@elevasis/sdk 0.5.16 → 0.5.17

package/dist/cli.cjs

@@ -40221,7 +40221,7 @@ var GPT5ConfigSchema = external_exports.object({
   apiKey: external_exports.string(),
   temperature: external_exports.literal(1),
   // Required to be exactly 1
-  maxTokens: external_exports.number().min(4e3).optional(),
+  maxOutputTokens: external_exports.number().min(4e3).optional(),
   topP: external_exports.number().min(0).max(1).optional(),
   modelOptions: GPT5OptionsSchema.optional()
 });
@@ -40230,7 +40230,7 @@ var MockConfigSchema = external_exports.object({
   provider: external_exports.enum(["mock"]),
   apiKey: external_exports.string(),
   temperature: external_exports.number().min(0).max(2).optional(),
-  maxTokens: external_exports.number().min(500).optional(),
+  maxOutputTokens: external_exports.number().min(500).optional(),
   topP: external_exports.number().min(0).max(1).optional(),
   modelOptions: external_exports.object({}).strict().optional()
   // No options supported
@@ -40250,7 +40250,7 @@ var OpenRouterConfigSchema = external_exports.object({
   provider: external_exports.literal("openrouter"),
   apiKey: external_exports.string(),
   temperature: external_exports.number().min(0).max(2).optional(),
-  maxTokens: external_exports.number().min(500).optional(),
+  maxOutputTokens: external_exports.number().min(500).optional(),
   topP: external_exports.number().min(0).max(1).optional(),
   modelOptions: OpenRouterOptionsSchema.optional()
 });
@@ -40264,7 +40264,7 @@ var GoogleConfigSchema = external_exports.object({
   apiKey: external_exports.string(),
   temperature: external_exports.literal(1).optional(),
   // Must be 1 for Gemini 3 (changing degrades performance)
-  maxTokens: external_exports.number().min(500).optional(),
+  maxOutputTokens: external_exports.number().min(500).optional(),
   topP: external_exports.number().min(0).max(1).optional(),
   modelOptions: GoogleOptionsSchema.optional()
 });
@@ -40274,7 +40274,7 @@ var AnthropicConfigSchema = external_exports.object({
   provider: external_exports.literal("anthropic"),
   apiKey: external_exports.string(),
   temperature: external_exports.number().min(0).max(1).optional(),
-  maxTokens: external_exports.number().min(1e3).optional(),
+  maxOutputTokens: external_exports.number().min(1e3).optional(),
   // Anthropic requires max_tokens
   topP: external_exports.number().min(0).max(1).optional(),
   modelOptions: AnthropicOptionsSchema.optional()
@@ -43228,6 +43228,17 @@ function buildEdges(resources) {
 }
 
