@lota-sdk/core 0.4.40 → 0.4.42

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lota-sdk/core",
-  "version": "0.4.40",
+  "version": "0.4.42",
   "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": "0.4.40",
+    "@lota-sdk/shared": "0.4.42",
     "@mendable/firecrawl-js": "^4.20.0",
     "@surrealdb/node": "^3.0.3",
     "ai": "^6.0.170",

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

@@ -10,7 +10,6 @@ import { ERROR_TAGS, AiGenerationError, ConfigurationError } from '../effect/err
 import { RuntimeConfigServiceTag } from '../effect/services'
 import { openRouterEmbeddingModel } from '../embeddings/openrouter'
 import { isRecord, readString } from '../utils/string'
-import { buildAiGatewayCacheHeaders } from './cache-headers'
 
 type AiGatewayChatResponse = { body?: unknown }
 type AiGatewayTransformParamsOptions = Parameters<NonNullable<LanguageModelMiddleware['transformParams']>>[0]
@@ -473,23 +472,6 @@ function withAiGatewayStreamIdleTimeout(
   })
 }
 
-function mergeAiGatewayHeaders(
-  existingHeaders: AiGatewayCallOptions['headers'] | undefined,
-  additionalHeaders: Record<string, string>,
-): Record<string, string> {
-  const merged = new Headers(existingHeaders as HeadersInit | undefined)
-  for (const [key, value] of Object.entries(additionalHeaders)) {
-    if (!merged.has(key)) {
-      merged.set(key, value)
-    }
-  }
-  return Object.fromEntries(merged.entries())
-}
-
-function withDefaultAiGatewayCacheHeaders(params: AiGatewayCallOptions): AiGatewayCallOptions {
-  return { ...params, headers: mergeAiGatewayHeaders(params.headers, buildAiGatewayCacheHeaders('lota-sdk')) }
-}
-
 function normalizeAiGatewayUrlEffect(value: string): Effect.Effect<string, ConfigurationError> {
   const trimmed = value.trim()
   if (!trimmed) {
@@ -1122,13 +1104,11 @@ function createAiGatewayLanguageModelMiddleware(
     specificationVersion: 'v3',
     transformParams: ({ params, type }) =>
       Promise.resolve(
-        withDefaultAiGatewayCacheHeaders(
-          addAiGatewayReasoningRawChunks(
-            normalizeAiGatewayJsonSchemas(
-              providerId === OPENAI_CHAT_PROVIDER_ID ? normalizeAiGatewayChatProviderOptions(params, modelId) : params,
-            ),
-            type,
+        addAiGatewayReasoningRawChunks(
+          normalizeAiGatewayJsonSchemas(
+            providerId === OPENAI_CHAT_PROVIDER_ID ? normalizeAiGatewayChatProviderOptions(params, modelId) : params,
           ),
+          type,
         ),
       ),
     wrapGenerate: ({ params }) => {

package/src/ai-gateway/index.ts

@@ -20,4 +20,3 @@ export {
   normalizeAiGatewayUrl,
 } from './ai-gateway'
 export type { AiGatewayDeps, AiGatewayModels, RuntimeBridge } from './ai-gateway'
-export * from './cache-headers'

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 {

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

@@ -2,7 +2,6 @@ import type { CreateHelperToolLoopAgentOptions } from '@lota-sdk/shared'
 import { ToolLoopAgent } from 'ai'
 
 import type { AiGatewayModels } from '../ai-gateway/ai-gateway'
-import { buildAiGatewayDirectCacheHeaders } from '../ai-gateway/cache-headers'
 import { OPENROUTER_STRUCTURED_HELPER_MODEL_ID } from '../config/model-constants'
 import { resolveHelperAgentOptions } from './helper-agent-options'
 
@@ -33,7 +32,6 @@ export function makeContextCompactionAgentFactory(models: AiGatewayModels) {
     new ToolLoopAgent({
       id: 'context-compaction',
       model: models.chatModel(OPENROUTER_STRUCTURED_HELPER_MODEL_ID),
-      headers: buildAiGatewayDirectCacheHeaders('lota-sdk'),
       ...resolveHelperAgentOptions(options, { instructions: CONTEXT_COMPACTION_PROMPT }),
     })
 }

