageProcessor {
return; return;
} }
// Before swapping to the fallback model, give the primary one more
// shot with a pointed reminder. The model may have written PROMPT.md
// but stopped without calling fn_review_spec — that's recoverable
// with a nudge, no need to discard the session and pay the cold-start
// tax of a new triage on a different model.
const MAX_REVIEW_REMINDERS = 2;
let reviewReminders = 0;
while (
specReviewVerdictRef.current !== "APPROVE" &&
!this.pauseAborted.has(task.id) &&
!this.stuckAborted.has(task.id) &&
createdSubtasksRef.current.length === 0 &&
reviewReminders < MAX_REVIEW_REMINDERS
) {
reviewReminders += 1;
const verdictDesc =
specReviewVerdictRef.current === null
? "fn_review_spec was never called"
: `verdict was ${specReviewVerdictRef.current}`;
triageLog.warn(
`${task.id} primary planning model returned without APPROVE (${verdictDesc}) — reminder ${reviewReminders}/${MAX_REVIEW_REMINDERS}`,
);
await this.store.logEntry(
task.id,
`Primary planning model returned without APPROVE (${verdictDesc}) — reminder ${reviewReminders}/${MAX_REVIEW_REMINDERS}`,
);
const reminder =
specReviewVerdictRef.current === null
? "You wrote the PROMPT.md but did not call `fn_review_spec()`. Call `fn_review_spec()` now to validate the spec. Do not stop until the verdict is APPROVE."
: `Spec review verdict was ${specReviewVerdictRef.current}. Address the feedback, rewrite the PROMPT.md as needed, and call \`fn_review_spec()\` again. Do not stop until the verdict is APPROVE.`;
stuckDetector?.recordActivity(task.id);
await promptWithFallback(session, reminder);
checkSessionError(session);
if (this.pauseAborted.has(task.id) || this.stuckAborted.has(task.id)) {
break;
}
}
const planningFallbackProvider = settings.planningFallbackProvider; const planningFallbackProvider = settings.planningFallbackProvider;
const planningFallbackModelId = settings.planningFallbackModelId; const planningFallbackModelId = settings.planningFallbackModelId;
const canRetryWithPlanningFallback = const canRetryWithPlanningFallback =

View File

@@ -444,6 +444,40 @@ describe("createEventBridge", () => {
expect(event.toolCall.arguments).toEqual({ path: "/foo.ts" }); expect(event.toolCall.arguments).toEqual({ path: "/foo.ts" });
}); });
it("emits {} for parameterless MCP tool calls (no input_json_delta)", () => {
// Parameterless MCP tools (e.g. fn_review_spec, schema
// {type:"object", properties:{}}) emit ZERO input_json_delta events.
// Without the empty-partialJson guard, finalArgs would fall through to
// the raw-string fallback ("") and pi's TypeBox validator would reject
// the call with "Validation failed for tool ...: root: must be object".
const bridge = createBridgeWithStart();
bridge.handleEvent({
type: "content_block_start",
index: 0,
content_block: {
type: "tool_use",
id: "toolu_01XYZ",
name: "mcp__custom-tools__fn_review_spec",
},
});
// No content_block_delta with input_json_delta — Claude emits none for
// parameterless tools.
stream.push.mockClear();
stream.events.length = 0;
bridge.handleEvent({
type: "content_block_stop",
index: 0,
});
expect(stream.push).toHaveBeenCalledTimes(1);
const event = stream.events[0] as any;
expect(event.type).toBe("toolcall_end");
expect(event.toolCall.arguments).toEqual({});
// The MCP prefix should be stripped: pi sees the bare tool name.
expect(event.toolCall.name).toBe("fn_review_spec");
});
it("tracks multiple tool_use blocks independently by Claude event.index", () => { it("tracks multiple tool_use blocks independently by Claude event.index", () => {
const bridge = createBridgeWithStart(); const bridge = createBridgeWithStart();

