ai-whisper 0.5.4 → 0.5.6

package/README.md

@@ -1,6 +1,6 @@
 # ai-whisper
 
-ai-whisper turns two coding agents — Claude and Codex — into a terminal-native pair that hand work back and forth under a single baton, so one agent implements while the other reviews, and a structured workflow drives the loop to a finished, reviewed deliverable without a human babysitting every round.
+ai-whisper pairs two coding agents — mount any two of Claude, Codex, and ezio — into a terminal-native pair that hand work back and forth under a single baton, so one agent implements while the other reviews, and a structured workflow drives the loop to a finished, reviewed deliverable without a human babysitting every round.
 
 ## Magic moment
 
@@ -52,10 +52,11 @@ It is **not** for:
 
 ## Prerequisites
 
-ai-whisper drives the *real* Claude and Codex CLIs, so you need both installed and authenticated first:
+You pair any two of three agents — `claude`, `codex`, and `ezio`. ai-whisper drives the *real* Claude and Codex CLIs, so install and authenticate whichever of those two you plan to mount first; `ezio` is protocol-native and ships with ai-whisper, so it needs no separate CLI.
 
 - **[Claude Code CLI](https://claude.com/claude-code)** — the `claude` command, signed in.
 - **[Codex CLI](https://github.com/openai/codex)** — the `codex` command, signed in.
+- **ezio** *(optional)* — bundled with ai-whisper; mount it with `whisper collab mount ezio`, no separate install.
 - **Node.js 22+**.
 - **An LLM evaluator with credentials** — workflows are gated by it and refuse to start without it. See [Evaluator configuration](docs/evaluator-configuration.md).
 - **tmux** *(optional)* — only for `whisper collab start`, which auto-arranges both agents into panes. The mount flow below does not need it.
@@ -110,9 +111,9 @@ A run that stops short usually **escalates** — it does not crash. When the eva
 
 ## Core concepts
 
-ai-whisper is **not a swarm**. The agents never type at once — work moves by a single baton, one owner at a time. Mounted sessions are *real* Claude and Codex sessions in your terminal, and those sessions are the source of truth. Autonomy is supervised: every handoff, verdict, and round is inspectable, and runs are resumable rather than fire-and-forget. Work is organized as structured workflows — explicit loops and state transitions, not a free-form chat.
+ai-whisper is **not a swarm**. The agents never type at once — work moves by a single baton, one owner at a time. Mounted sessions are *real* agent sessions in your terminal — Claude or Codex CLIs, or ezio — and those sessions are the source of truth. Autonomy is supervised: every handoff, verdict, and round is inspectable, and runs are resumable rather than fire-and-forget. Work is organized as structured workflows — explicit loops and state transitions, not a free-form chat.
 
-Claude and Codex are supported today; the architecture is provider-agnostic by design, so other coding-agent CLIs can be added behind the same relay.
+Claude, Codex, and ezio are supported today — you mount any two of them; the architecture is provider-agnostic by design, so other coding-agent CLIs can be added behind the same relay.
 
 For the full mental model, read [Concepts](docs/concepts.md).
 

package/dist/bin/broker-daemon.js

@@ -106,6 +106,26 @@ var attachTargetModes = [
   "mount_current_tty"
 ];
 
+// ../shared/dist/repo-relative-path.js
+import { basename, isAbsolute, relative } from "node:path";
+function repoRelativePath(absPath, root) {
+  if (!absPath)
+    return "";
+  if (!root || !isAbsolute(absPath) || !isAbsolute(root))
+    return absPath;
+  const rel = relative(root, absPath);
+  if (rel === "" || rel.startsWith("..") || isAbsolute(rel))
+    return absPath;
+  return rel;
+}
+function displayArtifactPath(absPath, root) {
+  const trimmed = (absPath ?? "").trim();
+  if (!trimmed)
+    return null;
+  const rel = repoRelativePath(trimmed, root).trim();
+  return rel || basename(trimmed).trim() || null;
+}
+
 // ../shared/dist/broker-status.js
 import { z as z3 } from "zod";
 var brokerStatusSchema = z3.object({
@@ -1255,8 +1275,78 @@ function deleteEvaluatorDiagnosticsOlderThan(db, cutoffIso) {
   return result.changes;
 }
 