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

@@ -2,7 +2,6 @@ import type { CreateHelperToolLoopAgentOptions } from '@lota-sdk/shared'
 import { ToolLoopAgent } from 'ai'
 
 import type { AiGatewayModels } from '../ai-gateway/ai-gateway'
-import { buildAiGatewayDirectCacheHeaders } from '../ai-gateway/cache-headers'
 import {
   OPENROUTER_LOW_REASONING_PROVIDER_OPTIONS,
   OPENROUTER_STRUCTURED_HELPER_MODEL_ID,
@@ -34,7 +33,6 @@ export function makeMemoryRerankerAgentFactory(models: AiGatewayModels) {
     new ToolLoopAgent({
       id: 'memory-reranker',
       model: models.chatModel(OPENROUTER_STRUCTURED_HELPER_MODEL_ID),
-      headers: buildAiGatewayDirectCacheHeaders('lota-sdk'),
       providerOptions: OPENROUTER_LOW_REASONING_PROVIDER_OPTIONS,
       ...resolveHelperAgentOptions(options),
     })

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

@@ -2,7 +2,6 @@ import type { CreateHelperToolLoopAgentOptions } from '@lota-sdk/shared'
 import { ToolLoopAgent } from 'ai'
 
 import type { AiGatewayModels } from '../ai-gateway/ai-gateway'
-import { buildAiGatewayDirectCacheHeaders } from '../ai-gateway/cache-headers'
 import {
   OPENROUTER_STRUCTURED_HELPER_MODEL_ID,
   OPENROUTER_HIGH_REASONING_PROVIDER_OPTIONS,
@@ -54,7 +53,6 @@ export function makeOrgMemoryAgentFactory(models: AiGatewayModels) {
     new ToolLoopAgent({
       id: 'org-memory',
       model: models.chatModel(OPENROUTER_STRUCTURED_HELPER_MODEL_ID),
-      headers: buildAiGatewayDirectCacheHeaders('lota-sdk'),
       providerOptions: OPENROUTER_HIGH_REASONING_PROVIDER_OPTIONS,
       ...resolveHelperAgentOptions(options),
     })

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

@@ -2,7 +2,6 @@ import type { CreateHelperToolLoopAgentOptions } from '@lota-sdk/shared'
 import { ToolLoopAgent } from 'ai'
 
 import type { AiGatewayModels } from '../ai-gateway/ai-gateway'
-import { buildAiGatewayDirectCacheHeaders } from '../ai-gateway/cache-headers'
 import type { ResolvedAgentConfig } from '../config/agent-defaults'
 import { OPENROUTER_STRUCTURED_HELPER_MODEL_ID } from '../config/model-constants'
 import { resolveHelperAgentOptions } from './helper-agent-options'
@@ -78,7 +77,6 @@ export function makeRecentActivityTitleRefinerAgentFactory(models: AiGatewayMode
     new ToolLoopAgent({
       id: 'recent-activity-title-refiner',
       model: models.chatModel(OPENROUTER_STRUCTURED_HELPER_MODEL_ID),
-      headers: buildAiGatewayDirectCacheHeaders('lota-sdk'),
       ...resolveHelperAgentOptions(options, {
         instructions: buildRecentActivityTitleRefinerPrompt(agentConfig),
         maxOutputTokens: RECENT_ACTIVITY_TITLE_MAX_TOKENS,

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

@@ -2,7 +2,6 @@ import type { CreateHelperToolLoopAgentOptions } from '@lota-sdk/shared'
 import { ToolLoopAgent } from 'ai'
 
 import type { AiGatewayModels } from '../ai-gateway/ai-gateway'
-import { buildAiGatewayDirectCacheHeaders } from '../ai-gateway/cache-headers'
 import {
   OPENROUTER_STRUCTURED_HELPER_MODEL_ID,
   OPENROUTER_HIGH_REASONING_PROVIDER_OPTIONS,
@@ -29,7 +28,6 @@ export function makeRegularChatMemoryDigestAgentFactory(models: AiGatewayModels)
     new ToolLoopAgent({
       id: 'regular-chat-memory-digest',
       model: models.chatModel(OPENROUTER_STRUCTURED_HELPER_MODEL_ID),
-      headers: buildAiGatewayDirectCacheHeaders('lota-sdk'),
       providerOptions: OPENROUTER_HIGH_REASONING_PROVIDER_OPTIONS,
       ...resolveHelperAgentOptions(options, {
         instructions: regularChatMemoryDigestPrompt,

