@moda-ai/cli 1.39.0 → 1.41.0

package/dist/{cli-a7ypptsf.js → cli-pr6s56fc.js}

@@ -5595,7 +5595,7 @@ async function runPromptAb(flags, profileOptions, context) {
   const baselineSource = flags.baseline || flags["baseline-file"] || flags["baseline-key"];
   const candidateSource = flags.candidate || flags["candidate-file"] || flags["candidate-key"];
   if (!baselineSource || !candidateSource) {
-    throw new Error("Usage: moda prompts ab --baseline=<file|key> --candidate=<file|key> [--auto-generate [--cases=1-100]|--set-id=ID|--traces=id1,id2 (max 100)] [--seeds=1-10] [--model=<slug>] [--yes] (legacy alias: --conversations=)");
+    throw new Error("Usage: moda prompts ab --baseline=<file|key> --candidate=<file|key> [--auto-generate [--cases=1-100]|--set-id=ID|--traces=id1,id2[@msg_index] (max 100)] [--seeds=1-10] [--model=<slug>] [--yes] (legacy alias: --conversations=)");
   }
   assertArmSourceShape(baselineSource, "baseline");
   assertArmSourceShape(candidateSource, "candidate");
@@ -5637,6 +5637,8 @@ async function runPromptAb(flags, profileOptions, context) {
       ...plan.kind === "existing" ? replayCasePin(plan.caseIds) : {}
     })
   }, profileOptions, { timeoutMs: 120000, retries: 0 });
+  if (context.outputMode === "human")
+    warnIfQueued(enqueue);
   const noWait = flags.wait === "false" || flags["no-wait"] === "true";
   if (noWait) {
     context.output.writeData({
@@ -5658,7 +5660,7 @@ async function runPromptAb(flags, profileOptions, context) {
       break;
     }
     if (context.outputMode === "human") {
-      process.stderr.write(`Waiting for replay run (${status || "pending"})...
+      process.stderr.write(`Waiting for replay run (${waitLabel(latest.run)})...
 `);
     }
     await sleep(pollIntervalMs);
@@ -5680,6 +5682,23 @@ async function runPromptAb(flags, profileOptions, context) {
   }
   context.output.writeData(result);
 }
+function warnIfQueued(enqueue) {
+  const { runsAhead, maxActiveRuns } = enqueue;
+  if (typeof runsAhead !== "number" || typeof maxActiveRuns !== "number")
+    return;
+  if (runsAhead < maxActiveRuns)
+    return;
+  process.stderr.write(`Note: ${runsAhead} replay runs for this tenant are already running or queued ` + `(${maxActiveRuns} at a time); run ${enqueue.runId} waits in the queue until a slot frees.
+`);
+}
+function waitLabel(run) {
+  if (!run?.status)
+    return "pending";
+  if (run.status === "accepted") {
+    return typeof run.runsAhead === "number" ? `queued, ${run.runsAhead} run${run.runsAhead === 1 ? "" : "s"} ahead` : "queued";
+  }
+  return run.status;
+}
 var FILE_ARM_RE = /[\\/]|\.(?:md|txt|json|ya?ml|jinja2?|j2|hbs|tmpl|prompt|[cm]?[jt]sx?|py)$/i;
 var PROMPT_KEY_RE = /^[^\s"'`\x00-\x1f\x7f]+$/;
 function assertArmSourceShape(source, label) {
@@ -5858,6 +5877,12 @@ async function ensureReplaySet(plan, flags, tenantId, profileOptions) {
   }
   return generated.id;
 }
+function parseTraceRef(ref) {
+  const match = /^(.+)@(\d+)$/.exec(ref.trim());
+  if (!match)
+    return { conversationId: ref.trim() };
+  return { conversationId: match[1], startMsgIndex: Number(match[2]) };
+}
 async function createSetFromConversations(conversationIds, flags, tenantId, profileOptions) {
   const name = (flags.name ?? flags["set-name"] ?? `Prompt A/B ${conversationIds.length} trace(s)`).trim();
   const created = await callControlAPI(`/tenants/${encodeURIComponent(tenantId)}/replay-sets`, {
@@ -5865,7 +5890,8 @@ async function createSetFromConversations(conversationIds, flags, tenantId, prof
     body: JSON.stringify({ name, description: "Created by moda prompts ab" })
   }, profileOptions);
   let position = 0;
-  for (const conversationId of conversationIds) {
+  for (const ref of conversationIds) {
+    const { conversationId, startMsgIndex } = parseTraceRef(ref);
     const scenario = await loadScenarioFromConversation(conversationId);
     await callControlAPI(`/tenants/${encodeURIComponent(tenantId)}/replay-sets/${encodeURIComponent(created.id)}/cases`, {
       method: "POST",
@@ -5873,6 +5899,7 @@ async function createSetFromConversations(conversationIds, flags, tenantId, prof
         title: conversationId,
         scenario,
         sourceConversationId: conversationId,
+        ...startMsgIndex !== undefined ? { sourceStartMsgIndex: startMsgIndex } : {},
         successCriteria: [
           "The agent understands the user request",
           "The agent uses tools appropriately when needed",
@@ -5975,6 +6002,8 @@ async function enqueueAndPollComparison(args, profileOptions) {
       ...replayCasePin(args.caseIds)
     })
   }, profileOptions, { timeoutMs: 120000, retries: 0 });
+  if (args.onWaiting)
+    warnIfQueued(enqueue);
   const timeoutMs = args.timeoutMs ?? DEFAULT_WAIT_TIMEOUT_MS;
   const pollIntervalMs = args.pollIntervalMs ?? DEFAULT_POLL_INTERVAL_MS;
   const deadline = Date.now() + timeoutMs;
@@ -5986,7 +6015,7 @@ async function enqueueAndPollComparison(args, profileOptions) {
     if (TERMINAL_RUN_STATUSES.has(status) && runId === enqueue.runId) {
       return latest;
     }
-    args.onWaiting?.(status || "pending");
+    args.onWaiting?.(waitLabel(latest.run));
     await sleep(pollIntervalMs);
   }
   if (!latest?.run || latest.run.runId !== enqueue.runId || !TERMINAL_RUN_STATUSES.has(latest.run.status)) {

