@runtypelabs/cli 2.21.6 → 2.22.1

package/dist/index.js

@@ -15128,7 +15128,7 @@ function date4(params) {
 // ../../node_modules/.pnpm/zod@4.4.2/node_modules/zod/v4/classic/external.js
 config(en_default());
 
-// ../shared/dist/chunk-R2PFJC7N.mjs
+// ../shared/dist/chunk-IZI2WRCS.mjs
 var apiReleaseChannelSchema = external_exports.enum(["staging", "production"]);
 var API_ARTIFACT_SCRIPT_PATTERN = /^api-[a-z0-9][a-z0-9-]{0,62}$/;
 var apiArtifactScriptNameSchema = external_exports.string().regex(
@@ -15168,6 +15168,18 @@ var apiRoutingDocSchema = external_exports.object({
   message: "activeScript may only be null while legacyFallback is true",
   path: ["api", "activeScript"]
 });
+var edgeContextPayloadSchema = external_exports.object({
+  /** Resolved Runtype user id (`user_…`). */
+  userId: external_exports.string().min(1),
+  /** Resolved org id (`org_…`); null for a personal (no-org) credential. */
+  orgId: external_exports.string().min(1).nullable(),
+  /** Effective permission scopes for the credential (e.g. `["*"]`). */
+  scopes: external_exports.array(external_exports.string()),
+  /** Token id (one per mint) — replay/audit handle, never reused as a secret. */
+  jti: external_exports.string().min(1),
+  /** Expiry as epoch MILLISECONDS; the token is rejected once `exp <= now`. */
+  exp: external_exports.number().int().positive()
+});
 
 // ../shared/dist/chunk-KR4LFVFE.mjs
 function getNestedValue(obj, path18) {
@@ -36768,36 +36780,36 @@ var caseExpectedSchema = external_exports.object({
 });
 var DEFAULT_MODELS_FOR_NEW_ACCOUNTS = [
   // Mixlayer models
-  { provider: "runtype", modelId: "qwen/qwen3.5-9b", isDefault: true },
-  { provider: "runtype", modelId: "qwen3.5-plus", isDefault: false },
+  { provider: "runtype", modelId: "qwen/qwen3.6-27b", isDefault: false },
+  { provider: "runtype", modelId: "qwen3.6-plus", isDefault: false },
   // General Compute (platform key) - explicit provider prefix, single-provider model
   { provider: "runtype", modelId: "general-compute/minimax-m2.7", isDefault: false },
   // Routed models - benefit from automatic provider optimization
   // OpenAI models (routed via OpenAI direct, with Vercel fallback)
-  { provider: "runtype", modelId: "gpt-5.4", isDefault: false },
+  { provider: "runtype", modelId: "gpt-5.5", isDefault: false },
   { provider: "runtype", modelId: "gpt-5.4-mini", isDefault: false },
   { provider: "runtype", modelId: "gpt-5.4-nano", isDefault: false },
   // Anthropic models (routed via Anthropic direct, with Vertex/Bedrock fallback)
-  { provider: "runtype", modelId: "claude-sonnet-4-6", isDefault: false },
   { provider: "runtype", modelId: "claude-haiku-4-5", isDefault: false },
-  { provider: "runtype", modelId: "claude-opus-4-6", isDefault: false },
+  { provider: "runtype", modelId: "claude-sonnet-4-6", isDefault: false },
+  { provider: "runtype", modelId: "claude-opus-4-8", isDefault: false },
   // Google models (routed via Google direct, with Vertex/Vercel fallback)
   { provider: "runtype", modelId: "gemini-3-1-flash-lite", isDefault: false },
+  { provider: "runtype", modelId: "gemini-3.5-flash", isDefault: false },
   { provider: "runtype", modelId: "gemini-3-1-flash-image", isDefault: false },
   { provider: "runtype", modelId: "gemini-3.1-pro", isDefault: false },
-  // xAI models (routed via Vercel, with xAI fallback when available)
-  { provider: "runtype", modelId: "grok-4.3", isDefault: false },
-  { provider: "runtype", modelId: "grok-4.1-fast-non-reasoning", isDefault: false },
   // Inception models (routed via Vercel)
   { provider: "runtype", modelId: "mercury-2", isDefault: false },
-  // Moonshot models (routed via Cloudflare Workers AI)
-  { provider: "runtype", modelId: "kimi-k2.6", isDefault: false },
-  // NVIDIA models (routed via Cloudflare Workers AI)
-  { provider: "runtype", modelId: "nemotron-3-120b-a12b", isDefault: false },
+  // Moonshot models (routed via Vercel)
+  // Kimi K2.6 is the default model: a reasoning + tool-use + vision model
+  // already trusted as the product-generator default.