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

@@ -3,7 +3,6 @@ import { ToolLoopAgent } from 'ai'
 import { z } from 'zod'
 
 import type { AiGatewayModels } from '../ai-gateway/ai-gateway'
-import { buildAiGatewayDirectCacheHeaders } from '../ai-gateway/cache-headers'
 import {
   OPENROUTER_STRUCTURED_HELPER_MODEL_ID,
   OPENROUTER_HIGH_REASONING_PROVIDER_OPTIONS,
@@ -47,7 +46,6 @@ export function makeSkillExtractorAgentFactory(models: AiGatewayModels) {
     new ToolLoopAgent({
       id: 'skill-extractor',
       model: models.chatModel(OPENROUTER_STRUCTURED_HELPER_MODEL_ID),
-      headers: buildAiGatewayDirectCacheHeaders('lota-sdk'),
       providerOptions: OPENROUTER_HIGH_REASONING_PROVIDER_OPTIONS,
       ...resolveHelperAgentOptions(options, {
         instructions: skillExtractorPrompt,

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

@@ -3,7 +3,6 @@ import { ToolLoopAgent } from 'ai'
 import { z } from 'zod'
 
 import type { AiGatewayModels } from '../ai-gateway/ai-gateway'
-import { buildAiGatewayDirectCacheHeaders } from '../ai-gateway/cache-headers'
 import {
   OPENROUTER_STRUCTURED_HELPER_MODEL_ID,
   OPENROUTER_HIGH_REASONING_PROVIDER_OPTIONS,
@@ -71,7 +70,6 @@ export function makeSkillManagerAgentFactory(models: AiGatewayModels) {
     new ToolLoopAgent({
       id: 'skill-manager',
       model: models.chatModel(OPENROUTER_STRUCTURED_HELPER_MODEL_ID),
-      headers: buildAiGatewayDirectCacheHeaders('lota-sdk'),
       providerOptions: OPENROUTER_HIGH_REASONING_PROVIDER_OPTIONS,
       ...resolveHelperAgentOptions(options, {
         instructions: skillManagerPrompt,

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

@@ -3,7 +3,6 @@ import { Effect } from 'effect'
 import { z } from 'zod'
 
 import type { AiGatewayModels } from '../ai-gateway/ai-gateway'
-import { buildAiGatewayDirectCacheHeaders } from '../ai-gateway/cache-headers'
 import type { ResolvedAgentConfig } from '../config/agent-defaults'
 import { chatLogger } from '../config/logger'
 import type { ValidationError } from '../effect/errors'
@@ -182,7 +181,6 @@ function generateRouterObjectEffect<TSchema extends z.ZodTypeAny>(params: {
     try: () =>
       generateObject({
         model: params.aiGatewayModels.chatModel(modelId),
-        headers: buildAiGatewayDirectCacheHeaders('lota-sdk'),
         schema: params.schema,
         system: params.system,
         prompt: params.prompt,

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

@@ -2,7 +2,6 @@ import type { CreateHelperToolLoopAgentOptions } from '@lota-sdk/shared'
 import { ToolLoopAgent } from 'ai'
 
