llm-agent-runtime 0.1.16 → 0.1.21

package/USAGE.md

@@ -48,10 +48,16 @@ OPENROUTER_API_KEY=sk-or-v1-...
 OPENROUTER_MODEL=meta-llama/llama-3.1-70b-instruct
 OPENROUTER_INTENT_ROUTER_MODEL=meta-llama/llama-3.1-70b-instruct
 
-# Embeddings — omit EMBEDDINGS_PROVIDER → xenova when chat is openrouter
+# Embeddings — local default for openrouter is xenova (free)
 EMBEDDINGS_PROVIDER=xenova
 # XENOVA_EMBEDDINGS_MODEL=Xenova/all-MiniLM-L6-v2
 
+# Vercel / Lambda / Netlify: xenova will NOT load. Set:
+# EMBEDDINGS_PROVIDER=openai
+# OPENAI_API_KEY=sk-...
+# (If OPENAI_API_KEY is set on those hosts, the package auto-selects openai
+#  even when EMBEDDINGS_PROVIDER is unset or xenova.)
+
 # Optional Redis (flow state / cache). Plain Redis works for cache;
 # Redis Stack (RedisJSON) needed for durable LangGraph checkpoints.
 REDIS_URL=redis://localhost:6379
@@ -66,9 +72,9 @@ LANGSMITH_PROJECT=my-agent
 
 | `EMBEDDINGS_PROVIDER` | Cost | When to use |
 | --- | --- | --- |
-| `xenova` | Free | Default for OpenRouter/OpenAI chat when unset |
+| `xenova` | Free | Local default for OpenRouter/OpenAI chat when unset |
 | `ollama` | Free (self-host) | Default when `LLM_PROVIDER=ollama` |
-| `openai` | Paid | Strongest cloud embeddings |
+| `openai` | Paid | **Required on Vercel/serverless**; strongest cloud embeddings |
 
 ---
 
@@ -577,11 +583,12 @@ await runtime.invoke({
     turn:  { inputTokens, outputTokens, totalTokens }  // this invoke
     flow:  { …, intent }                               // sticky flow bucket
     total: { inputTokens, outputTokens, totalTokens }  // whole thread
+    intents: Record<string, { inputTokens, outputTokens, totalTokens }> // every intent
   }
 }
 ```
 
-Token totals are stored in **Redis** when `redisUrl` is set (`usage:thread:{id}`, `usage:flow:{id}:{intent}`), otherwise in memory.
+Token totals are stored in **Redis** when `redisUrl` is set (`usage:thread:{id}`, `usage:flow:{id}:{intent}`, `usage:intent:{id}:{intent}`), otherwise in memory.
 
 ### Clear session (frontend)
 
@@ -589,7 +596,8 @@ Token totals are stored in **Redis** when `redisUrl` is set (`usage:thread:{id}`
 // Clears checkpoint history + sticky flow + token usage for this threadId
 await runtime.clearSession(threadId);
 // aliases: clearThread(threadId)  |  clearFlow(threadId) only sticky state
-const total = await runtime.getUsage(threadId);
+const usage = await runtime.getUsage(threadId);
+// usage.total / usage.flow / usage.intents
 ```
 
 Example HTTP for the UI “New chat / Clear session” button:
@@ -684,4 +692,4 @@ import {
 
 ## Version
 
-Current local package: **0.1.16** (`defineIntents` / `withGeneralIntents` unified intent objects, `getRagStore` singleton, Example B aligned with `plain-agent-demo`, Xenova embeddings, token usage + `clearSession`).
+Current local package: **0.1.19** (`TokenUsageReport` with `turn` / `flow` / `total` / `intents`, serverless OpenAI embeddings, `getRagStore`, unified intents).

package/dist/agent.d.ts

@@ -18,4 +18,3 @@ export declare function createAgentFactory(options: {
     redisUrl?: string;
     checkpointer?: BaseCheckpointSaver;
 }): Promise<AgentFactory>;
-//# sourceMappingURL=agent.d.ts.map