 // ../core/src/platform/registry/resource-registry.ts
+function filterArchived(org) {
+  return {
+    ...org,
+    workflows: org.workflows?.filter((w) => !w.config.archived),
+    agents: org.agents?.filter((a) => !a.config.archived),
+    triggers: org.triggers?.filter((t) => !t.archived),
+    integrations: org.integrations?.filter((i) => !i.archived),
+    externalResources: org.externalResources?.filter((e) => !e.archived),
+    humanCheckpoints: org.humanCheckpoints?.filter((h) => !h.archived)
+  };
+}
 var ResourceRegistry = class {
   constructor(registry2) {
     this.registry = registry2;
@@ -43359,6 +43370,7 @@ var ResourceRegistry = class {
    * @throws Error if incoming deployment contains duplicate resourceIds
    */
   registerOrganization(orgName, org, remote) {
+    org = filterArchived(org);
     const incomingWorkflowIds = (org.workflows ?? []).map((w) => w.config.resourceId);
     const incomingAgentIds = (org.agents ?? []).map((a) => a.config.resourceId);
     const incomingIds = [...incomingWorkflowIds, ...incomingAgentIds];
@@ -43420,6 +43432,7 @@ var ResourceRegistry = class {
    * @param org - Resource definitions with real handlers (not stubs)
    */
   registerStaticResources(orgName, org) {
+    org = filterArchived(org);
     const incomingWorkflowIds = (org.workflows ?? []).map((w) => w.config.resourceId);
     const incomingAgentIds = (org.agents ?? []).map((a) => a.config.resourceId);
     const incomingIds = [...incomingWorkflowIds, ...incomingAgentIds];
@@ -43851,7 +43864,7 @@ async function apiDelete(endpoint, apiUrl = resolveApiUrl()) {
 // package.json
 var package_default = {
   name: "@elevasis/sdk",
-  version: "0.5.16",
+  version: "0.5.17",
   description: "SDK for building Elevasis organization resources",
   type: "module",
   bin: {
@@ -44422,6 +44435,8 @@ function registerDeployCommand(program3) {
       }
       const validateSpinner = ora("Validating...").start();
       let org;
+      let activeWorkflows = [];
+      let activeAgents = [];
       try {
         const jiti = (0, import_jiti.createJiti)(import_meta.url);
         const entryModule = await jiti.import((0, import_path.resolve)(entryPath));
@@ -44432,20 +44447,24 @@ function registerDeployCommand(program3) {
           throw new Error("Invalid entry: no default export found");
         }
         new ResourceRegistry({ [orgName]: org });
-        const workflowCount = org.workflows?.length ?? 0;
-        const agentCount = org.agents?.length ?? 0;
-        const totalCount = workflowCount + agentCount;
+        activeWorkflows = (org.workflows ?? []).filter((w) => !w.config.archived);
+        activeAgents = (org.agents ?? []).filter((a) => !a.config.archived);
+        const archivedCount = (org.workflows?.length ?? 0) + (org.agents?.length ?? 0) - activeWorkflows.length - activeAgents.length;
+        const totalCount = activeWorkflows.length + activeAgents.length;
         validateSpinner.succeed(
           source_default.green("Validating...") + source_default.white("              done") + source_default.gray(` (${totalCount} resource${totalCount !== 1 ? "s" : ""}, 0 errors)`)
         );
+        if (archivedCount > 0) {
+          console.log(source_default.gray(`  Skipping ${archivedCount} archived resource${archivedCount !== 1 ? "s" : ""}`));
+        }
         console.log("");
         console.log(source_default.gray(`  Org:    ${orgName}`));
         console.log(source_default.gray(`  Target: ${apiUrl} (${env2})`));
         console.log("");
-        for (const w of org.workflows ?? []) {
+        for (const w of activeWorkflows) {
           console.log(source_default.gray(`  workflow  ${source_default.white(w.config.resourceId)}  v${w.config.version}`));
         }
-        for (const a of org.agents ?? []) {
+        for (const a of activeAgents) {
           console.log(source_default.gray(`  agent     ${source_default.white(a.config.resourceId)}  v${a.config.version}`));
         }
         console.log("");
@@ -44489,7 +44508,7 @@ function registerDeployCommand(program3) {
         );
       }
       const schemaWarnings = [];
-      const workflows = (org.workflows ?? []).map((w) => {
+      const workflows = activeWorkflows.map((w) => {
         const meta = {
           resourceId: w.config.resourceId,
           name: w.config.name,
@@ -44514,7 +44533,7 @@ function registerDeployCommand(program3) {
         }
         return meta;
       });
-      const agents = (org.agents ?? []).map((a) => {
+      const agents = activeAgents.map((a) => {
         const meta = {
           resourceId: a.config.resourceId,
           name: a.config.name,
@@ -44615,7 +44634,7 @@ startWorker(org)
         }
         const result = await response.json();
         uploadSpinner.succeed(source_default.green("Uploading...") + source_default.white("               done"));
-        const totalResources = (org.workflows?.length ?? 0) + (org.agents?.length ?? 0);
+        const totalResources = activeWorkflows.length + activeAgents.length;
         const elapsed = ((Date.now() - startTime) / 1e3).toFixed(1);
         if (result.status === "active") {
           console.log("");