View File

@@ -21,7 +21,7 @@ describe("buildPrompt", () => {
expect(buildPrompt(context)).toBe("ASSISTANT:\nHi there"); expect(buildPrompt(context)).toBe("ASSISTANT:\nHi there");
}); });
it("produces 'TOOL RESULT (historical {claudeName}):\\n{content}' for a tool result message", () => { it("produces 'TOOL RESULT ({claudeName}):\\n{content}' for a tool result message", () => {
const context = { const context = {
messages: [ messages: [
{ {
@@ -33,7 +33,7 @@ describe("buildPrompt", () => {
} as unknown as any; } as unknown as any;
// Pi tool name "read" should be mapped to Claude name "Read" in the label // Pi tool name "read" should be mapped to Claude name "Read" in the label
expect(buildPrompt(context)).toBe( expect(buildPrompt(context)).toBe(
"TOOL RESULT (historical Read):\nfile contents here", "TOOL RESULT (Read):\nfile contents here",
); );
}); });
@@ -57,7 +57,7 @@ describe("buildPrompt", () => {
"What is in file.ts?", "What is in file.ts?",
"ASSISTANT:", "ASSISTANT:",
"Let me read that file.", "Let me read that file.",
"TOOL RESULT (historical Read):", "TOOL RESULT (Read):",
"export const x = 1;", "export const x = 1;",
"USER:", "USER:",
"Now explain it.", "Now explain it.",
@@ -109,7 +109,7 @@ describe("buildPrompt", () => {
// Tool name should be mapped from pi "read" to Claude "Read" // Tool name should be mapped from pi "read" to Claude "Read"
// Arg "path" should be mapped from pi format to Claude "file_path" // Arg "path" should be mapped from pi format to Claude "file_path"
expect(result).toContain( expect(result).toContain(
'Historical tool call (non-executable): Read args={"file_path":"/file.ts"}', '[Prior tool call — already executed; result follows in TOOL RESULT (Read):] args={"file_path":"/file.ts"}',
); );
}); });
@@ -158,7 +158,7 @@ describe("buildPrompt", () => {
const result = buildPrompt(context); const result = buildPrompt(context);
// Pi "bash" maps to Claude "Bash" // Pi "bash" maps to Claude "Bash"
expect(result).toContain( expect(result).toContain(
"Historical tool call (non-executable): Bash args={}", "[Prior tool call — already executed; result follows in TOOL RESULT (Bash):] args={}",
); );
}); });
@@ -177,7 +177,7 @@ describe("buildPrompt", () => {
} as unknown as any; } as unknown as any;
const result = buildPrompt(context); const result = buildPrompt(context);
expect(result).toBe("TOOL RESULT (historical Bash):\nline 1\nline 2"); expect(result).toBe("TOOL RESULT (Bash):\nline 1\nline 2");
}); });
describe("tool name and argument reverse mapping", () => { describe("tool name and argument reverse mapping", () => {
@@ -237,7 +237,7 @@ describe("buildPrompt", () => {
} as unknown as any; } as unknown as any;
const result = buildPrompt(context); const result = buildPrompt(context);
expect(result).toContain("TOOL RESULT (historical Read):"); expect(result).toContain("TOOL RESULT (Read):");
}); });
it("prefixes custom (non-built-in) tool names with MCP prefix", () => { it("prefixes custom (non-built-in) tool names with MCP prefix", () => {
@@ -281,7 +281,8 @@ describe("buildPrompt", () => {
const result = buildPrompt(context); const result = buildPrompt(context);
// String arguments should be serialized as JSON string // String arguments should be serialized as JSON string
expect(result).toContain('Read args="raw string args"'); expect(result).toContain('TOOL RESULT (Read):');
expect(result).toContain('args="raw string args"');
}); });
}); });
}); });
@@ -688,7 +689,7 @@ describe("custom tool history replay", () => {
} as unknown as any; } as unknown as any;
const result = buildPrompt(context); const result = buildPrompt(context);
expect(result).toContain("TOOL RESULT (historical Read):"); expect(result).toContain("TOOL RESULT (Read):");
expect(result).not.toContain("mcp__custom-tools__"); expect(result).not.toContain("mcp__custom-tools__");
}); });
@@ -1019,7 +1020,7 @@ describe("buildResumePrompt", () => {
], ],
}; };
const result = buildResumePrompt(context) as string; const result = buildResumePrompt(context) as string;
expect(result).toContain("TOOL RESULT (historical Read):"); expect(result).toContain("TOOL RESULT (Read):");
expect(result).toContain("file contents here"); expect(result).toContain("file contents here");
expect(result).toContain("Now explain it"); expect(result).toContain("Now explain it");
}); });