+  { provider: "runtype", modelId: "kimi-k2.6", isDefault: true },
+  // NVIDIA Nemotron Ultra (reasoning + tool-use flagship, routed via Vercel)
+  { provider: "runtype", modelId: "nemotron-3-ultra-550b-a55b", isDefault: false },
   // Google models (routed via Cloudflare Workers AI)
   { provider: "runtype", modelId: "gemma-4-26b-a4b-it", isDefault: false }
 ];
-var DEFAULT_MODEL_ID = DEFAULT_MODELS_FOR_NEW_ACCOUNTS.find((m2) => m2.isDefault)?.modelId ?? "qwen/qwen3.5-9b";
+var DEFAULT_MODEL_ID = DEFAULT_MODELS_FOR_NEW_ACCOUNTS.find((m2) => m2.isDefault)?.modelId ?? "kimi-k2.6";
 var userProfileFeaturesSchema = external_exports.object({
   enableAgentSkills: external_exports.boolean(),
   // Routed model id the product generator dispatches with. Driven by the
@@ -52726,6 +52738,32 @@ function listForkCandidates(log) {
   return candidates;
 }
 
+// src/marathon/steering-queue.ts
+function createSteeringQueue(options = {}) {
+  const items = [];
+  const notify = () => {
+    options.onSizeChange?.(items.length);
+  };
+  return {
+    push(message) {
+      items.push(message);
+      notify();
+    },
+    drain() {
+      if (items.length === 0) return [];
+      const drained = items.splice(0);
+      notify();
+      return drained;
+    },
+    size() {
+      return items.length;
+    },
+    isEmpty() {
+      return items.length === 0;
+    }
+  };
+}
+
 // src/ink/marathon/CheckpointPrompt.tsx
 import { useState as useState26, useEffect as useEffect21, useRef as useRef8 } from "react";
 import { Box as Box23, Text as Text26 } from "ink";
@@ -55133,14 +55171,12 @@ function MarathonApp({
       getCallbacks: () => callbacks,
       getState: () => getHydratedState(),
       drainSteeringQueue: () => {
-        const drained = steeringQueueRef.current.splice(0);
-        if (drained.length > 0) setQueuedSteerCount(0);
+        const drained = steeringQueue.drain();
         return drained.length > 0 ? drained : void 0;
       },
-      hasQueuedSteering: () => steeringQueueRef.current.length > 0,
+      hasQueuedSteering: () => !steeringQueue.isEmpty(),
       drainFollowUpQueue: () => {
-        const drained = followUpQueueRef.current.splice(0);
-        if (drained.length > 0) setQueuedFollowUpCount(0);
+        const drained = followUpQueue.drain();
         return drained.length > 0 ? drained : void 0;
       },
       appendSessionSnapshot: (snapshot) => {
@@ -55222,10 +55258,18 @@ function MarathonApp({
       setEscAbortArmed(false);
     }
   }, [isAgentWorkingPhase, escAbortArmed]);
-  const steeringQueueRef = useRef9([]);
-  const followUpQueueRef = useRef9([]);
   const [queuedSteerCount, setQueuedSteerCount] = useState31(0);
   const [queuedFollowUpCount, setQueuedFollowUpCount] = useState31(0);
+  const steeringQueueHolder = useRef9(null);
+  if (steeringQueueHolder.current === null) {
+    steeringQueueHolder.current = createSteeringQueue({ onSizeChange: setQueuedSteerCount });
+  }
+  const steeringQueue = steeringQueueHolder.current;
+  const followUpQueueHolder = useRef9(null);
+  if (followUpQueueHolder.current === null) {
+    followUpQueueHolder.current = createSteeringQueue({ onSizeChange: setQueuedFollowUpCount });
+  }
+  const followUpQueue = followUpQueueHolder.current;
   const [showSteerComposer, setShowSteerComposer] = useState31(false);
   const [steerDraft, setSteerDraft] = useState31("");
   const [activeScreen, setActiveScreen] = useState31("overview");
@@ -55724,8 +55768,7 @@ function MarathonApp({
           if (escAbortTimeout.current) clearTimeout(escAbortTimeout.current);
           setEscAbortArmed(false);
           if (onAbortTurn?.() && !noCheckpoint) {
-            const queued = steeringQueueRef.current.splice(0);
-            setQueuedSteerCount(0);
+            const queued = steeringQueue.drain();
             const restored = [...queued, steerDraft].map((text) => text.trim()).filter(Boolean).join("\n\n");
             setSteerDraft("");
             setCheckpointInitialMessage(restored || void 0);