package/dist/index.d.ts

@@ -488,6 +488,8 @@ interface SerializedAgentDefinition {
         version: string;
         type: 'agent';
         status: 'dev' | 'prod';
+        /** Whether this resource is archived and should be excluded from registration and deployment */
+        archived?: boolean;
         systemPrompt: string;
         constraints?: {
             maxIterations?: number;
@@ -503,7 +505,7 @@ interface SerializedAgentDefinition {
         model: string;
         apiKey: string;
         temperature: number;
-        maxTokens: number;
+        maxOutputTokens: number;
         topP?: number;
         modelOptions?: Record<string, unknown>;
     };
@@ -541,6 +543,8 @@ interface SerializedWorkflowDefinition {
         version: string;
         type: 'workflow';
         status: 'dev' | 'prod';
+        /** Whether this resource is archived and should be excluded from registration and deployment */
+        archived?: boolean;
     };
     entryPoint: string;
     steps: Array<{
@@ -653,7 +657,8 @@ interface ModelConfig {
     provider: 'openai' | 'anthropic' | 'openrouter' | 'google' | 'mock';
     apiKey: string;
     temperature?: number;
-    maxTokens?: number;
+    /** Maximum output tokens per LLM call. NOT the model's context window — see ModelInfo.maxTokens for that. */
+    maxOutputTokens?: number;
     topP?: number;
     /**
      * Model-specific options (flat structure)
@@ -783,7 +788,8 @@ interface LLMMessage {
 interface LLMGenerateRequest {
     messages: LLMMessage[];
     responseSchema: unknown;
-    maxTokens?: number;
+    /** Maximum output tokens per LLM call. NOT the model's context window — see ModelInfo.maxTokens for that. */
+    maxOutputTokens?: number;
     temperature?: number;
     topP?: number;
     signal?: AbortSignal;
@@ -5650,12 +5656,27 @@ interface ExecutionMetadata {
  * Unified message event type - covers all message types in sessions
  * Replaces separate SessionTurnMessages and AgentActivityEvent mechanisms
  */
+/**
+ * Structured action metadata attached to assistant messages.
+ * Frontend reads this instead of parsing text prefixes.
+ */
+type AssistantAction = {
+    kind: 'navigate';
+    path: string;
+    reason: string;
+} | {
+    kind: 'update_filters';
+    timeRange: string | null;
+    statusFilter: string | null;
+    searchQuery: string | null;
+};
 type MessageEvent = {
     type: 'user_message';
     text: string;
 } | {
     type: 'assistant_message';
     text: string;
+    _action?: AssistantAction;
 } | {
     type: 'agent:started';
 } | {
@@ -5791,6 +5812,8 @@ interface ResourceDefinition {
     sessionCapable?: boolean;
     /** Whether the resource is local (monorepo) or remote (externally deployed) */
     origin?: 'local' | 'remote';
+    /** Whether this resource is archived and should be excluded from registration and deployment */
+    archived?: boolean;
 }
 /**
  * Domain definition for Command View filtering

package/dist/index.js

@@ -197,7 +197,7 @@ var GPT5ConfigSchema = z.object({
   apiKey: z.string(),
   temperature: z.literal(1),
   // Required to be exactly 1
-  maxTokens: z.number().min(4e3).optional(),
+  maxOutputTokens: z.number().min(4e3).optional(),
   topP: z.number().min(0).max(1).optional(),
   modelOptions: GPT5OptionsSchema.optional()
 });
@@ -206,7 +206,7 @@ var MockConfigSchema = z.object({
   provider: z.enum(["mock"]),
   apiKey: z.string(),
   temperature: z.number().min(0).max(2).optional(),
-  maxTokens: z.number().min(500).optional(),
+  maxOutputTokens: z.number().min(500).optional(),
   topP: z.number().min(0).max(1).optional(),
   modelOptions: z.object({}).strict().optional()
   // No options supported
@@ -226,7 +226,7 @@ var OpenRouterConfigSchema = z.object({
   provider: z.literal("openrouter"),
   apiKey: z.string(),
   temperature: z.number().min(0).max(2).optional(),
-  maxTokens: z.number().min(500).optional(),
+  maxOutputTokens: z.number().min(500).optional(),
   topP: z.number().min(0).max(1).optional(),
   modelOptions: OpenRouterOptionsSchema.optional()
 });
@@ -240,7 +240,7 @@ var GoogleConfigSchema = z.object({
   apiKey: z.string(),
   temperature: z.literal(1).optional(),
   // Must be 1 for Gemini 3 (changing degrades performance)
-  maxTokens: z.number().min(500).optional(),
+  maxOutputTokens: z.number().min(500).optional(),
   topP: z.number().min(0).max(1).optional(),
   modelOptions: GoogleOptionsSchema.optional()
 });
@@ -250,7 +250,7 @@ var AnthropicConfigSchema = z.object({
   provider: z.literal("anthropic"),
   apiKey: z.string(),
   temperature: z.number().min(0).max(1).optional(),
-  maxTokens: z.number().min(1e3).optional(),
+  maxOutputTokens: z.number().min(1e3).optional(),
   // Anthropic requires max_tokens
   topP: z.number().min(0).max(1).optional(),
   modelOptions: AnthropicOptionsSchema.optional()
@@ -3160,6 +3160,17 @@ function buildEdges(resources) {
 }
 