package/dist/agent.js

@@ -1,69 +1 @@
-import { createReactAgent } from "@langchain/langgraph/prebuilt";
-import { MemorySaver } from "@langchain/langgraph-checkpoint";
-import { RedisSaver } from "@langchain/langgraph-checkpoint-redis";
-import { createClient } from "redis";
-/**
- * RedisSaver needs Redis Stack (RedisJSON + RediSearch).
- * Plain redis://localhost has neither — probe and fall back to MemorySaver.
- */
-async function createCheckpointer(redisUrl, existing) {
-    if (existing)
-        return existing;
-    if (!redisUrl)
-        return new MemorySaver();
-    const client = createClient({ url: redisUrl });
-    client.on("error", () => {
-        /* probe client — swallow to avoid process crash */
-    });
-    try {
-        await client.connect();
-        const probeKey = `__llm_agent_runtime_probe_${Date.now()}`;
-        await client.sendCommand(["JSON.SET", probeKey, "$", "{}"]);
-        await client.del(probeKey);
-        await client.quit().catch(() => undefined);
-    }
-    catch {
-        await client.quit().catch(() => undefined);
-        console.warn("[llm-agent-runtime] Redis Stack (RedisJSON) not available — using in-memory checkpointer. " +
-            "Flow cache can still use Redis; install Redis Stack for durable LangGraph checkpoints.");
-        return new MemorySaver();
-    }
-    try {
-        return await RedisSaver.fromUrl(redisUrl);
-    }
-    catch (err) {
-        console.warn("[llm-agent-runtime] RedisSaver.fromUrl failed — using in-memory checkpointer:", err instanceof Error ? err.message : err);
-        return new MemorySaver();
-    }
-}
-export async function createAgentFactory(options) {
-    const checkpointer = await createCheckpointer(options.redisUrl, options.checkpointer);
-    const fallbackTools = () => {
-        const first = Object.values(options.intentRegistry)[0];
-        return first?.tools("") ?? [];
-    };
-    const stateReader = createReactAgent({
-        llm: options.llm,
-        tools: fallbackTools(),
-        checkpointSaver: checkpointer,
-    });
-    return {
-        checkpointer,
-        createAgent: (phoneKey, intent) => {
-            const tools = intent
-                ? (options.intentRegistry[intent]?.tools(phoneKey) ?? [])
-                : fallbackTools();
-            return createReactAgent({
-                llm: options.llm,
-                tools,
-                checkpointSaver: checkpointer,
-            });
-        },
-        getAgentState: async (threadId) => {
-            const state = await stateReader.getState({
-                configurable: { thread_id: threadId },
-            });
-            return state;
-        },
-    };
-}
+function a0_0x2798(_0x46626c,_0x3671ca){_0x46626c=_0x46626c-0xbf;const _0x27767b=a0_0x2776();let _0x279836=_0x27767b[_0x46626c];return _0x279836;}function a0_0x2776(){const _0x15919e=['268HUNDSt','2124916MfwNoy','jEnaW','QIekl','209511tnJlgA','cbJdr','values','1057220bTCxgu','getState','del','message','OYmgB','fromUrl','connect','[llm-agent-runtime]\x20RedisSaver.fromUrl\x20failed\x20—\x20using\x20in-memory\x20checkpointer:','redisUrl','80PfCQEv','17495ojsEax','now','sendCommand','djDBR','6997566UkwJrw','JSON.SET','Unset\x20REDIS_URL\x20for\x20faster\x20local\x20startup,\x20or\x20set\x20REDIS_ENABLED=true\x20with\x20a\x20reachable\x20Redis\x20Stack.','pcTGn','tools','warn','llm','10736348zeMcuZ','1196394laVHmC','__llm_agent_runtime_probe_','ZgrNd','catch','quit'];a0_0x2776=function(){return _0x15919e;};return a0_0x2776();}(function(_0x5855e3,_0xc72a04){const _0x314b41=a0_0x2798,_0x2a9d65=_0x5855e3();while(!![]){try{const _0x3eb0fc=-parseInt(_0x314b41(0xde))/0x1+parseInt(_0x314b41(0xd8))/0x2+parseInt(_0x314b41(0xd2))/0x3+-parseInt(_0x314b41(0xd7))/0x4*(-parseInt(_0x314b41(0xc6))/0x5)+-parseInt(_0x314b41(0xca))/0x6+parseInt(_0x314b41(0xd1))/0x7+parseInt(_0x314b41(0xc5))/0x8*(-parseInt(_0x314b41(0xdb))/0x9);if(_0x3eb0fc===_0xc72a04)break;else _0x2a9d65['push'](_0x2a9d65['shift']());}catch(_0x22de68){_0x2a9d65['push'](_0x2a9d65['shift']());}}}(a0_0x2776,0xbcc3e));import{createReactAgent}from'@langchain/langgraph/prebuilt';import{MemorySaver}from'@langchain/langgraph-checkpoint';import{RedisSaver}from'@langchain/langgraph-checkpoint-redis';import{createClient}from'redis';async function createCheckpointer(_0x29f934,_0x3b1593){const _0x2a0439=a0_0x2798,_0x18c381={'QIekl':'error','ZgrNd':_0x2a0439(0xcb),'djDBR':function(_0x11c180,_0x5840ef){return _0x11c180+_0x5840ef;},'jEnaW':_0x2a0439(0xcc)};if(_0x3b1593)return _0x3b1593;if(!_0x29f934)return new MemorySaver();const _0xc8078f=createClient({'url':_0x29f934,'socket':{'connectTimeout':0x1388}});_0xc8078f['on'](_0x18c381[_0x2a0439(0xda)],()=>{});try{await _0xc8078f[_0x2a0439(0xc2)]();const _0x5257a5=_0x2a0439(0xd3)+Date[_0x2a0439(0xc7)]();await _0xc8078f[_0x2a0439(0xc8)]([_0x18c381[_0x2a0439(0xd4)],_0x5257a5,'$','{}']),await _0xc8078f[_0x2a0439(0xe0)](_0x5257a5),await _0xc8078f[_0x2a0439(0xd6)]()[_0x2a0439(0xd5)](()=>undefined);}catch{return await _0xc8078f[_0x2a0439(0xd6)]()[_0x2a0439(0xd5)](()=>undefined),console[_0x2a0439(0xcf)](_0x18c381[_0x2a0439(0xc9)]('[llm-agent-runtime]\x20Redis\x20unreachable\x20or\x20missing\x20RedisJSON\x20—\x20using\x20in-memory\x20checkpointer.\x20',_0x18c381[_0x2a0439(0xd9)])),new MemorySaver();}try{return await RedisSaver[_0x2a0439(0xc1)](_0x29f934);}catch(_0x47962d){return console[_0x2a0439(0xcf)](_0x2a0439(0xc3),_0x47962d instanceof Error?_0x47962d[_0x2a0439(0xbf)]:_0x47962d),new MemorySaver();}}export async function createAgentFactory(_0x2c7f9b){const _0x42e45c=a0_0x2798,_0x581f66={'OYmgB':function(_0x254acd,_0x2e9bf5){return _0x254acd(_0x2e9bf5);},'pcTGn':function(_0xf0b3ed,_0x523610,_0x5e4527){return _0xf0b3ed(_0x523610,_0x5e4527);},'cbJdr':function(_0x14e109){return _0x14e109();}},_0x717be5=await _0x581f66[_0x42e45c(0xcd)](createCheckpointer,_0x2c7f9b[_0x42e45c(0xc4)],_0x2c7f9b['checkpointer']),_0x236621=()=>{const _0x1b8d93=_0x42e45c,_0x23453e=Object[_0x1b8d93(0xdd)](_0x2c7f9b['intentRegistry'])[0x0];return _0x23453e?.[_0x1b8d93(0xce)]('')??[];},_0x8fff0b=_0x581f66[_0x42e45c(0xc0)](createReactAgent,{'llm':_0x2c7f9b[_0x42e45c(0xd0)],'tools':_0x581f66[_0x42e45c(0xdc)](_0x236621),'checkpointSaver':_0x717be5});return{'checkpointer':_0x717be5,'createAgent':(_0x374733,_0x11a7e9)=>{const _0x2d666a=_0x42e45c,_0x11a00f=_0x11a7e9?_0x2c7f9b['intentRegistry'][_0x11a7e9]?.['tools'](_0x374733)??[]:_0x236621();return _0x581f66[_0x2d666a(0xc0)](createReactAgent,{'llm':_0x2c7f9b[_0x2d666a(0xd0)],'tools':_0x11a00f,'checkpointSaver':_0x717be5});},'getAgentState':async _0x2a6bf5=>{const _0x36d3e7=_0x42e45c,_0x535c37=await _0x8fff0b[_0x36d3e7(0xdf)]({'configurable':{'thread_id':_0x2a6bf5}});return _0x535c37;}};}