+// ../broker/dist/storage/repositories/relay-turn-event-diagnostics-repository.js
+var insertSeq = 0;
+function insertTurnEventDiagnostic(db, input) {
+  const eventId = `tevt_${input.receivedAt.replace(/[^0-9]/g, "")}_${(input.handoffId ?? input.workspaceId).slice(-8)}_${input.action}_${process.pid}_${(insertSeq++).toString(36)}`;
+  db.prepare(`INSERT INTO relay_turn_event_diagnostics (
+			event_id, received_at, provider, workspace_id, cwd,
+			session_or_thread_id, turn_id, workflow_active, collab_id, workflow_id,
+			chain_id, handoff_id, input_correlated, containment_score,
+			fidelity_verdict, defer_count, action, message_len, message_sample
+		) VALUES (
+			@eventId, @receivedAt, @provider, @workspaceId, @cwd,
+			@sessionOrThreadId, @turnId, @workflowActive, @collabId, @workflowId,
+			@chainId, @handoffId, @inputCorrelated, @containmentScore,
+			@fidelityVerdict, @deferCount, @action, @messageLen, @messageSample
+		)`).run({
+    eventId,
+    receivedAt: input.receivedAt,
+    provider: input.provider,
+    workspaceId: input.workspaceId,
+    cwd: input.cwd,
+    sessionOrThreadId: input.sessionOrThreadId,
+    turnId: input.turnId,
+    workflowActive: input.workflowActive ? 1 : 0,
+    collabId: input.collabId,
+    workflowId: input.workflowId,
+    chainId: input.chainId,
+    handoffId: input.handoffId,
+    inputCorrelated: input.inputCorrelated === null ? null : input.inputCorrelated ? 1 : 0,
+    containmentScore: input.containmentScore,
+    fidelityVerdict: input.fidelityVerdict,
+    deferCount: input.deferCount,
+    action: input.action,
+    messageLen: input.messageLen,
+    messageSample: input.messageSample
+  });
+  return { eventId };
+}
+function rowToRecord4(r) {
+  return {
+    eventId: r.event_id,
+    receivedAt: r.received_at,
+    provider: r.provider,
+    workspaceId: r.workspace_id,
+    cwd: r.cwd,
+    sessionOrThreadId: r.session_or_thread_id,
+    turnId: r.turn_id,
+    workflowActive: r.workflow_active === 1,
+    collabId: r.collab_id,
+    workflowId: r.workflow_id,
+    chainId: r.chain_id,
+    handoffId: r.handoff_id,
+    inputCorrelated: r.input_correlated === null ? null : r.input_correlated === 1,
+    containmentScore: r.containment_score,
+    fidelityVerdict: r.fidelity_verdict,
+    deferCount: r.defer_count,
+    action: r.action,
+    messageLen: r.message_len,
+    messageSample: r.message_sample
+  };
+}
+function listTurnEventDiagnosticsByCollab(db, collabId2, limit) {
+  const sql = "SELECT * FROM relay_turn_event_diagnostics WHERE collab_id = ? ORDER BY received_at DESC" + (limit !== null ? " LIMIT ?" : "");
+  const rows = limit !== null ? db.prepare(sql).all(collabId2, limit) : db.prepare(sql).all(collabId2);
+  return rows.map(rowToRecord4);
+}
+function deleteTurnEventDiagnosticsOlderThan(db, cutoffIso) {
+  const info = db.prepare("DELETE FROM relay_turn_event_diagnostics WHERE received_at < ?").run(cutoffIso);
+  return info.changes;
+}
+
 // ../broker/dist/storage/repositories/relay-handoff-repository.js
-function rowToRecord4(row) {
+function rowToRecord5(row) {
   return {
     handoffId: row.handoff_id,
     collabId: row.collab_id,
@@ -1293,7 +1383,7 @@ function queryRelayHandoff(db, handoffId) {
 			 FROM relay_handoff h
 			 JOIN collab c ON c.collab_id = h.collab_id
 			 WHERE h.handoff_id = ?`).get(handoffId);
-  return row ? rowToRecord4(row) : null;
+  return row ? rowToRecord5(row) : null;
 }
 function createRelayHandoffTxn(db, input) {
   return db.transaction(() => {
@@ -1445,7 +1535,7 @@ function queryLatestHandedBackHandoff(db, collabId2) {
 			 WHERE h.collab_id = ? AND h.status = 'handed_back'
 			 ORDER BY h.resolved_at DESC
 			 LIMIT 1`).get(collabId2);
-  return row ? rowToRecord4(row) : null;
+  return row ? rowToRecord5(row) : null;
 }
 function failRelayHandoffOnDisconnectTxn(db, input) {
   db.transaction(() => {
@@ -1516,7 +1606,7 @@ function listRelayHandoffsPendingOrchestration(db, collabId2) {
 			   AND h.status = 'handed_back'
 			   AND h.orchestrator_status = 'idle'
 			   AND (w.status IS NULL OR w.status <> 'paused')`).all(collabId2);
-  return rows.map(rowToRecord4);
+  return rows.map(rowToRecord5);
 }
 function createLoopRelayHandoffTxn(db, input) {
   return db.transaction(() => {
@@ -1712,7 +1802,7 @@ function getHandoffWithWorkflowMetaById(db, handoffId) {
 			 WHERE h.handoff_id = ?`).get(handoffId);
   if (!row)
     return null;
-  const base = rowToRecord4(row);
+  const base = rowToRecord5(row);
   return {
     ...base,
     handoffStep: row.handoff_step,
@@ -1811,7 +1901,7 @@ function cleanupOrchestrationOnShutdownTxn(db, input) {
 }
 