 // ../core/src/platform/registry/resource-registry.ts
+function filterArchived(org) {
+  return {
+    ...org,
+    workflows: org.workflows?.filter((w) => !w.config.archived),
+    agents: org.agents?.filter((a) => !a.config.archived),
+    triggers: org.triggers?.filter((t) => !t.archived),
+    integrations: org.integrations?.filter((i) => !i.archived),
+    externalResources: org.externalResources?.filter((e) => !e.archived),
+    humanCheckpoints: org.humanCheckpoints?.filter((h) => !h.archived)
+  };
+}
 var ResourceRegistry = class {
   constructor(registry) {
     this.registry = registry;
@@ -3291,6 +3302,7 @@ var ResourceRegistry = class {
    * @throws Error if incoming deployment contains duplicate resourceIds
    */
   registerOrganization(orgName, org, remote) {
+    org = filterArchived(org);
     const incomingWorkflowIds = (org.workflows ?? []).map((w) => w.config.resourceId);
     const incomingAgentIds = (org.agents ?? []).map((a) => a.config.resourceId);
     const incomingIds = [...incomingWorkflowIds, ...incomingAgentIds];
@@ -3352,6 +3364,7 @@ var ResourceRegistry = class {
    * @param org - Resource definitions with real handlers (not stubs)
    */
   registerStaticResources(orgName, org) {
+    org = filterArchived(org);
     const incomingWorkflowIds = (org.workflows ?? []).map((w) => w.config.resourceId);
     const incomingAgentIds = (org.agents ?? []).map((a) => a.config.resourceId);
     const incomingIds = [...incomingWorkflowIds, ...incomingAgentIds];

