@lota-sdk/core 0.4.39 → 0.4.41

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lota-sdk/core",
-  "version": "0.4.39",
+  "version": "0.4.41",
   "files": [
     "src",
     "infrastructure/schema"
@@ -32,7 +32,7 @@
     "@ai-sdk/provider": "^3.0.9",
     "@chat-adapter/slack": "^4.26.0",
     "@chat-adapter/state-ioredis": "^4.26.0",
-    "@lota-sdk/shared": "0.4.39",
+    "@lota-sdk/shared": "0.4.41",
     "@mendable/firecrawl-js": "^4.20.0",
     "@surrealdb/node": "^3.0.3",
     "ai": "^6.0.170",

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

@@ -1,7 +1,7 @@
 import { devToolsMiddleware } from '@ai-sdk/devtools'
 import { createOpenAI } from '@ai-sdk/openai'
 import type { JSONSchema7 } from '@ai-sdk/provider'
-import { wrapEmbeddingModel, wrapLanguageModel } from 'ai'
+import { wrapLanguageModel } from 'ai'
 import type { LanguageModelMiddleware } from 'ai'
 import { Cause, Clock, Context, Duration, Effect, Fiber, Layer, Semaphore } from 'effect'
 
@@ -16,7 +16,6 @@ type AiGatewayChatResponse = { body?: unknown }
 type AiGatewayTransformParamsOptions = Parameters<NonNullable<LanguageModelMiddleware['transformParams']>>[0]
 type WrapStreamOptions = Parameters<NonNullable<LanguageModelMiddleware['wrapStream']>>[0]
 type AiGatewayLanguageModel = Parameters<typeof wrapLanguageModel>[0]['model']
-type AiGatewayEmbeddingModel = Parameters<typeof wrapEmbeddingModel>[0]['model']
 type AiGatewayCallOptions = WrapStreamOptions['params']
 type AiGatewayFunctionTool = Extract<NonNullable<AiGatewayCallOptions['tools']>[number], { type: 'function' }>
 type AiGatewayGenerateResult = Awaited<ReturnType<WrapStreamOptions['doGenerate']>>
@@ -45,8 +44,6 @@ const AI_GATEWAY_MAX_RETRIES = 4
 const AI_GATEWAY_MAX_RETRY_DELAY_MS = 15_000
 const OPENAI_RESPONSES_PROVIDER_ID = 'openai.responses'
 const OPENAI_CHAT_PROVIDER_ID = 'openai.chat'
-const OPENAI_EMBEDDING_PROVIDER_ID = 'openai.embedding'
-const OPENAI_EMBEDDING_MAX_PER_CALL = 2_048
 const RETRYABLE_NETWORK_ERROR_CODES = new Set([
   'ECONNABORTED',
   'ECONNREFUSED',
@@ -1244,23 +1241,6 @@ function createAiGatewayLanguageModelPlaceholder(modelId: string, providerId: st
   }
 }
 
-function createAiGatewayEmbeddingModelPlaceholder(modelId: string): AiGatewayEmbeddingModel {
-  return {
-    specificationVersion: 'v3',
-    provider: OPENAI_EMBEDDING_PROVIDER_ID,
-    modelId,
-    maxEmbeddingsPerCall: OPENAI_EMBEDDING_MAX_PER_CALL,
-    supportsParallelCalls: true,
-    doEmbed: () =>
-      Promise.reject(
-        new Error(
-          `[ai-gateway] AiGateway embedding model ${modelId}.doEmbed was invoked without the gateway middleware; ` +
-            'this call path should be fully handled by aiGatewayEmbeddingModel middleware.',
-        ),
-      ),
-  }
-}
-
 export function aiGatewayModel(modelId: string, deps?: AiGatewayDeps) {
   if (isOpenRouterModel(modelId)) {
     return aiGatewayChatModel(modelId, deps)
@@ -1287,28 +1267,8 @@ export function aiGatewayChatModel(modelId: string, deps?: AiGatewayDeps) {
   )
 }
 
-export function aiGatewayEmbeddingModel(modelId: string, deps?: AiGatewayDeps) {
-  return wrapEmbeddingModel({
-    model: createAiGatewayEmbeddingModelPlaceholder(modelId),
-    middleware: {
-      specificationVersion: 'v3',
-      wrapEmbed: ({ params }) => {
-        const resolvedDeps = resolveAiGatewayDeps(deps)
-        const embeddingModel = resolvedDeps.gateway.provider.embeddingModel(modelId)
-        return resolvedDeps.runPromise(
-          withAiGatewayConcurrency(
-            withAiGatewayResilience(
-              'ai-gateway.embed',
-              Effect.tryPromise({
-                try: () => embeddingModel.doEmbed(params),
-                catch: (cause) => classifyAiGatewayError('ai-gateway.embed', cause),
-              }),
-            ).pipe(Effect.withSpan('AiGateway.embed'), Effect.annotateSpans({ modelId })),
-          ).pipe(Effect.provideService(AiGatewayTag, resolvedDeps.gateway)),
-        )
-      },
-    },
-  })
+export function aiGatewayEmbeddingModel(modelId: string, _deps?: AiGatewayDeps) {
+  return openRouterEmbeddingModel(modelId)
 }
 