 // ../broker/dist/storage/repositories/dashboard-repository.js
-import { basename } from "node:path";
+import { basename as basename2 } from "node:path";
 function listActiveCollabSummaries(db, input) {
   const nowMs = Date.parse(input.now ?? (/* @__PURE__ */ new Date()).toISOString());
   const base = Number.isFinite(nowMs) ? nowMs : Date.now();
@@ -1855,7 +1945,7 @@ function buildCollabSummary(db, collabId2) {
 				   FROM collab WHERE collab_id = ?`).get(e.collabId);
     const wf = db.prepare(`SELECT workflow_id AS workflowId, workflow_type AS workflowType,
 				        name, status, current_phase_index AS currentPhaseIndex,
-				        created_at AS createdAt
+				        created_at AS createdAt, spec_path AS specPath
 				   FROM workflows WHERE collab_id = ?
 				  ORDER BY (status = 'running') DESC, created_at DESC
 				  LIMIT 1`).get(e.collabId);
@@ -1898,7 +1988,7 @@ function buildCollabSummary(db, collabId2) {
 				 ) ranked
 				 WHERE rn = 1
 				 ORDER BY agentType ASC`).all(e.collabId);
-    const label = wf?.name && wf.name.trim() || collab2?.displayName && collab2.displayName.trim() || (collab2?.workspaceRoot ? basename(collab2.workspaceRoot) : "") || e.collabId.slice(0, 12);
+    const label = wf?.name && wf.name.trim() || collab2?.displayName && collab2.displayName.trim() || (collab2?.workspaceRoot ? basename2(collab2.workspaceRoot) : "") || e.collabId.slice(0, 12);
     const runLastActRow = wf ? db.prepare(`SELECT COALESCE(MAX(last_activity_at), '') AS lastAct
 						   FROM relay_handoff
 						  WHERE collab_id = ? AND workflow_id = ?`).get(e.collabId, wf.workflowId) : db.prepare(`SELECT COALESCE(MAX(last_activity_at), '') AS lastAct
@@ -1924,6 +2014,7 @@ function buildCollabSummary(db, collabId2) {
       },
       sessions,
       workflowCreatedAt: wf?.createdAt ?? null,
+      specPath: wf ? displayArtifactPath(wf.specPath, collab2?.workspaceRoot ?? "") : null,
       lastActivityAt: runLastAct
     };
   }
@@ -1966,7 +2057,7 @@ function listRunCostRows(db, input) {
 import { randomUUID } from "node:crypto";
 