package/dist/cli.js

@@ -8,7 +8,7 @@ import {
   runPromptsCommand,
   runSkillsCommand,
   runStatusCommand
-} from "./cli-a7ypptsf.js";
+} from "./cli-pr6s56fc.js";
 import {
   ApiError,
   HARNESS_REPORT_APPROVAL_PATH,
@@ -7296,6 +7296,7 @@ Examples:
   moda prompts sync
   moda prompts promote support.triage --label=prod --version=pver_abc123
   moda prompts ab --baseline=prompts/agent.prompt.md --candidate=prompts/agent-v2.prompt.md --traces=conv_1,conv_2
+  moda prompts ab --baseline=prompts/agent.prompt.md --candidate=prompts/agent-v2.prompt.md --traces=conv_1@412   # replay from msg 412
   moda skills pull
   moda fixes
   moda fix start <problem_id> --wait
@@ -9168,7 +9169,7 @@ var commandRegistry = createCommandRegistry([
     resetApiRequestCountBeforeRun: true,
     telemetry: "result",
     handler: async (context) => {
-      const { runInit } = await import("./index-qgc37efj.js");
+      const { runInit } = await import("./index-ah6yf5x2.js");
       if (context.outputMode === "agent-stream") {
         context.output.writeEvent({
           event: "started",

package/dist/{index-qgc37efj.js → index-ah6yf5x2.js}

@@ -3,7 +3,7 @@ import {
   initPrompts,
   runPromptSync,
   runSkillSync
-} from "./cli-a7ypptsf.js";
+} from "./cli-pr6s56fc.js";
 import {
   codingAgentDisplayName,
   describeCodingAgentEvent,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@moda-ai/cli",
-  "version": "1.39.0",
+  "version": "1.41.0",
   "description": "CLI for Moda - AI agent analytics and observability",
   "type": "module",
   "bin": {

package/skills/integration/index.json

@@ -1,7 +1,7 @@
 {
   "schema_version": "moda.skill_index.v1",
-  "bundled_at": "2026-10-05T22:35:48.948Z",
-  "cli_version": "1.39.0",
+  "bundled_at": "2026-10-09T01:07:28.008Z",
+  "cli_version": "1.41.0",
   "skills": [
     {
       "id": "integration-cloudflare-think",

package/skills/moda-cli/SKILL.md

@@ -759,7 +759,9 @@ anything they compute planned playouts (cases × seeds × 2 arms) and print it
 (`plannedPlayouts`). Above 200, or when an existing set's case count can't be
 read, they fail (exit 1) with the number and the exact command to re-run with `--yes`.
 Only add `--yes` when the user has agreed to that spend. Limits: `--cases`
-default 5, max 100; `--seeds` default 3, max 10; `--traces` at most 100;
+default 5, max 100; `--seeds` default 3, max 10; `--traces` at most 100
+(`--traces=conv_id@412` replays that trace from msg_index 412, with the
+earlier turns as history);
 `--model` / `--assistant-model` must be one of `openai/gpt-5.6-luna`,
 `openai/gpt-4o-mini`, `openai/gpt-4o`, `anthropic/claude-sonnet-4-5`,
 `anthropic/claude-haiku-4-5`, `google/gemini-2.5-flash`,