 import type { AiGatewayModels } from '../ai-gateway/ai-gateway'
-import { buildAiGatewayDirectCacheHeaders } from '../ai-gateway/cache-headers'
 import {
   OPENROUTER_FAST_REASONING_MODEL_ID,
   OPENROUTER_MINIMAL_REASONING_PROVIDER_OPTIONS,
@@ -35,7 +34,6 @@ export function makeThreadTitleGeneratorAgentFactory(models: AiGatewayModels) {
     new ToolLoopAgent({
       id: 'thread-title-generator',
       model: models.chatModel(OPENROUTER_FAST_REASONING_MODEL_ID),
-      headers: buildAiGatewayDirectCacheHeaders('lota-sdk'),
       providerOptions: OPENROUTER_MINIMAL_REASONING_PROVIDER_OPTIONS,
       ...resolveHelperAgentOptions(options, {
         instructions: THREAD_TITLE_GENERATOR_PROMPT,

package/src/tools/research-topic.tool.ts

@@ -3,7 +3,6 @@ import type { Effect } from 'effect'
 
 import type { AiGatewayModels } from '../ai-gateway/ai-gateway'
 import { aiGatewayChatModel, getDefaultAiGatewayRunPromise } from '../ai-gateway/ai-gateway'
-import { buildAiGatewayStrictSemanticCacheHeaders } from '../ai-gateway/cache-headers'
 import {
   OPENROUTER_FAST_REASONING_MODEL_ID,
   OPENROUTER_LOW_REASONING_PROVIDER_OPTIONS,
@@ -28,7 +27,6 @@ export const researchTopicTool = createDelegatedAgentToolWithContext<ResearchTop
       ? aiGatewayModels.chatModel(OPENROUTER_FAST_REASONING_MODEL_ID)
       : aiGatewayChatModel(OPENROUTER_FAST_REASONING_MODEL_ID),
   providerOptions: OPENROUTER_LOW_REASONING_PROVIDER_OPTIONS,
-  headers: buildAiGatewayStrictSemanticCacheHeaders('researchTopic'),
   instructions: RESEARCHER_PROMPT,
   createTools: ({ firecrawl, runPromise }) => ({
     searchWeb: searchWebTool.create({ firecrawl, runPromise }),

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

@@ -1,42 +0,0 @@
-const AI_GATEWAY_CACHE_KEY_HEADER = 'x-bf-cache-key'
-const AI_GATEWAY_CACHE_TTL_HEADER = 'x-bf-cache-ttl'
-const AI_GATEWAY_CACHE_THRESHOLD_HEADER = 'x-bf-cache-threshold'
-const AI_GATEWAY_CACHE_TYPE_HEADER = 'x-bf-cache-type'
-
-export const AI_GATEWAY_STRICT_SEMANTIC_CACHE_THRESHOLD = 0.975
-
-export type AiGatewayCacheType = 'direct' | 'semantic'
-
-export function toAiGatewayCacheKeyPart(value: string): string {
-  const normalized = value
-    .trim()
-    .toLowerCase()
-    .replace(/[^a-z0-9:_-]+/g, '-')
-    .replace(/-+/g, '-')
-  return normalized.replace(/^-+|-+$/g, '') || 'request'
-}
-
-export function buildAiGatewayCacheHeaders(
-  cacheKey: string,
-  ttl?: string,
-  threshold?: number,
-  cacheType?: AiGatewayCacheType,
-): Record<string, string> {
-  const headers: Record<string, string> = { [AI_GATEWAY_CACHE_KEY_HEADER]: cacheKey }
-  if (ttl) headers[AI_GATEWAY_CACHE_TTL_HEADER] = ttl
-  if (typeof threshold === 'number') headers[AI_GATEWAY_CACHE_THRESHOLD_HEADER] = String(threshold)
-  if (cacheType) headers[AI_GATEWAY_CACHE_TYPE_HEADER] = cacheType
-  return headers
-}
-
-export function buildAiGatewayDirectCacheHeaders(cacheKey: string, ttl?: string): Record<string, string> {
-  return buildAiGatewayCacheHeaders(cacheKey, ttl, undefined, 'direct')
-}
-
-export function buildAiGatewayStrictSemanticCacheHeaders(
-  cacheKey: string,
-  ttl?: string,
-  threshold = AI_GATEWAY_STRICT_SEMANTIC_CACHE_THRESHOLD,
-): Record<string, string> {
-  return buildAiGatewayCacheHeaders(cacheKey, ttl, threshold, 'semantic')
-}