 // ../broker/dist/storage/repositories/workflow-phase-repository.js
-function rowToRecord5(row) {
+function rowToRecord6(row) {
   return {
     phaseRunId: row.phase_run_id,
     workflowId: row.workflow_id,
@@ -1985,7 +2076,7 @@ function insertWorkflowPhaseRun(db, input) {
 }
 function listPhaseRunsForWorkflow(db, workflowId) {
   const rows = db.prepare("SELECT * FROM workflow_phases WHERE workflow_id = ? ORDER BY started_at ASC").all(workflowId);
-  return rows.map(rowToRecord5);
+  return rows.map(rowToRecord6);
 }
 function closeWorkflowPhaseRun(db, input) {
   db.prepare("UPDATE workflow_phases SET ended_at = ?, outcome = ? WHERE phase_run_id = ?").run(input.now, input.outcome, input.phaseRunId);
@@ -1998,7 +2089,7 @@ function hasOpenPhaseRunForIndex(db, input) {
 }
 
 // ../broker/dist/storage/repositories/relay-chain-repository.js
-function rowToRecord6(row) {
+function rowToRecord7(row) {
   return {
     chainId: row.chain_id,
     collabId: row.collab_id,
@@ -2018,7 +2109,7 @@ function insertRelayChain(db, input) {
 }
 function getRelayChain(db, chainId) {
   const row = db.prepare("SELECT * FROM relay_chains WHERE chain_id = ?").get(chainId);
-  return row ? rowToRecord6(row) : null;
+  return row ? rowToRecord7(row) : null;
 }
 function incrementChainRound(db, input) {
   db.prepare("UPDATE relay_chains SET current_round = current_round + 1, updated_at = ? WHERE chain_id = ?").run(input.now, input.chainId);
@@ -2375,11 +2466,11 @@ function derivePlanPath(specPath, dateIso) {
   if (!/^\d{4}-\d{2}-\d{2}/.test(dateIso)) {
     throw new Error(`derivePlanPath: dateIso must start with YYYY-MM-DD, got "${dateIso}"`);
   }
-  const basename2 = specPath.split("/").pop() ?? "";
-  if (!basename2.endsWith("-design.md")) {
+  const basename3 = specPath.split("/").pop() ?? "";
+  if (!basename3.endsWith("-design.md")) {
     throw new Error(`derivePlanPath: specPath must end with "-design.md", got "${specPath}"`);
   }
-  const withoutExt = basename2.replace(/\.md$/, "");
+  const withoutExt = basename3.replace(/\.md$/, "");
   const slug = withoutExt.replace(/-design$/, "").replace(/^\d{4}-\d{2}-\d{2}-/, "");
   const date = dateIso.slice(0, 10);
   return `docs/superpowers/plans/${date}-${slug}.md`;
@@ -4170,6 +4261,15 @@ function createControlService(db, events) {
     sweepEvaluatorDiagnostics(input) {
       return deleteEvaluatorDiagnosticsOlderThan(db, input.cutoffIso);
     },
+    recordTurnEventDiagnostic(input) {
+      return insertTurnEventDiagnostic(db, input);
+    },
+    listTurnEventDiagnosticsByCollab(collabId2, limit) {
+      return listTurnEventDiagnosticsByCollab(db, collabId2, limit);
+    },
+    sweepTurnEventDiagnostics(input) {
+      return deleteTurnEventDiagnosticsOlderThan(db, input.cutoffIso);
+    },
     cleanupOrchestration(input) {
       return cleanupOrchestrationOnShutdownTxn(db, input);
     }
@@ -4310,7 +4410,7 @@ function sweepStaleCaptureLease(db, options = {}) {
 }
 