@@ -56457,11 +56500,9 @@ function MarathonApp({
             initialDraft: steerDraft,
             onSubmit: (message, kind) => {
               if (kind === "steer") {
-                steeringQueueRef.current.push(message);
-                setQueuedSteerCount(steeringQueueRef.current.length);
+                steeringQueue.push(message);
               } else {
-                followUpQueueRef.current.push(message);
-                setQueuedFollowUpCount(followUpQueueRef.current.length);
+                followUpQueue.push(message);
               }
               setSteerDraft("");
               setShowSteerComposer(false);
@@ -59455,7 +59496,40 @@ function resolveModelForPhase(phase, cliOverrides, milestoneModels) {
   }
   return cliOverrides.defaultModel;
 }
-function resolveErrorHandlingForPhase(phase, cliFallbackModel, milestoneFallbackModels) {
+function resolveMaxTokensForPhase(phase, cliMaxTokens, milestoneMaxTokens, defaultMaxTokens) {
+  if (phase && milestoneMaxTokens?.[phase] !== void 0) return milestoneMaxTokens[phase];
+  if (cliMaxTokens !== void 0) return cliMaxTokens;
+  return defaultMaxTokens;
+}
+function resolveTemperatureForPhase(phase, cliTemperature, milestoneTemperature, defaultTemperature) {
+  if (phase && milestoneTemperature?.[phase] !== void 0) return milestoneTemperature[phase];
+  if (cliTemperature !== void 0) return cliTemperature;
+  return defaultTemperature;
+}
+function parseMaxTokensFlag(raw) {
+  if (raw === void 0) return void 0;
+  const value = Number(raw);
+  if (!Number.isInteger(value) || value <= 0) {
+    throw new Error(`Invalid --max-tokens "${raw}": expected a positive integer`);
+  }
+  return value;
+}
+function parseTemperatureFlag(raw) {
+  if (raw === void 0) return void 0;
+  const value = Number(raw);
+  if (!Number.isFinite(value) || value < 0 || value > 2) {
+    throw new Error(`Invalid --temperature "${raw}": expected a number between 0 and 2`);
+  }
+  return value;
+}
+function resolveFallbackOnEmptyForPhase(phase, milestoneFallbackOnEmpty, defaultFallbackOnEmpty) {
+  if (phase && milestoneFallbackOnEmpty?.[phase] !== void 0) {
+    return milestoneFallbackOnEmpty[phase];
+  }
+  return defaultFallbackOnEmpty ?? false;
+}
+function resolveErrorHandlingForPhase(phase, cliFallbackModel, milestoneFallbackModels, fallbackOnEmpty) {
+  const triggers = fallbackOnEmpty ? [{ type: "empty-output" }, { type: "error" }] : void 0;
   const phaseFallbacks = phase ? milestoneFallbackModels?.[phase] : void 0;
   if (phaseFallbacks?.length) {
     return {
@@ -59468,7 +59542,8 @@ function resolveErrorHandlingForPhase(phase, cliFallbackModel, milestoneFallback
           ...fb.temperature !== void 0 ? { temperature: fb.temperature } : {},
           ...fb.maxTokens !== void 0 ? { maxTokens: fb.maxTokens } : {}
         }))
-      ]
+      ],
+      ...triggers ? { triggers } : {}
     };
   }
   if (cliFallbackModel) {
@@ -59477,7 +59552,8 @@ function resolveErrorHandlingForPhase(phase, cliFallbackModel, milestoneFallback
       fallbacks: [
         { type: "retry", delay: 5e3 },
         { type: "model", model: cliFallbackModel }
-      ]
+      ],
+      ...triggers ? { triggers } : {}
     };
   }
   return void 0;
@@ -59632,6 +59708,15 @@ function collectPlaybookWarnings(config3) {
       );
     }
   }
+  const emptyDefault = config3.fallbackOnEmpty === true;
+  for (const m2 of config3.milestones) {
+    const enabled = m2.fallbackOnEmpty ?? emptyDefault;
+    if (enabled && !m2.fallbackModels?.length) {
+      warnings.push(
+        `Playbook '${config3.name}': milestone '${m2.name}' enables fallbackOnEmpty but defines no fallbackModels \u2014 empty output has nothing to fall back to.`
+      );
+    }
+  }
   return warnings;
 }
 var PLUGIN_EXTENSIONS = /* @__PURE__ */ new Set([".js", ".mjs", ".cjs"]);
@@ -59709,17 +59794,29 @@ async function loadPlaybook(nameOrPath, cwd) {
     });
     const milestoneModels = {};
     const milestoneFallbackModels = {};
+    const milestoneMaxTokens = {};