package/dist/worker/index.js

@@ -2551,10 +2551,13 @@ function buildReasoningRequest(iterationContext) {
     systemPrompt,
     tools: toolDefinitions,
     constraints: {
-      maxTokens: iterationContext.modelConfig.maxTokens,
+      maxOutputTokens: iterationContext.modelConfig.maxOutputTokens,
       temperature: 1
     },
-    memoryContext: iterationContext.memoryManager.toContext(iterationContext.iteration, iterationContext.executionContext.sessionTurnNumber),
+    memoryContext: iterationContext.memoryManager.toContext(
+      iterationContext.iteration,
+      iterationContext.executionContext.sessionTurnNumber
+    ),
     includeMessageAction: isSessionCapable,
     includeNavigateKnowledge: hasKnowledgeMap,
     includeMemoryOps
@@ -2595,7 +2598,7 @@ var GPT5ConfigSchema = z.object({
   apiKey: z.string(),
   temperature: z.literal(1),
   // Required to be exactly 1
-  maxTokens: z.number().min(4e3).optional(),
+  maxOutputTokens: z.number().min(4e3).optional(),
   topP: z.number().min(0).max(1).optional(),
   modelOptions: GPT5OptionsSchema.optional()
 });
@@ -2604,7 +2607,7 @@ var MockConfigSchema = z.object({
   provider: z.enum(["mock"]),
   apiKey: z.string(),
   temperature: z.number().min(0).max(2).optional(),
-  maxTokens: z.number().min(500).optional(),
+  maxOutputTokens: z.number().min(500).optional(),
   topP: z.number().min(0).max(1).optional(),
   modelOptions: z.object({}).strict().optional()
   // No options supported
@@ -2624,7 +2627,7 @@ var OpenRouterConfigSchema = z.object({
   provider: z.literal("openrouter"),
   apiKey: z.string(),
   temperature: z.number().min(0).max(2).optional(),
-  maxTokens: z.number().min(500).optional(),
+  maxOutputTokens: z.number().min(500).optional(),
   topP: z.number().min(0).max(1).optional(),
   modelOptions: OpenRouterOptionsSchema.optional()
 });
@@ -2638,7 +2641,7 @@ var GoogleConfigSchema = z.object({
   apiKey: z.string(),
   temperature: z.literal(1).optional(),
   // Must be 1 for Gemini 3 (changing degrades performance)
-  maxTokens: z.number().min(500).optional(),
+  maxOutputTokens: z.number().min(500).optional(),
   topP: z.number().min(0).max(1).optional(),
   modelOptions: GoogleOptionsSchema.optional()
 });
@@ -2648,7 +2651,7 @@ var AnthropicConfigSchema = z.object({
   provider: z.literal("anthropic"),
   apiKey: z.string(),
   temperature: z.number().min(0).max(1).optional(),
-  maxTokens: z.number().min(1e3).optional(),
+  maxOutputTokens: z.number().min(1e3).optional(),
   // Anthropic requires max_tokens
   topP: z.number().min(0).max(1).optional(),
   modelOptions: AnthropicOptionsSchema.optional()
@@ -2904,9 +2907,9 @@ var AgentIterationOutputSchema = z.object({
   memoryOps: MemoryOperationsSchema.optional(),
   nextActions: z.array(AgentActionSchema)
 });
-function validateTokenConfiguration(model, maxTokens) {
+function validateTokenConfiguration(model, maxOutputTokens) {
   const modelInfo = getModelInfo(model);
-  const configured = maxTokens || 1e3;
+  const configured = maxOutputTokens || 1e3;
   if (!modelInfo) {
     if (configured < 2e3) {
       throw new InsufficientTokensError(
@@ -2929,15 +2932,20 @@ function validateTokenConfiguration(model, maxTokens) {
   }
 }
 async function callLLMForAgentIteration(adapter, request) {
-  validateTokenConfiguration(request.model, request.constraints.maxTokens);
+  validateTokenConfiguration(request.model, request.constraints.maxOutputTokens);
   const messages = [
     { role: "system", content: request.systemPrompt },
     { role: "user", content: request.memoryContext }
   ];
   const response = await adapter.generate({
     messages,
-    responseSchema: buildIterationResponseSchema(request.tools, request.includeMessageAction, request.includeNavigateKnowledge, request.includeMemoryOps),
-    maxTokens: request.constraints.maxTokens,
+    responseSchema: buildIterationResponseSchema(
+      request.tools,
+      request.includeMessageAction,
+      request.includeNavigateKnowledge,
+      request.includeMemoryOps
+    ),
+    maxOutputTokens: request.constraints.maxOutputTokens,
     temperature: request.constraints.temperature,
     signal: request.signal
   });
@@ -2949,16 +2957,13 @@ async function callLLMForAgentIteration(adapter, request) {
       nextActions: validated.nextActions
     };
   } catch (error) {
-    throw new AgentOutputValidationError(
-      "Agent iteration output validation failed",
-      {
-        zodError: error instanceof ZodError ? error.format() : error
-      }
-    );
+    throw new AgentOutputValidationError("Agent iteration output validation failed", {
+      zodError: error instanceof ZodError ? error.format() : error
+    });
   }
 }
 async function callLLMForAgentCompletion(adapter, request) {
-  validateTokenConfiguration(request.model, request.constraints.maxTokens);
+  validateTokenConfiguration(request.model, request.constraints.maxOutputTokens);
   const response = await adapter.generate({
     messages: [
       { role: "system", content: request.systemPrompt },
@@ -2967,7 +2972,7 @@ async function callLLMForAgentCompletion(adapter, request) {
     responseSchema: request.outputSchema,
     // Use output schema directly
     temperature: request.constraints.temperature || 0.3,
-    maxTokens: request.constraints.maxTokens,
+    maxOutputTokens: request.constraints.maxOutputTokens,
     signal: request.signal
   });
   return response.output;
@@ -3681,8 +3686,19 @@ function estimateTokens(text) {
 var MAX_SESSION_MEMORY_KEYS = 25;
 var MAX_MEMORY_TOKENS = 32e3;
 var MAX_SINGLE_ENTRY_TOKENS = 2e3;
+var MAX_TOOL_RESULT_TOKENS = 4e3;
 