package/dist/create-runtime.d.ts

@@ -11,8 +11,11 @@ export type AgentRuntime<TContext = unknown> = {
      * Same as clearThread + usage wipe.
      */
     clearSession: (threadId: string) => Promise<void>;
-    /** Read accumulated token usage for a thread (Redis or memory). */
-    getUsage: (threadId: string) => Promise<TokenUsageReport["total"]>;
+    /**
+     * Full token usage for a thread: `total`, sticky `flow`, and per-intent `intents`.
+     * `turn` is zeros here (only populated on `invoke`).
+     */
+    getUsage: (threadId: string) => Promise<TokenUsageReport>;
     getCachedAnswer: (message: string) => Promise<string | undefined>;
     setCachedAnswer: (message: string, answer: string) => Promise<void>;
     /** Invalidate embedding cache after changing intent examples at runtime. */
@@ -26,4 +29,3 @@ export type AgentRuntime<TContext = unknown> = {
  * `createAgentRuntime<{ userInfo?: { name?: string } }>({ ... })`
  */
 export declare function createAgentRuntime<TContext = unknown>(config: AgentRuntimeConfig<TContext>): Promise<AgentRuntime<TContext>>;
-//# sourceMappingURL=create-runtime.d.ts.map

package/dist/create-runtime.js

@@ -1,321 +1 @@
-import { createAgentFactory } from "./agent.js";
-import { createModels } from "./llm.js";
-import { llmConfigFromEnv } from "./env-llm.js";
-import { createRedis } from "./redis.js";
-import { createAgentRunner } from "./services/agent-runner.js";
-import { createChatCache } from "./services/chat-cache.js";
-import { createConversationService } from "./services/conversation.js";
-import { createToolOutputExtractor, extractRawToolOutputs } from "./services/extract.js";
-import { createCompleteOnFinalize, toolOutputMapToExtractors, } from "./services/simple-config.js";
-import { createFlowRouter, stickySetFromIntents, } from "./services/flow-router.js";
-import { createIntentEmbeddings } from "./services/intent-embeddings.js";
-import { createIntentRouter } from "./services/intent-router.js";
-import { extractTokenUsageFromMessages } from "./services/token-usage.js";
-import { createUsageStore } from "./services/usage-store.js";
-import { assertInvokeInput } from "./errors.js";
-import { getLastMessageContent } from "./utils/message.js";
-/** Messages for this user turn only (used to extract tool outputs). */
-const getTurnMessages = (allMessages, userMessage) => {
-    const normalized = userMessage.trim();
-    const list = allMessages;
-    for (let i = list.length - 1; i >= 0; i--) {
-        const content = String(list[i]?.content ?? "").trim();
-        if (content === normalized) {
-            return list.slice(i);
-        }
-    }
-    return list.slice(-6);
-};
-/**
- * Factory for a reusable agent runtime.
- * Engine pipeline: resolveFlow → runAgent → getTurnMessages → extract → finalize
- *
- * Pass your context type so prompts get typed `ctx.context`:
- * `createAgentRuntime<{ userInfo?: { name?: string } }>({ ... })`
- */
-export async function createAgentRuntime(config) {
-    // Wire LangSmith tracing before models are created.
-    // Respect LANGSMITH_PROJECT / LANGCHAIN_PROJECT (do not force a package default over env).
-    const langsmithKey = process.env.LANGSMITH_API_KEY?.trim() ||
-        process.env.LANGCHAIN_API_KEY?.trim() ||
-        undefined;
-    const tracingEnabled = config.tracing?.enabled ?? Boolean(langsmithKey);
-    if (tracingEnabled) {
-        if (!langsmithKey) {
-            console.warn("[llm-agent-runtime] tracing.enabled but no LANGSMITH_API_KEY / LANGCHAIN_API_KEY set");
-        }
-        else {
-            const project = config.tracing?.projectName?.trim() ||
-                process.env.LANGSMITH_PROJECT?.trim() ||
-                process.env.LANGCHAIN_PROJECT?.trim() ||
-                "llm-agent-runtime";
-            process.env.LANGCHAIN_TRACING_V2 = "true";
-            process.env.LANGSMITH_TRACING = "true";
-            process.env.LANGCHAIN_API_KEY = langsmithKey;