+    const milestoneTemperature = {};
+    const milestoneFallbackOnEmpty = {};
     for (const m2 of config3.milestones) {
       if (m2.model) milestoneModels[m2.name] = m2.model;
       if (m2.fallbackModels?.length) {
         milestoneFallbackModels[m2.name] = m2.fallbackModels.map(normalizeFallbackModel);
       }
+      if (m2.maxTokens !== void 0) milestoneMaxTokens[m2.name] = m2.maxTokens;
+      if (m2.temperature !== void 0) milestoneTemperature[m2.name] = m2.temperature;
+      if (m2.fallbackOnEmpty !== void 0) milestoneFallbackOnEmpty[m2.name] = m2.fallbackOnEmpty;
     }
     return {
       workflow,
       milestones: config3.milestones.map((m2) => m2.name),
       milestoneModels: Object.keys(milestoneModels).length > 0 ? milestoneModels : void 0,
       milestoneFallbackModels: Object.keys(milestoneFallbackModels).length > 0 ? milestoneFallbackModels : void 0,
+      milestoneMaxTokens: Object.keys(milestoneMaxTokens).length > 0 ? milestoneMaxTokens : void 0,
+      milestoneTemperature: Object.keys(milestoneTemperature).length > 0 ? milestoneTemperature : void 0,
+      milestoneFallbackOnEmpty: Object.keys(milestoneFallbackOnEmpty).length > 0 ? milestoneFallbackOnEmpty : void 0,
+      ...config3.maxTokens !== void 0 ? { defaultMaxTokens: config3.maxTokens } : {},
+      ...config3.temperature !== void 0 ? { defaultTemperature: config3.temperature } : {},
+      ...config3.fallbackOnEmpty !== void 0 ? { defaultFallbackOnEmpty: config3.fallbackOnEmpty } : {},
       verification: config3.verification,
       rules: config3.rules,
       policy: config3.policy,
@@ -60282,6 +60379,12 @@ async function taskAction(agent, options) {
   let playbookMilestones;
   let playbookMilestoneModels;
   let playbookMilestoneFallbackModels;
+  let playbookMilestoneMaxTokens;
+  let playbookMilestoneTemperature;
+  let playbookMilestoneFallbackOnEmpty;
+  let playbookDefaultMaxTokens;
+  let playbookDefaultTemperature;
+  let playbookDefaultFallbackOnEmpty;
   let playbookPolicy;
   if (options.playbook) {
     const result = await loadPlaybook(options.playbook);
@@ -60289,6 +60392,12 @@ async function taskAction(agent, options) {
     playbookMilestones = result.milestones;
     playbookMilestoneModels = result.milestoneModels;
     playbookMilestoneFallbackModels = result.milestoneFallbackModels;
+    playbookMilestoneMaxTokens = result.milestoneMaxTokens;
+    playbookMilestoneTemperature = result.milestoneTemperature;
+    playbookMilestoneFallbackOnEmpty = result.milestoneFallbackOnEmpty;
+    playbookDefaultMaxTokens = result.defaultMaxTokens;
+    playbookDefaultTemperature = result.defaultTemperature;
+    playbookDefaultFallbackOnEmpty = result.defaultFallbackOnEmpty;
     playbookPolicy = result.policy;
     for (const warning of result.warnings) {
       if (useStartupShell) {
@@ -60300,6 +60409,8 @@ async function taskAction(agent, options) {
   } else {
     playbookPolicy = void 0;
   }
+  const cliMaxTokens = parseMaxTokensFlag(options.maxTokens);
+  const cliTemperature = parseTemperatureFlag(options.temperature);
   if (useStartupShell && !options.model?.trim()) {
     if (playbookMilestoneModels && Object.keys(playbookMilestoneModels).length > 0 && startupShellRef.current) {
       let editableModels = { ...playbookMilestoneModels };
@@ -60439,7 +60550,12 @@ ${rulesContext}`;
     const initialErrorHandling = resolveErrorHandlingForPhase(
       currentPhase,
       options.fallbackModel,
-      playbookMilestoneFallbackModels
+      playbookMilestoneFallbackModels,
+      resolveFallbackOnEmptyForPhase(
+        currentPhase,
+        playbookMilestoneFallbackOnEmpty,
+        playbookDefaultFallbackOnEmpty
+      )
     );
     if (initialErrorHandling) {
       await client.agents.update(agentId, { config: { errorHandling: initialErrorHandling } }).catch(() => {
@@ -60773,6 +60889,18 @@ Saving state... done. Session saved to ${filePath}`);
         },
         playbookMilestoneModels
       );
+      const phaseMaxTokens = resolveMaxTokensForPhase(