View File

@@ -323,13 +323,24 @@ export function createEventBridge(
partial: output, partial: output,
}); });
} else if (block.type === "tool_use") { } else if (block.type === "tool_use") {
// Final JSON parse with fallback to raw string // Final JSON parse with fallback to raw string.
// Special case: parameterless MCP tools (e.g. fn_review_spec, schema
// `{type:"object", properties:{}}`) emit ZERO input_json_delta events,
// so `partialJson` stays "". Without this guard we'd JSON.parse("")
// → throw → fall through to `finalArgs = ""` (raw string), and pi's
// TypeBox validator then rejects with "root: must be object" because
// an empty string is not an object. Default to `{}` so the call lands.
let finalArgs: Record<string, unknown> | string; let finalArgs: Record<string, unknown> | string;
try { const trimmedJson = block.partialJson.trim();
const parsed = JSON.parse(block.partialJson); if (trimmedJson === "") {
finalArgs = translateClaudeArgsToPi(block.claudeName, parsed); finalArgs = {};
} catch { } else {
finalArgs = block.partialJson; try {
const parsed = JSON.parse(trimmedJson);
finalArgs = translateClaudeArgsToPi(block.claudeName, parsed);
} catch {
finalArgs = block.partialJson;
}
} }
// Update output.content with final arguments // Update output.content with final arguments