-            process.env.LANGSMITH_API_KEY = langsmithKey;
-            process.env.LANGCHAIN_PROJECT = project;
-            process.env.LANGSMITH_PROJECT = project;
-            console.log("[llm-agent-runtime] LangSmith tracing on", { project });
-        }
-    }
-    const defaultIntent = config.defaultIntent ?? config.intents[0]?.name ?? "greetings";
-    const intentNames = new Set(config.intents.map((i) => i.name));
-    const aliases = config.intentAliases ?? {};
-    const normalizeIntent = (intent) => {
-        const key = intent
-            .trim()
-            .toLowerCase()
-            .replace(/[\s-]+/g, "_");
-        if (intentNames.has(key))
-            return key;
-        return aliases[key] ?? defaultIntent;
-    };
-    const stickyFlows = stickySetFromIntents(config.intents);
-    const menuMap = config.menuMap ?? {};
-    const llm = config.llm ?? llmConfigFromEnv();
-    const created = createModels(llm);
-    const chat = config.models?.chat ?? created.chat;
-    const intentRouterModel = config.models?.intentRouter ?? created.intentRouter;
-    const embeddings = config.models?.embeddings ?? created.embeddings;
-    console.log("[llm-agent-runtime] models", {
-        provider: created.config.provider,
-        chat: created.config.model,
-        intentRouter: created.config.intentRouterModel,
-        embeddingsProvider: created.config.embeddingsProvider,
-        embeddings: created.config.embeddingsModel,
-        baseUrl: llm.baseUrl,
-        embeddingsBaseUrl: llm.embeddingsBaseUrl,
-    });
-    const fromMap = config.toolOutputMap
-        ? toolOutputMapToExtractors(config.toolOutputMap)
-        : null;
-    const toolExtractors = config.toolExtractors ?? fromMap?.toolExtractors;
-    const initialToolOutputs = config.initialToolOutputs ?? fromMap?.initialToolOutputs;
-    const completeOn = config.completeOn ??
-        (config.toolOutputMap ? Object.keys(config.toolOutputMap) : undefined);
-    const finalizeFlow = config.finalizeFlow ??
-        (completeOn?.length
-            ? createCompleteOnFinalize({
-                completeOn,
-                toolOutputMap: config.toolOutputMap,
-            })
-            : undefined);
-    // Redis client when redisUrl is set (flow state + optional answer cache + usage)
-    const redis = config.redisUrl
-        ? createRedis(config.redisUrl)
-        : undefined;
-    const chatCache = createChatCache({
-        redis: redis?.client,
-        config: {
-            enabled: config.cache?.enabled ?? false,
-            ttlSeconds: config.cache?.ttlSeconds,
-            flowTtlSeconds: config.cache?.flowTtlSeconds,
-            scope: config.cache?.scope,
-        },
-    });
-    const usageStore = createUsageStore({
-        redis: redis?.client,
-        ttlSeconds: config.cache?.ttlSeconds ?? 60 * 60 * 24 * 7,
-    });
-    const agentFactory = await createAgentFactory({
-        llm: chat,
-        intentRegistry: config.intentRegistry,
-        redisUrl: config.redisUrl,
-        checkpointer: config.checkpointer,
-    });
-    const conversation = createConversationService(agentFactory);
-    const { runAgent } = createAgentRunner({
-        agentFactory,
-        recursionLimit: config.recursionLimit,
-    });
-    const intentEmbeddings = createIntentEmbeddings({
-        embeddings,
-        intents: config.intents,
-        normalizeIntent,
-    });
-    const intentRouter = createIntentRouter({
-        model: intentRouterModel,
-        intents: config.intents,
-        normalizeIntent,
-        defaultIntent,
-        routerSystemPrompt: config.routerSystemPrompt,
-        llmConfidence: config.thresholds?.llmConfidence,
-        embeddingConfidence: config.thresholds?.embeddingConfidence,
-    });
-    const { resolveFlow, finalizeAgentFlow } = createFlowRouter({
-        stickyFlows,
-        menuMap,
-        normalizeIntent,
-        intents: config.intents,
-        getEmbeddedIntents: intentEmbeddings.getEmbeddedIntents,
