@lota-sdk/core 0.3.2 → 0.4.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lota-sdk/core",
-  "version": "0.3.2",
+  "version": "0.4.0",
   "type": "module",
   "main": "./src/index.ts",
   "types": "./src/index.ts",
@@ -32,7 +32,7 @@
     "@chat-adapter/slack": "^4.23.0",
     "@chat-adapter/state-ioredis": "^4.23.0",
     "@logtape/logtape": "^2.0.5",
-    "@lota-sdk/shared": "0.3.2",
+    "@lota-sdk/shared": "0.4.0",
     "@mendable/firecrawl-js": "^4.18.1",
     "@surrealdb/node": "^3.0.3",
     "ai": "^6.0.145",

package/src/ai-gateway/ai-gateway.ts

@@ -5,7 +5,7 @@ import type { LanguageModelMiddleware } from 'ai'
 
 import { getRuntimeConfig } from '../runtime/runtime-config'
 import { isRecord, readString } from '../utils/string'
-import { buildAiGatewayCacheHeaders, toAiGatewayCacheKeyPart } from './cache-headers'
+import { buildAiGatewayCacheHeaders } from './cache-headers'
 
 type AiGatewayLanguageModel = Parameters<typeof wrapLanguageModel>[0]['model']
 type AiGatewayExtraParams = Record<string, unknown>
@@ -55,14 +55,8 @@ function parseAiGatewayJsonRequestBody(body: BodyInit | null | undefined): Recor
   return isRecord(parsed) ? parsed : null
 }
 
-function withDefaultAiGatewayCacheHeaders(params: AiGatewayCallOptions, modelId: string): AiGatewayCallOptions {
-  return {
-    ...params,
-    headers: mergeAiGatewayHeaders(
-      params.headers,
-      buildAiGatewayCacheHeaders(`model:${toAiGatewayCacheKeyPart(modelId)}`),
-    ),
-  }
+function withDefaultAiGatewayCacheHeaders(params: AiGatewayCallOptions): AiGatewayCallOptions {
+  return { ...params, headers: mergeAiGatewayHeaders(params.headers, buildAiGatewayCacheHeaders('lota-sdk')) }
 }
 
 function normalizeAiGatewayUrl(value: string): string {
@@ -417,7 +411,7 @@ export function aiGatewayModel(modelId: string) {
       middleware: {
         specificationVersion: 'v3',
         transformParams: async ({ params, type }) =>
-          withDefaultAiGatewayCacheHeaders(addAiGatewayReasoningRawChunks(params, type), modelId),
+          withDefaultAiGatewayCacheHeaders(addAiGatewayReasoningRawChunks(params, type)),
         wrapStream: async ({ doStream, params }) => {
           const result = await doStream()
           if (!isReasoningEnabled(params)) return result
@@ -435,7 +429,7 @@ export function aiGatewayOpenRouterResponseHealingModel(modelId: string) {
       model: getAiGatewayOpenRouterResponseHealingProvider()(modelId),
       middleware: {
         specificationVersion: 'v3',
-        transformParams: async ({ params }) => withDefaultAiGatewayCacheHeaders(params, modelId),
+        transformParams: async ({ params }) => withDefaultAiGatewayCacheHeaders(params),
       },
     }),
   )
@@ -449,7 +443,7 @@ export function aiGatewayChatModel(modelId: string) {
         specificationVersion: 'v3',
         transformParams: async ({ params, type }) =>
           normalizeAiGatewayChatProviderOptions(
-            withDefaultAiGatewayCacheHeaders(addAiGatewayReasoningRawChunks(params, type), modelId),
+            withDefaultAiGatewayCacheHeaders(addAiGatewayReasoningRawChunks(params, type)),
           ),
         wrapGenerate: async ({ doGenerate }) => {
           const result = await doGenerate()

package/src/db/sdk-database.ts

@@ -1 +1 @@
-export const LOTA_SDK_DATABASE_NAME = 'lotasdk'
+export const LOTA_SDK_DATABASE_NAME = 'lotasdk_thread'

package/src/redis/stream-context.ts

@@ -38,7 +38,10 @@ let sharedSubscriber: { client: Redis; subscriber: Subscriber } | undefined
 