 // ../broker/dist/storage/apply-migrations.js
-var CURRENT_SCHEMA_VERSION = 5;
+var CURRENT_SCHEMA_VERSION = 6;
 var initMigrationSql = `
 CREATE TABLE IF NOT EXISTS broker_state (
   id INTEGER PRIMARY KEY CHECK (id = 1),
@@ -4825,6 +4925,37 @@ function runMigrationBody(db) {
   if (!brokerDaemonColumns.some((column) => column.name === "evaluator_status")) {
     db.exec("ALTER TABLE broker_daemon ADD COLUMN evaluator_status TEXT");
   }
+  db.exec(`
+		CREATE TABLE IF NOT EXISTS relay_turn_event_diagnostics (
+			event_id TEXT PRIMARY KEY,
+			received_at TEXT NOT NULL,
+			provider TEXT NOT NULL,
+			workspace_id TEXT NOT NULL,
+			cwd TEXT NOT NULL,
+			session_or_thread_id TEXT,
+			turn_id TEXT,
+			workflow_active INTEGER NOT NULL DEFAULT 0,
+			collab_id TEXT,
+			workflow_id TEXT,
+			chain_id TEXT,
+			handoff_id TEXT,
+			input_correlated INTEGER,
+			containment_score REAL,
+			fidelity_verdict TEXT NOT NULL DEFAULT 'n/a',
+			defer_count INTEGER NOT NULL DEFAULT 0,
+			action TEXT NOT NULL,
+			message_len INTEGER NOT NULL DEFAULT 0,
+			message_sample TEXT
+		);
+		CREATE INDEX IF NOT EXISTS idx_relay_turn_event_diag_collab_received
+			ON relay_turn_event_diagnostics (collab_id, received_at DESC);
+		CREATE INDEX IF NOT EXISTS idx_relay_turn_event_diag_handoff
+			ON relay_turn_event_diagnostics (handoff_id);
+		CREATE INDEX IF NOT EXISTS idx_relay_turn_event_diag_action
+			ON relay_turn_event_diagnostics (action);
+		CREATE INDEX IF NOT EXISTS idx_relay_turn_event_diag_received
+			ON relay_turn_event_diagnostics (received_at);
+	`);
 }
 
 // ../broker/dist/storage/open-database.js
@@ -4975,8 +5106,12 @@ function createDaemonHeartbeat(input) {
 }
 
 // ../broker/dist/runtime/diagnostics-sweep.js
+import { readdirSync, unlinkSync, statSync as statSync2 } from "node:fs";
+import { join as join2 } from "node:path";
 var DEFAULT_INTERVAL_MS = 60 * 60 * 1e3;
 var DEFAULT_RETENTION_DAYS = 30;
+var DEFAULT_EVENT_LOG_RETENTION_DAYS = 3;
+var TURN_EVENT_FILE_RE = /^turn-events-(\d{4}-\d{2}-\d{2})\.jsonl$/;
 function resolveIntervalMs(override) {
   if (override !== void 0)
     return override;
@@ -4993,6 +5128,38 @@ function resolveRetentionDays(override) {
     return Number(envVal);
   return DEFAULT_RETENTION_DAYS;
 }
+function resolveEventLogRetentionDays(override) {
+  if (override !== void 0)
+    return override;
+  const raw = process.env["AI_WHISPER_EVENT_LOG_RETENTION_DAYS"];
+  if (raw && Number.isFinite(Number(raw)) && Number(raw) > 0)
+    return Number(raw);
+  return DEFAULT_EVENT_LOG_RETENTION_DAYS;
+}
+function sweepTurnEventLogs(logsDir, cutoffDate) {
+  let entries;
+  try {
+    entries = readdirSync(logsDir);
+  } catch {
+    return [];
+  }
+  const removed = [];
+  for (const name of entries) {
+    const m = TURN_EVENT_FILE_RE.exec(name);
+    if (!m)
+      continue;
+    if (m[1] < cutoffDate) {
+      const full = join2(logsDir, name);
+      try {
+        statSync2(full);
+        unlinkSync(full);
+        removed.push(full);
+      } catch {
+      }
+    }
+  }
+  return removed;
+}
 function createDiagnosticsSweep(deps) {
   const intervalMs = resolveIntervalMs(deps.intervalMs);
   const retentionDays = resolveRetentionDays(deps.retentionDays);
@@ -5002,6 +5169,12 @@ function createDiagnosticsSweep(deps) {
     const cutoffIso = cutoff.toISOString();
     deps.broker.control.sweepCaptureDiagnostics({ cutoffIso });
     deps.broker.control.sweepEvaluatorDiagnostics({ cutoffIso });
+    const eventRetentionDays = resolveEventLogRetentionDays(deps.eventLogRetentionDays);
+    const eventCutoff = new Date(Date.now() - eventRetentionDays * 864e5);
+    deps.broker.control.sweepTurnEventDiagnostics({ cutoffIso: eventCutoff.toISOString() });
+    if (deps.logsDir) {
+      sweepTurnEventLogs(deps.logsDir, eventCutoff.toISOString().slice(0, 10));
+    }
   }
   return {
     start() {
@@ -5020,15 +5193,15 @@ function createDiagnosticsSweep(deps) {
 