-        scoreIntentsForMessage: intentEmbeddings.scoreIntentsForMessage,
-        classifyIntentWithLlm: intentRouter.classifyIntentWithLlm,
-        mergeIntentResults: intentRouter.mergeIntentResults,
-        chatCache,
-        finalizeFlow,
-        thresholds: config.thresholds,
-    });
-    const extractToolOutputs = config.extractToolOutputs ??
-        (toolExtractors
-            ? createToolOutputExtractor(toolExtractors, initialToolOutputs, {
-                includeUnknown: config.includeUnknownToolOutputs ?? false,
-            })
-            : extractRawToolOutputs);
-    const resolveSharedPrompt = (ctx) => {
-        const shared = config.sharedPrompt;
-        if (!shared)
-            return "";
-        return typeof shared === "function" ? shared(ctx) : shared;
-    };
-    const buildIntentPrompt = (ctx) => {
-        const binding = config.intentRegistry[ctx.intent];
-        if (!binding) {
-            const fallback = Object.values(config.intentRegistry)[0];
-            return fallback?.prompt(ctx) ?? "";
-        }
-        return binding.prompt(ctx);
-    };
-    const buildSystemPrompt = config.buildSystemPrompt ??
-        ((ctx) => {
-            const shared = resolveSharedPrompt(ctx).trim();
-            const intentPrompt = buildIntentPrompt(ctx).trim();
-            if (shared && intentPrompt)
-                return `${shared}\n\n${intentPrompt}`;
-            return shared || intentPrompt;
-        });
-    const resolvePhoneKey = config.resolvePhoneKey ??
-        (({ threadId, phoneKey }) => phoneKey ?? threadId);
-    const clearSession = async (threadId) => {
-        const saver = agentFactory.checkpointer;
-        if (typeof saver.deleteThread === "function") {
-            await saver.deleteThread(threadId);
-        }
-        await chatCache.clearChatCacheFlowState(threadId);
-        await usageStore.clearThreadUsage(threadId);
-    };
-    const invoke = async (input) => {
-        assertInvokeInput(input);
-        if (chatCache.enabled) {
-            const cached = await chatCache.getCachedAnswer(input.message);
-            if (cached) {
-                const total = await usageStore.getThreadUsage(input.threadId);
-                return {
-                    messages: [],
-                    answer: cached,
-                    intent: { intent: defaultIntent, score: 1 },
-                    flow: { active: false, intent: null },
-                    toolOutputs: {},
-                    flowCompleted: false,
-                    extras: { cacheHit: true },
-                    usage: {
-                        turn: { inputTokens: 0, outputTokens: 0, totalTokens: 0 },
-                        flow: {
-                            inputTokens: 0,
-                            outputTokens: 0,
-                            totalTokens: 0,
-                            intent: null,
-                        },
-                        total,
-                    },
-                };
-            }
-        }
-        const cachedFlowState = await chatCache.getChatCacheFlowState(input.threadId);
-        const [messages, conversationContext] = await Promise.all([
-            conversation.getManagedConversationMessages(input.threadId),
-            conversation.getConversationContextForIntent(input.threadId),
-        ]);
-        const { messageIntent, flow, promptIntent, llmResult } = await resolveFlow({
-            message: input.message,
-            explicitIntent: input.explicitIntent,
-            cachedFlowState,
-            conversationContext,
-        });
-        const phoneKey = resolvePhoneKey({
-            threadId: input.threadId,
-            phoneKey: input.phoneKey,
-            context: input.context,
-        });
-        const systemPrompt = buildSystemPrompt({
-            hasConversationHistory: messages.length > 0,
-            intent: promptIntent,