View File

@@ -62,7 +62,7 @@ type AnthropicContentBlock =
* Each message is labeled with its role: * Each message is labeled with its role:
* - USER: for user messages * - USER: for user messages
* - ASSISTANT: for assistant messages * - ASSISTANT: for assistant messages
* - TOOL RESULT (historical {toolName}): for tool result messages * - TOOL RESULT ({toolName}): for tool result messages
*/ */
/** Module-level counter for placeholder images, reset per buildPrompt call. */ /** Module-level counter for placeholder images, reset per buildPrompt call. */
let placeholderImageCount = 0; let placeholderImageCount = 0;
@@ -213,7 +213,7 @@ export function buildResumePrompt(context: PiContext): string | AnthropicContent
const claudeToolName = msg.toolName const claudeToolName = msg.toolName
? mapPiToolNameToClaude(msg.toolName) ? mapPiToolNameToClaude(msg.toolName)
: "unknown"; : "unknown";
parts.push(`TOOL RESULT (historical ${claudeToolName}):`); parts.push(`TOOL RESULT (${claudeToolName}):`);
} }
parts.push(toolResultContentToText(msg.content)); parts.push(toolResultContentToText(msg.content));
} else if (msg.role === "user") { } else if (msg.role === "user") {
@@ -283,7 +283,7 @@ export function buildPrompt(context: PiContext): string | AnthropicContentBlock[
const claudeToolName = message.toolName const claudeToolName = message.toolName
? mapPiToolNameToClaude(message.toolName) ? mapPiToolNameToClaude(message.toolName)
: "unknown"; : "unknown";
historyParts.push(`TOOL RESULT (historical ${claudeToolName}):`); historyParts.push(`TOOL RESULT (${claudeToolName}):`);
} }
// Extract text portion of tool result // Extract text portion of tool result
historyParts.push(toolResultContentToText(message.content)); historyParts.push(toolResultContentToText(message.content));
@@ -347,7 +347,7 @@ export function buildPrompt(context: PiContext): string | AnthropicContentBlock[
const claudeToolName = message.toolName const claudeToolName = message.toolName
? mapPiToolNameToClaude(message.toolName) ? mapPiToolNameToClaude(message.toolName)
: "unknown"; : "unknown";
parts.push(`TOOL RESULT (historical ${claudeToolName}):`); parts.push(`TOOL RESULT (${claudeToolName}):`);
} }
parts.push(toolResultContentToText(message.content)); parts.push(toolResultContentToText(message.content));
} }
@@ -451,8 +451,6 @@ function rewriteCustomToolReferences(
} }
let result = prompt; let result = prompt;
let totalRewrites = 0;
const rewritten: string[] = [];
for (const tool of tools) { for (const tool of tools) {
if (BUILT_IN_PI_TOOLS.has(tool.name)) continue; if (BUILT_IN_PI_TOOLS.has(tool.name)) continue;
// \b doesn't treat `_` as a word boundary the way we want here, so anchor // \b doesn't treat `_` as a word boundary the way we want here, so anchor
@@ -464,19 +462,7 @@ function rewriteCustomToolReferences(
`(?<![A-Za-z0-9_])(?<!mcp__custom-tools__)${escaped}(?![A-Za-z0-9_])`, `(?<![A-Za-z0-9_])(?<!mcp__custom-tools__)${escaped}(?![A-Za-z0-9_])`,
"g", "g",
); );
const before = result;
result = result.replace(pattern, `mcp__custom-tools__${tool.name}`); result = result.replace(pattern, `mcp__custom-tools__${tool.name}`);
if (result !== before) {
const matches = before.match(pattern);
const count = matches?.length ?? 0;
totalRewrites += count;
rewritten.push(`${tool.name}×${count}`);
}
}
if (totalRewrites > 0) {
console.error(
`[pi-claude-cli] system prompt: rewrote ${totalRewrites} custom tool ref(s) [${rewritten.join(", ")}]`,
);
} }
return result; return result;
} }
@@ -580,7 +566,7 @@ function contentToText(content: string | unknown[]): string {
: typeof rawArgs === "string" : typeof rawArgs === "string"
? JSON.stringify(rawArgs) ? JSON.stringify(rawArgs)
: "{}"; : "{}";
return `Historical tool call (non-executable): ${claudeName} args=${argsStr}`; return `[Prior tool call — already executed; result follows in TOOL RESULT (${claudeName}):] args=${argsStr}`;
} }
// Unknown block types are represented as a placeholder // Unknown block types are represented as a placeholder
return `[${String(block.type)}]`; return `[${String(block.type)}]`;