 function getSharedSubscriber(): Subscriber {
   if (!sharedSubscriber) {
-    const client = getRedisConnection().duplicate()
+    // Disable enableReadyCheck — the ready check sends INFO which is rejected
+    // on connections in subscribe mode, causing unhandled ioredis error events.
+    const client = getRedisConnection().duplicate({ enableReadyCheck: false })
+    client.on('error', () => {}) // prevent [ioredis] Unhandled error event logs
     sharedSubscriber = { client, subscriber: toSubscriber(client) }
   }
   return sharedSubscriber.subscriber

package/src/services/plan-run.service.ts

@@ -368,7 +368,7 @@ class PlanRunService {
 
       // Slim mode: non-active/ready nodes get summary only (used for prompt injection via JSON.stringify).
       // The cast is safe — this data is only consumed by formatExecutionPlansForPrompt, not by Zod validation.
-      // Plan introspection tools (getExecutionPlanDetails) call toSerializablePlan without slim=true.
+      // executionPlanQuery calls toSerializablePlan without slim=true for full inspection payloads.
       if (slim && !isActiveOrReady) {
         return {
           id: nodeSpec.nodeId,

package/src/services/thread.service.ts

@@ -712,6 +712,19 @@ class ThreadService extends BaseService<typeof ThreadSchema> {
     return true
   }
 
+  async clearThread(threadId: RecordIdRef): Promise<void> {
+    const threadRef = ensureRecordId(threadId, TABLES.THREAD)
+    await databaseService.deleteWhere(TABLES.THREAD_MESSAGE, { threadId: threadRef })
+    await databaseService.query<unknown>(surql`
+      UPDATE ONLY ${threadRef}
+      SET turnCount = 0,
+          compactionSummary = NONE,
+          lastCompactedMessageId = NONE,
+          activeRunId = NONE,
+          activeStreamId = NONE
+    `)
+  }
+
   async deleteThread(threadId: RecordIdRef): Promise<void> {
     const existing = await this.getById(threadId)
     assertMutableThread(existing)

package/src/system-agents/context-compaction.agent.ts

@@ -35,7 +35,7 @@ export function createContextCompactionAgent(options: CreateHelperToolLoopAgentO
   return new ToolLoopAgent({
     id: 'context-compaction',
     model: aiGatewayOpenRouterResponseHealingModel(OPENROUTER_STRUCTURED_HELPER_MODEL_ID),
-    headers: buildAiGatewayDirectCacheHeaders('context-compaction'),
+    headers: buildAiGatewayDirectCacheHeaders('lota-sdk'),
     providerOptions: OPENROUTER_HIGH_REASONING_PROVIDER_OPTIONS,
     ...resolveHelperAgentOptions(options, { instructions: CONTEXT_COMPACTION_PROMPT }),
   })

package/src/system-agents/memory-reranker.agent.ts

@@ -33,7 +33,7 @@ export function createMemoryRerankerAgent(options: CreateHelperToolLoopAgentOpti
   return new ToolLoopAgent({
     id: 'memory-reranker',
     model: aiGatewayOpenRouterResponseHealingModel(OPENROUTER_STRUCTURED_HELPER_MODEL_ID),
-    headers: buildAiGatewayDirectCacheHeaders('memory-reranker'),
+    headers: buildAiGatewayDirectCacheHeaders('lota-sdk'),
     providerOptions: OPENROUTER_LOW_REASONING_PROVIDER_OPTIONS,
     ...resolveHelperAgentOptions(options),
   })

package/src/system-agents/memory.agent.ts

@@ -53,7 +53,7 @@ export function createOrgMemoryAgent(options: CreateHelperToolLoopAgentOptions)
   return new ToolLoopAgent({
     id: 'org-memory',
     model: aiGatewayOpenRouterResponseHealingModel(OPENROUTER_STRUCTURED_HELPER_MODEL_ID),
-    headers: buildAiGatewayDirectCacheHeaders('org-memory'),
+    headers: buildAiGatewayDirectCacheHeaders('lota-sdk'),
     providerOptions: OPENROUTER_HIGH_REASONING_PROVIDER_OPTIONS,
     ...resolveHelperAgentOptions(options),
   })