 // ../core/src/execution/engine/agent/memory/manager.ts
+function truncateToolResult(content, maxTokens) {
+  const estimated = estimateTokens(content);
+  if (estimated <= maxTokens) return content;
+  const maxChars = Math.floor(maxTokens * 3.5);
+  const truncated = content.slice(0, maxChars);
+  const omitted = estimated - maxTokens;
+  return truncated + `
+
+[Response truncated \u2014 estimated ${omitted} tokens omitted. Use more specific filters to get smaller results.]`;
+}
 var MemoryManager = class {
   constructor(memory, constraints = {}, logger) {
     this.memory = memory;
@@ -3750,17 +3766,31 @@ var MemoryManager = class {
    */
   addToHistory(entry) {
     if (entry.turnNumber === void 0 && entry.type !== "context") {
-      throw new AgentMemoryValidationError(
-        "turnNumber required for history entries (use null for session memory)",
-        {
-          entryType: entry.type,
-          missingField: "turnNumber",
-          iterationNumber: entry.iterationNumber
-        }
-      );
+      throw new AgentMemoryValidationError("turnNumber required for history entries (use null for session memory)", {
+        entryType: entry.type,
+        missingField: "turnNumber",
+        iterationNumber: entry.iterationNumber
+      });
+    }
+    let content = entry.content;
+    if (entry.type === "tool-result") {
+      const before = content;
+      content = truncateToolResult(content, MAX_TOOL_RESULT_TOKENS);
+      if (content !== before) {
+        const truncateTime = Date.now();
+        this.logger?.action(
+          "memory-tool-result-truncate",
+          `Tool result truncated (${estimateTokens(before)} -> ${MAX_TOOL_RESULT_TOKENS} tokens)`,
+          entry.iterationNumber ?? 0,
+          truncateTime,
+          truncateTime,
+          0
+        );
+      }
     }
     this.memory.history.push({
       ...entry,
+      content,
       timestamp: Date.now()
     });
     this.autoCompact();
@@ -4416,7 +4446,7 @@ var Agent = class {
         memoryContext: this.memoryManager.toContext(this.iterationNumber, this.executionContext?.sessionTurnNumber),
         outputSchema,
         constraints: {
-          maxTokens: this.modelConfig.maxTokens,
+          maxOutputTokens: this.modelConfig.maxOutputTokens,
           temperature
         },
         model: this.modelConfig.model,
@@ -4742,7 +4772,7 @@ var PostMessageLLMAdapter = class {
         messages: request.messages,
         responseSchema: request.responseSchema,
         temperature: request.temperature,
-        maxTokens: request.maxTokens
+        maxOutputTokens: request.maxOutputTokens
       }
     });
     return { output: result };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@elevasis/sdk",
-  "version": "0.5.16",
+  "version": "0.5.17",
   "description": "SDK for building Elevasis organization resources",
   "type": "module",
   "bin": {

package/reference/framework/agent.mdx

@@ -81,7 +81,7 @@ The `/work` command manages in-progress task tracking across sessions. It uses `
 - **Auto-create** -- When the user describes work that doesn't match an existing task, the agent creates a task doc automatically. If intent is clear, it skips the interview; if ambiguous, it asks 1-2 focused questions.
 - **Auto-save** -- The agent saves progress silently when the conversation context is getting heavy, the user is wrapping up, or 2+ steps have been completed without saving. Updates progress markers, files modified, and resume context.
 - **Auto-resume** -- When the user picks an existing task (by number, name, or keyword), the agent loads context and resumes automatically.
-- **Suggest complete** -- When all plan steps are marked COMPLETE, the agent suggests finalizing: "All steps for '{task}' look done. Want me to finalize it?" It never auto-invokes completion. The complete flow validates readiness, cleans the doc (strips resume context, removes progress markers, targets 200-400 lines), determines destination in `docs/`, confirms with user, moves, and verifies.
+- **Suggest complete** -- When all plan steps are marked COMPLETE, the agent suggests finalizing: "All steps for '`{task}`' look done. Want me to finalize it?" It never auto-invokes completion. The complete flow validates readiness, cleans the doc (strips resume context, removes progress markers, targets 200-400 lines), determines destination in `docs/`, confirms with user, moves, and verifies.
 
 **Directory conventions for `docs/in-progress/`:**