View File

@@ -136,11 +136,6 @@ export function streamViaCli(
}); });
const getStderr = captureStderr(proc); const getStderr = captureStderr(proc);
const spawnTime = Date.now(); const spawnTime = Date.now();
const procPid = proc.pid;
const traceMode = resumeSessionId ? "resume" : "new";
console.error(
`[pi-claude-cli] spawn pid=${procPid} model=${model.id} mode=${traceMode} effort=${effort ?? "default"} promptLen=${typeof prompt === "string" ? prompt.length : 0} systemPromptLen=${systemPrompt?.length ?? 0} mcp=${options?.mcpConfigPath ? "yes" : "no"}`,
);
// Register in global process registry for teardown cleanup // Register in global process registry for teardown cleanup
registerProcess(proc); registerProcess(proc);
@@ -231,12 +226,8 @@ export function streamViaCli(
}); });
// Handle subprocess close -- surface crashes with stderr and exit code // Handle subprocess close -- surface crashes with stderr and exit code
proc.on("close", (code: number | null, signal: string | null) => { proc.on("close", (code: number | null, _signal: string | null) => {
clearTimeout(inactivityTimer); clearTimeout(inactivityTimer);
const elapsedMs = Date.now() - spawnTime;
console.error(
`[pi-claude-cli] close pid=${procPid} code=${code ?? "null"} signal=${signal ?? "null"} elapsedMs=${elapsedMs} broken=${broken}`,
);
if (broken) return; // Break-early kill, expected if (broken) return; // Break-early kill, expected
if (code !== 0 && code !== null) { if (code !== 0 && code !== null) {
const stderr = getStderr(); const stderr = getStderr();
@@ -250,9 +241,6 @@ export function streamViaCli(
// Start inactivity timer after writing user message // Start inactivity timer after writing user message
resetInactivityTimer(); resetInactivityTimer();
let firstLineLoggedAt = 0;
let lineCount = 0;
// Process NDJSON lines from stdout using event-based callback // Process NDJSON lines from stdout using event-based callback
// NOTE: Using 'line' event instead of `for await` because the async // NOTE: Using 'line' event instead of `for await` because the async
// iterator batches lines, breaking real-time streaming to pi. // iterator batches lines, breaking real-time streaming to pi.
@@ -261,36 +249,10 @@ export function streamViaCli(
// Reset inactivity timer on each line of output // Reset inactivity timer on each line of output
resetInactivityTimer(); resetInactivityTimer();
lineCount++;
if (lineCount === 1) {
firstLineLoggedAt = Date.now();
console.error(
`[pi-claude-cli] first-stdout-line pid=${procPid} afterMs=${firstLineLoggedAt - spawnTime}`,
);
}
const msg = parseLine(line); const msg = parseLine(line);
if (!msg) return; if (!msg) return;
// Log init system event so we can see MCP server status / model on stderr
if (
msg.type === "system" &&
(msg as { subtype?: string }).subtype === "init"
) {
const init = msg as unknown as {
session_id?: string;
mcp_servers?: Array<{ name: string; status: string }>;
model?: string;
permissionMode?: string;
};
const mcps = (init.mcp_servers ?? [])
.map((s) => `${s.name}=${s.status}`)
.join(",");
console.error(
`[pi-claude-cli] init pid=${procPid} session=${init.session_id ?? "?"} model=${init.model ?? "?"} permissionMode=${init.permissionMode ?? "?"} mcp=[${mcps}]`,
);
}
if (msg.type === "stream_event") { if (msg.type === "stream_event") {
// Only forward top-level events to pi's event bridge. // Only forward top-level events to pi's event bridge.
// Sub-agent events (parent_tool_use_id !== null) are internal to the CLI. // Sub-agent events (parent_tool_use_id !== null) are internal to the CLI.
@@ -306,11 +268,6 @@ export function streamViaCli(
msg.event.content_block?.type === "tool_use" msg.event.content_block?.type === "tool_use"
) { ) {
const toolName = msg.event.content_block.name; const toolName = msg.event.content_block.name;
if (toolName) {
console.error(
`[pi-claude-cli] tool_use pid=${procPid} name=${toolName} piKnown=${isPiKnownClaudeTool(toolName)}`,
);
}
if (toolName && isPiKnownClaudeTool(toolName)) { if (toolName && isPiKnownClaudeTool(toolName)) {
// Built-in tool (Read/Write/etc.) OR custom MCP tool (mcp__custom-tools__*) // Built-in tool (Read/Write/etc.) OR custom MCP tool (mcp__custom-tools__*)
// Internal Claude Code tools (ToolSearch, Task, etc.) are excluded // Internal Claude Code tools (ToolSearch, Task, etc.) are excluded
@@ -325,9 +282,6 @@ export function streamViaCli(
msg.event.type === "message_stop" && msg.event.type === "message_stop" &&
sawBuiltInOrCustomTool sawBuiltInOrCustomTool
) { ) {
console.error(
`[pi-claude-cli] break-early pid=${procPid} elapsedMs=${Date.now() - spawnTime} lines=${lineCount}`,
);
broken = true; // Set guard BEFORE rl.close() to prevent buffered lines broken = true; // Set guard BEFORE rl.close() to prevent buffered lines
clearTimeout(inactivityTimer); clearTimeout(inactivityTimer);
// Pi will execute these tools. Kill subprocess to prevent CLI from executing them. // Pi will execute these tools. Kill subprocess to prevent CLI from executing them.