package/src/system-agents/recent-activity-title-refiner.agent.ts

@@ -80,7 +80,7 @@ export function createRecentActivityTitleRefinerAgent(options: CreateHelperToolL
   return new ToolLoopAgent({
     id: 'recent-activity-title-refiner',
     model: aiGatewayModel(OPENROUTER_STRUCTURED_HELPER_MODEL_ID),
-    headers: buildAiGatewayDirectCacheHeaders('recent-activity-title-refiner'),
+    headers: buildAiGatewayDirectCacheHeaders('lota-sdk'),
     providerOptions: OPENROUTER_HIGH_REASONING_PROVIDER_OPTIONS,
     ...resolveHelperAgentOptions(options, {
       instructions: buildRecentActivityTitleRefinerPrompt(),

package/src/system-agents/regular-chat-memory-digest.agent.ts

@@ -28,7 +28,7 @@ export function createRegularChatMemoryDigestAgent(options: CreateHelperToolLoop
   return new ToolLoopAgent({
     id: 'regular-chat-memory-digest',
     model: aiGatewayOpenRouterResponseHealingModel(OPENROUTER_STRUCTURED_HELPER_MODEL_ID),
-    headers: buildAiGatewayDirectCacheHeaders('regular-chat-memory-digest'),
+    headers: buildAiGatewayDirectCacheHeaders('lota-sdk'),
     providerOptions: OPENROUTER_HIGH_REASONING_PROVIDER_OPTIONS,
     ...resolveHelperAgentOptions(options, {
       instructions: regularChatMemoryDigestPrompt,

package/src/system-agents/skill-extractor.agent.ts

@@ -46,7 +46,7 @@ export function createSkillExtractorAgent(options: CreateHelperToolLoopAgentOpti
   return new ToolLoopAgent({
     id: 'skill-extractor',
     model: aiGatewayOpenRouterResponseHealingModel(OPENROUTER_STRUCTURED_HELPER_MODEL_ID),
-    headers: buildAiGatewayDirectCacheHeaders('skill-extractor'),
+    headers: buildAiGatewayDirectCacheHeaders('lota-sdk'),
     providerOptions: OPENROUTER_HIGH_REASONING_PROVIDER_OPTIONS,
     ...resolveHelperAgentOptions(options, {
       instructions: skillExtractorPrompt,

package/src/system-agents/skill-manager.agent.ts

@@ -70,7 +70,7 @@ export function createSkillManagerAgent(options: CreateHelperToolLoopAgentOption
   return new ToolLoopAgent({
     id: 'skill-manager',
     model: aiGatewayOpenRouterResponseHealingModel(OPENROUTER_STRUCTURED_HELPER_MODEL_ID),
-    headers: buildAiGatewayDirectCacheHeaders('skill-manager'),
+    headers: buildAiGatewayDirectCacheHeaders('lota-sdk'),
     providerOptions: OPENROUTER_HIGH_REASONING_PROVIDER_OPTIONS,
     ...resolveHelperAgentOptions(options, {
       instructions: skillManagerPrompt,

package/src/system-agents/thread-router.agent.ts

@@ -165,7 +165,7 @@ async function generateRouterObject<TSchema extends z.ZodTypeAny>(params: {
     const { object } = await withTimeout(
       generateObject({
         model: aiGatewayChatModel(modelId),
-        headers: buildAiGatewayDirectCacheHeaders('thread-router'),
+        headers: buildAiGatewayDirectCacheHeaders('lota-sdk'),
         providerOptions: { openai: { reasoningEffort: 'low' } },
         schema: params.schema,
         system: params.system,

package/src/system-agents/title-generator.agent.ts

@@ -34,7 +34,7 @@ export function createThreadTitleGeneratorAgent(options: CreateHelperToolLoopAge
   return new ToolLoopAgent({
     id: 'thread-title-generator',
     model: aiGatewayModel(OPENROUTER_FAST_REASONING_MODEL_ID),
-    headers: buildAiGatewayDirectCacheHeaders('thread-title-generator'),
+    headers: buildAiGatewayDirectCacheHeaders('lota-sdk'),
     providerOptions: OPENROUTER_MINIMAL_REASONING_PROVIDER_OPTIONS,
     ...resolveHelperAgentOptions(options, {
       instructions: THREAD_TITLE_GENERATOR_PROMPT,