 /**

package/src/config/model-constants.ts

@@ -1,4 +1,5 @@
 export {
+  AI_GATEWAY_FAST_RERANK_MODEL_ID,
   AI_GATEWAY_REASONING_SUMMARY_LEVEL,
   OPENROUTER_GEMINI_PRO_MODEL_ID,
   OPENAI_HIGH_REASONING_PROVIDER_OPTIONS,

package/src/runtime/runtime-config.ts

@@ -5,7 +5,7 @@ import { z } from 'zod'
 import type { CoreThreadProfile } from '../config/agent-defaults'
 import type { AgentFactory, AgentRuntimeConfigProvider, AgentToolBuilder } from '../config/agent-types'
 import { DEFAULT_AI_GATEWAY_URL } from '../config/constants'
-import { OPENROUTER_FAST_RERANK_MODEL_ID } from '../config/model-constants'
+import { AI_GATEWAY_FAST_RERANK_MODEL_ID } from '../config/model-constants'
 import type { LotaThreadConfig, ThreadBootstrapWelcomeConfig } from '../config/thread-defaults'
 import type { RecordIdRef } from '../db/record-id'
 import type { NotificationService } from '../services/notification.service'
@@ -239,13 +239,13 @@ export const LotaRuntimeConfigSchema = z.object({
       searchK: z.coerce.number().int().positive().default(6),
       embeddingCacheTtlSeconds: z.coerce.number().int().positive().default(7200),
       rerankerStrategy: MemoryRerankerStrategySchema.default('rerank'),
-      rerankerModelId: z.string().trim().min(1).default(OPENROUTER_FAST_RERANK_MODEL_ID),
+      rerankerModelId: z.string().trim().min(1).default(AI_GATEWAY_FAST_RERANK_MODEL_ID),
     })
     .default({
       searchK: 6,
       embeddingCacheTtlSeconds: 7200,
       rerankerStrategy: 'rerank',
-      rerankerModelId: OPENROUTER_FAST_RERANK_MODEL_ID,
+      rerankerModelId: AI_GATEWAY_FAST_RERANK_MODEL_ID,
     }),
   threads: threadConfigSchema.default({}),
   agents: agentsConfigSchema,
@@ -342,7 +342,7 @@ export const lotaRuntimeEnvConfig = Config.all({
   memorySearchK: Config.number('MEMORY_SEARCH_K').pipe(Config.withDefault(6)),
   memoryRerankerStrategy: Config.string('MEMORY_RERANKER_STRATEGY').pipe(Config.withDefault('rerank')),
   memoryRerankerModelId: Config.string('MEMORY_RERANKER_MODEL_ID').pipe(
-    Config.withDefault(OPENROUTER_FAST_RERANK_MODEL_ID),
+    Config.withDefault(AI_GATEWAY_FAST_RERANK_MODEL_ID),
   ),
   otlpBaseUrl: Config.string('OTLP_BASE_URL').pipe(Config.option),
   otlpServiceName: Config.string('OTLP_SERVICE_NAME').pipe(Config.withDefault('lota-sdk')),

package/src/services/memory/rerank.service.ts

@@ -2,7 +2,7 @@ import { Context, Effect, Layer } from 'effect'
 import * as Schema from 'effect/Schema'
 import { z } from 'zod'
 
-import { OPENROUTER_FAST_RERANK_MODEL_ID } from '../../config/model-constants'
+import { AI_GATEWAY_FAST_RERANK_MODEL_ID } from '../../config/model-constants'
 import { ConfigurationError, ERROR_TAGS } from '../../effect/errors'
 import { RuntimeConfigServiceTag } from '../../effect/services'
 import { toValidationError } from '../../effect/zod'
@@ -11,6 +11,7 @@ import type { ResolvedLotaRuntimeConfig } from '../../runtime/runtime-config'
 const AI_GATEWAY_VIRTUAL_KEY_HEADER = 'x-bf-vk' as const
 const EXPECTED_GATEWAY_KEY_PREFIX = 'sk-bf-' as const
 const OPENROUTER_MODEL_PREFIX = 'openrouter/' as const
+const LEGACY_COHERE_RERANK_MODEL_IDS = new Set(['cohere/rerank-4-fast', 'openrouter/cohere/rerank-4-fast'])
 
 const RerankRequestBodySchema = Schema.Struct({
   model: Schema.String,
@@ -58,6 +59,8 @@ function normalizeRerankModelId(modelId: string): string {
     throw new ConfigurationError({ message: 'Rerank model id is required.', key: 'rerankModelId' })
   }
 
+  if (LEGACY_COHERE_RERANK_MODEL_IDS.has(normalized)) return AI_GATEWAY_FAST_RERANK_MODEL_ID
+
   return normalized.startsWith(OPENROUTER_MODEL_PREFIX) ? normalized.slice(OPENROUTER_MODEL_PREFIX.length) : normalized
 }
 
@@ -137,7 +140,7 @@ export function makeRerankService(config: ResolvedLotaRuntimeConfig) {
     const configured = readConfiguredRerankModelId()
     if (configured) return normalizeRerankModelId(configured)
 
-    return OPENROUTER_FAST_RERANK_MODEL_ID
+    return AI_GATEWAY_FAST_RERANK_MODEL_ID
   }
 
   return {