 // ../broker/dist/runtime/ralph-setup.js
 import { mkdirSync, writeFileSync, existsSync as existsSync2 } from "node:fs";
-import { join as join2 } from "node:path";
+import { join as join3 } from "node:path";
 function ensureRalphWorkspace(workspaceRoot, workflowId) {
   const dir = ralphRunDir(workspaceRoot, workflowId);
   mkdirSync(dir, { recursive: true });
-  const ignorePath = join2(workspaceRoot, ".ai-whisper", ".gitignore");
+  const ignorePath = join3(workspaceRoot, ".ai-whisper", ".gitignore");
   if (!existsSync2(ignorePath)) {
     writeFileSync(ignorePath, "*\n");
   }
-  const ralphIgnorePath = join2(workspaceRoot, ".ai-whisper", "ralph", ".gitignore");
+  const ralphIgnorePath = join3(workspaceRoot, ".ai-whisper", "ralph", ".gitignore");
   if (!existsSync2(ralphIgnorePath)) {
     writeFileSync(ralphIgnorePath, "*\n");
   }
@@ -5037,15 +5210,15 @@ function ensureRalphWorkspace(workspaceRoot, workflowId) {
 
 // ../broker/dist/runtime/bugfix-setup.js
 import { mkdirSync as mkdirSync2, writeFileSync as writeFileSync2, existsSync as existsSync3 } from "node:fs";
-import { join as join3 } from "node:path";
+import { join as join4 } from "node:path";
 function ensureBugfixWorkspace(workspaceRoot, workflowId) {
   const dir = bugfixRunDir(workspaceRoot, workflowId);
   mkdirSync2(dir, { recursive: true });
-  const ignorePath = join3(workspaceRoot, ".ai-whisper", ".gitignore");
+  const ignorePath = join4(workspaceRoot, ".ai-whisper", ".gitignore");
   if (!existsSync3(ignorePath)) {
     writeFileSync2(ignorePath, "*\n");
   }
-  const bugfixIgnorePath = join3(workspaceRoot, ".ai-whisper", "bugfix", ".gitignore");
+  const bugfixIgnorePath = join4(workspaceRoot, ".ai-whisper", "bugfix", ".gitignore");
   if (!existsSync3(bugfixIgnorePath)) {
     writeFileSync2(bugfixIgnorePath, "*\n");
   }
@@ -6099,8 +6272,8 @@ function writeOwnPidToBrokerDaemon(db, input) {
 }
 
 // src/runtime/evaluator-config.ts
-import { readFileSync as readFileSync2, statSync as statSync2 } from "node:fs";
-import { join as join4 } from "node:path";
+import { readFileSync as readFileSync2, statSync as statSync3 } from "node:fs";
+import { join as join5 } from "node:path";
 
 // src/runtime/state-root.ts
 import os from "node:os";
@@ -6131,7 +6304,7 @@ function warnIfLoosePerms(path2) {
   if (process.platform === "win32") return;
   let mode;
   try {
-    mode = statSync2(path2).mode;
+    mode = statSync3(path2).mode;
   } catch {
     return;
   }
@@ -6159,15 +6332,15 @@ function loadEvaluatorConfig() {
   const root = getStateRoot();
   let dotenv = {};
   try {
-    dotenv = parseDotEnv(readFileSync2(join4(root, ".env"), "utf8"));
+    dotenv = parseDotEnv(readFileSync2(join5(root, ".env"), "utf8"));
   } catch (err) {
     if (err.code !== "ENOENT") throw err;
   }
   const envGet = (k) => process.env[k] ?? dotenv[k];
-  const authPath = join4(root, "auth.json");
+  const authPath = join5(root, "auth.json");
   warnIfLoosePerms(authPath);
   const auth = readJsonFile(authPath, "auth.json");
-  const config = readJsonFile(join4(root, "config.json"), "config.json");
+  const config = readJsonFile(join5(root, "config.json"), "config.json");
   const evalCfg = config?.evaluator ?? {};
   const evalOllama = evalCfg.ollama ?? {};
   const provider = (envGet("AI_WHISPER_EVALUATOR_PROVIDER") ?? evalCfg.provider) === "ollama" ? "ollama" : "anthropic";