package/src/tools/execution-plan.tool.ts

@@ -5,7 +5,7 @@ import {
   expandAgentPlanDraft,
   getLatestExecutionPlanResult,
 } from '@lota-sdk/shared'
-import type { CreateProjectWithPlanResultData, ExecutionPlanAction, ExecutionPlanArgs } from '@lota-sdk/shared'
+import type { CreateProjectWithPlanResultData } from '@lota-sdk/shared'
 import { tool } from 'ai'
 
 import type { RecordIdRef } from '../db/record-id'
@@ -26,11 +26,6 @@ type ExecutionPlanExecutionPlanService = Pick<
   | 'getActivePlansForThread'
 >
 
-function extractDraft(input: ExecutionPlanArgs) {
-  const { action, projectTitle, targetThreadId, runId, reason, ...draftInput } = input
-  return { action, projectTitle, targetThreadId, runId, reason, draft: expandAgentPlanDraft(draftInput) }
-}
-
 export function createExecutionPlanTool(params: {
   orgId: RecordIdRef
   userId: RecordIdRef
@@ -49,20 +44,26 @@ export function createExecutionPlanTool(params: {
       'Manage execution plans. Actions: create (inline, 1-2 nodes), create-project (dedicated project thread, 3+ nodes), replace (swap active plan), resume (resume interrupted plan).',
     inputSchema: ExecutionPlanArgsSchema,
     execute: async (input) => {
-      const { action, projectTitle, targetThreadId, runId, reason, draft } = extractDraft(input)
+      const parsed = ExecutionPlanArgsSchema.parse(input)
+      let result: unknown
 
-      const handler: Record<ExecutionPlanAction, () => Promise<unknown>> = {
-        create: async () => {
+      switch (parsed.action) {
+        case 'create': {
+          const { action: _action, targetThreadId, ...draftInput } = parsed
+          const draft = expandAgentPlanDraft(draftInput)
           params.validateInlinePlan?.(draft)
-          return await resolvedEpService.createPlan({
+          result = await resolvedEpService.createPlan({
             organizationId: params.orgId,
             threadId: targetThreadId ?? params.threadId,
             leadAgentId: params.agentId,
             input: draft,
           })
-        },
+          break
+        }
 
-        'create-project': async () => {
+        case 'create-project': {
+          const { action: _action, projectTitle, targetThreadId, ...draftInput } = parsed
+          const draft = expandAgentPlanDraft(draftInput)
           const targetThread = targetThreadId
             ? await resolvedWsService.getThread(targetThreadId)
             : await (() => {
@@ -94,14 +95,14 @@ export function createExecutionPlanTool(params: {
 
           const createdThread = !targetThreadId
           try {
-            const result = await resolvedEpService.createPlan({
+            const created = await resolvedEpService.createPlan({
               organizationId: params.orgId,
               threadId: targetThread.id,
               leadAgentId: params.agentId,
               input: draft,
             })
-            return {
-              ...result,
+            result = {
+              ...created,
               threadId: targetThread.id,
               threadTitle: targetThread.title,
               createdThread,
@@ -112,35 +113,30 @@ export function createExecutionPlanTool(params: {
             }
             throw error
           }
-        },
-
-        replace: async () => {
-          if (!runId || !reason) {
-            throw new Error('runId and reason are required when action is "replace".')
-          }
+          break
+        }
 
-          return await resolvedEpService.replacePlan({
+        case 'replace': {
+          const { action: _action, runId, reason, ...draftInput } = parsed
+          const draft = expandAgentPlanDraft(draftInput)
+          result = await resolvedEpService.replacePlan({
             organizationId: params.orgId,
             threadId: params.threadId,
             leadAgentId: params.agentId,
             input: { runId, reason, ...draft },
           })
-        },
-
-        resume: async () => {
-          if (!runId) {
-            throw new Error('runId is required when action is "resume".')
-          }
+          break
+        }
 
-          return await resolvedEpService.resumeRun({
+        case 'resume':
+          result = await resolvedEpService.resumeRun({
             threadId: params.threadId,
             emittedBy: params.agentId,
-            input: { runId },
+            input: { runId: parsed.runId },
           })
-        },
+          break
       }
 
-      const result = await handler[action]()
       params.onPlanChanged?.()
       return result
     },