+        resumeState?.workflowPhase,
+        cliMaxTokens,
+        playbookMilestoneMaxTokens,
+        playbookDefaultMaxTokens
+      );
+      const phaseTemperature = resolveTemperatureForPhase(
+        resumeState?.workflowPhase,
+        cliTemperature,
+        playbookMilestoneTemperature,
+        playbookDefaultTemperature
+      );
       const effectiveModelForContext = phaseModel || options.model || agentConfigModel || defaultConfiguredModel;
       const compactStrategy = resolveCompactStrategyForModel(effectiveModelForContext);
       const contextLimitTokens = resolveContextLimitForModel(effectiveModelForContext);
@@ -60787,6 +60915,8 @@ Saving state... done. Session saved to ${filePath}`);
         maxSessions: remainingSessions - accumulatedSessions,
         maxCost: currentRemainingCost,
         model: phaseModel || options.model,
+        ...phaseMaxTokens !== void 0 ? { maxTokens: phaseMaxTokens } : {},
+        ...phaseTemperature !== void 0 ? { temperature: phaseTemperature } : {},
         abortSignal: turnAbortController.signal,
         reasoning: !options.noReasoning,
         debugMode: options.debug ?? true,
@@ -60958,7 +61088,12 @@ Saving state... done. Session saved to ${filePath}`);
               const newErrorHandling = resolveErrorHandlingForPhase(
                 resumeState.workflowPhase,
                 options.fallbackModel,
-                playbookMilestoneFallbackModels
+                playbookMilestoneFallbackModels,
+                resolveFallbackOnEmptyForPhase(
+                  resumeState.workflowPhase,
+                  playbookMilestoneFallbackOnEmpty,
+                  playbookDefaultFallbackOnEmpty
+                )
               );
               client.agents.update(agentId, {
                 config: { errorHandling: newErrorHandling ?? null }
@@ -61468,7 +61603,13 @@ function resolveSandboxWorkflowSelection(message, sandboxProvider, resumeState)
   };
 }
 function applyTaskOptions(cmd) {
-  return cmd.argument("<agent>", "Agent ID or name").option("-g, --goal <text>", "Goal message for the agent").option("--max-sessions <n>", "Maximum sessions", "50").option("--max-cost <n>", "Budget in USD").option("--model <modelId>", "Model ID to use (overrides agent config)").option("--name <name>", "Task name (used for state file, defaults to agent name)").option("--session <name>", "Resume a specific session by name").option(
+  return cmd.argument("<agent>", "Agent ID or name").option("-g, --goal <text>", "Goal message for the agent").option("--max-sessions <n>", "Maximum sessions", "50").option("--max-cost <n>", "Budget in USD").option("--model <modelId>", "Model ID to use (overrides agent config)").option(
+    "--max-tokens <n>",
+    "Max output-token budget per session (overrides agent config; playbook milestones can override per phase)"
+  ).option(
+    "--temperature <n>",
+    "Sampling temperature 0-2 per session (overrides agent config; playbook milestones can override per phase)"
+  ).option("--name <name>", "Task name (used for state file, defaults to agent name)").option("--session <name>", "Resume a specific session by name").option(
     "--state-dir <path>",
     "Directory for state files (default: ~/.runtype/projects/<hash>/marathons/)"
   ).option(

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@runtypelabs/cli",
-  "version": "2.21.6",
+  "version": "2.22.1",
   "description": "Command-line interface for Runtype AI platform",
   "type": "module",
   "main": "dist/index.js",
@@ -24,7 +24,7 @@
     "rosie-skills": "0.8.1",
     "yaml": "^2.9.0",
     "@runtypelabs/ink-components": "0.3.2",
-    "@runtypelabs/sdk": "4.18.0",
+    "@runtypelabs/sdk": "4.19.1",
     "@runtypelabs/terminal-animations": "0.2.1"
   },
   "devDependencies": {
@@ -39,7 +39,7 @@
     "tsx": "^4.7.1",
     "typescript": "^5.3.3",
     "vitest": "^4.1.0",
-    "@runtypelabs/shared": "1.33.0"
+    "@runtypelabs/shared": "1.33.2"
   },
   "engines": {
     "node": ">=22.0.0"