-            intentScore: flow.active ? 1 : messageIntent.score,
-            message: input.message,
-            context: input.context,
-        });
-        const result = await runAgent({
-            message: input.message,
-            threadId: input.threadId,
-            messages,
-            systemPrompt,
-            phoneKey,
-            intent: promptIntent,
-        });
-        const resultMessages = result.messages;
-        const turnMessages = getTurnMessages(resultMessages, input.message);
-        const toolOutputs = extractToolOutputs(turnMessages);
-        const finalized = await finalizeAgentFlow({
-            threadId: input.threadId,
-            promptIntent,
-            flow,
-            toolOutputs,
-        });
-        const answer = getLastMessageContent(resultMessages);
-        if (chatCache.enabled && answer) {
-            await chatCache.setCachedAnswer(input.message, answer);
-        }
-        // Token usage: this turn + sticky-flow bucket + thread total (Redis when set)
-        const turnUsage = extractTokenUsageFromMessages(turnMessages);
-        const flowIntentForUsage = finalized.flow.active && finalized.flow.intent
-            ? finalized.flow.intent
-            : stickyFlows.has(promptIntent)
-                ? promptIntent
-                : null;
-        const { total, flow: flowUsage } = await usageStore.recordTurn({
-            threadId: input.threadId,
-            flowIntent: flowIntentForUsage,
-            turn: turnUsage,
-        });
-        if (finalized.flowCompleted && flowIntentForUsage) {
-            // Keep completed-flow totals in this response; reset bucket for next flow
-            await usageStore.clearFlowUsage(input.threadId, flowIntentForUsage);
-        }
-        const usage = {
-            turn: turnUsage,
-            flow: {
-                ...flowUsage,
-                intent: flowIntentForUsage,
-            },
-            total,
-        };
-        return {
-            messages: resultMessages,
-            answer,
-            intent: messageIntent,
-            flow: finalized.flow,
-            toolOutputs,
-            flowCompleted: finalized.flowCompleted,
-            extras: finalized.extras,
-            usage,
-            ...(llmResult
-                ? { router: { action: llmResult.action, reason: llmResult.reason } }
-                : {}),
-        };
-    };
-    return {
-        invoke,
-        connect: async () => {
-            await redis?.connect();
-        },
-        disconnect: async () => {
-            await redis?.disconnect();
-        },
-        clearFlow: chatCache.clearChatCacheFlowState,
-        clearThread: clearSession,
-        clearSession,
-        getUsage: usageStore.getThreadUsage,
-        getCachedAnswer: chatCache.getCachedAnswer,
-        setCachedAnswer: chatCache.setCachedAnswer,
-        invalidateIntentEmbeddings: intentEmbeddings.invalidate,
-    };
-}
+(function(_0x37f28f,_0x1efeb1){const _0x2312bb=a1_0x26c0,_0xd94309=_0x37f28f();while(!![]){try{const _0x38301d=-parseInt(_0x2312bb(0x126))/0x1*(parseInt(_0x2312bb(0xf6))/0x2)+parseInt(_0x2312bb(0x134))/0x3+-parseInt(_0x2312bb(0xfc))/0x4+parseInt(_0x2312bb(0x125))/0x5+parseInt(_0x2312bb(0xc9))/0x6+-parseInt(_0x2312bb(0xdb))/0x7+parseInt(_0x2312bb(0x107))/0x8;if(_0x38301d===_0x1efeb1)break;else _0xd94309['push'](_0xd94309['shift']());}catch(_0x486d89){_0xd94309['push'](_0xd94309['shift']());}}}(a1_0xdce1,0x3b779));import{createAgentFactory}from'./agent.js';import{createModels}from'./llm.js';import{llmConfigFromEnv}from'./env-llm.js';function a1_0x26c0(_0x2dd741,_0x438f87){_0x2dd741=_0x2dd741-0xb3;const _0xdce16d=a1_0xdce1();let _0x26c02a=_0xdce16d[_0x2dd741];return _0x26c02a;}import{createRedis}from'./redis.js';import{createAgentRunner}from'./services/agent-runner.js';import{createChatCache}from'./services/chat-cache.js';import{createConversationService}from'./services/conversation.js';import{createToolOutputExtractor,extractRawToolOutputs}from'./services/extract.js';import{createCompleteOnFinalize,toolOutputMapToExtractors}from'./services/simple-config.js';import{createFlowRouter,stickySetFromIntents}from'./services/flow-router.js';import{createIntentEmbeddings}from'./services/intent-embeddings.js';import{createIntentRouter}from'./services/intent-router.js';function a1_0xdce1(){const _0x5aa009=['initialToolOutputs','phoneKey','vHCvy','config','6763376zhnpID','QXagV','provider','yufJG','GLhOT','getUsageReport','Gxuoi','BEBvA','all','keys','scope','reason','active','ZXWKW','LANGSMITH_PROJECT','YVMIP','xFldO','baseUrl','toolOutputMap','aAuDB','LANGCHAIN_TRACING_V2','uxFQj','llmConfidence','EECKM','intentRegistry','mergeIntentResults','xkeqP','score','jQeQk','intentRouterModel','620230WoxoDA','3799dCkoIe','MkDlt','sharedPrompt','slice','defaultIntent','rkNNI','embeddingsModel','QlqeD','VmghZ','toLowerCase','RhptV','intents','LANGSMITH_API_KEY','threadId','139398SgSPOd','nRfCR','embeddingsProvider','djvnR','length','fssOw','utjSj','getManagedConversationMessages','values','redisUrl','enjVU','embeddingConfidence','thresholds','cdvcu','LANGCHAIN_PROJECT','PCXYe','XHRnh','rPYWG','menuMap','message','getConversationContextForIntent','models','projectName','flow','has','recursionLimit','trim','scoreIntentsForMessage','name','910566LvEQQS','flowTtlSeconds','getCachedAnswer','checkpointer','llm-agent-runtime','messages','tracing','llm','UIxRp','GeWDw','Fhucx','function','nUekq','embeddings','content','buildSystemPrompt','bKhuW','[llm-agent-runtime]\x20models','2969064wsGfal','model','vhEuv','cNHEV','extras','ttlSeconds','log','toolExtractors','uipWy','uZhOM','cache','WJCkC','client','LANGCHAIN_API_KEY','completeOn','LANGSMITH_TRACING','warn','replace','deleteThread','BwQvj','[llm-agent-runtime]\x20LangSmith\x20tracing\x20on','context','iheci','enabled','intent','intentRouter','env','30nVUINe','setCachedAnswer','nOhCN','UIwOX','classifyIntentWithLlm','OtHgO','1771924JoYyHg','flowCompleted','fwBMf','extractToolOutputs','chat','prompt','tsdEE'];a1_0xdce1=function(){return _0x5aa009;};return a1_0xdce1();}import{extractTokenUsageFromMessages}from'./services/token-usage.js';import{createUsageStore}from'./services/usage-store.js';import{assertInvokeInput}from'./errors.js';import{getLastMessageContent}from'./utils/message.js';const getTurnMessages=(_0x4f4d26,_0x5f48a2)=>{const _0x3a9b7a=a1_0x26c0,_0x3b4b9d={'OtHgO':function(_0x2048da,_0x1db9e4){return _0x2048da-_0x1db9e4;},'RSoXj':function(_0x54c578,_0x34919b){return _0x54c578>=_0x34919b;},'YVMIP':function(_0x2d75ba,_0x2f7863){return _0x2d75ba(_0x2f7863);},'rkNNI':function(_0x5b6fba,_0x4a44b8){return _0x5b6fba!==_0x4a44b8;},'WoIvH':'LMUqd'},_0x17069e=_0x5f48a2[_0x3a9b7a(0xc6)](),_0x1799ef=_0x4f4d26;for(let _0x5b7efc=_0x3b4b9d[_0x3a9b7a(0xfb)](_0x1799ef[_0x3a9b7a(0x138)],0x1);_0x3b4b9d['RSoXj'](_0x5b7efc,0x0);_0x5b7efc--){const _0x16cd1f=_0x3b4b9d[_0x3a9b7a(0x116)](String,_0x1799ef[_0x5b7efc]?.[_0x3a9b7a(0xd7)]??'')[_0x3a9b7a(0xc6)]();if(_0x16cd1f===_0x17069e){if(_0x3b4b9d[_0x3a9b7a(0x12b)](_0x3b4b9d['WoIvH'],_0x3b4b9d['WoIvH'])){const _0x365287=_0x70732a[_0x3a9b7a(0xc6)]()['toLowerCase']()['replace'](/[\s-]+/g,'_');if(_0x533cca[_0x3a9b7a(0xc4)](_0x365287))return _0x365287;return _0x14b8c9[_0x365287]??_0x40f3c3;}else return _0x1799ef[_0x3a9b7a(0x129)](_0x5b7efc);}}return _0x1799ef[_0x3a9b7a(0x129)](-0x6);};export async function createAgentRuntime(_0x870a0d){const _0x1f35dc=a1_0x26c0,_0x27c671={'ZXWKW':function(_0x2c8e3e,_0x537387){return _0x2c8e3e!==_0x537387;},'rPYWG':_0x1f35dc(0x102),'BwQvj':function(_0x26769e,_0x130dc5){return _0x26769e===_0x130dc5;},'vHCvy':function(_0x149c72,_0x38b3ec){return _0x149c72(_0x38b3ec);},'WJCkC':_0x1f35dc(0xd4),'nRfCR':_0x1f35dc(0xcd),'nUekq':'true','yufJG':function(_0x490811,_0x1e5dd5){return _0x490811===_0x1e5dd5;},'QXagV':_0x1f35dc(0x11e),'uxFQj':function(_0x4764a1,_0x2b57e9){return _0x4764a1(_0x2b57e9);},'uZhOM':function(_0x15e9a4,_0x2f2eae){return _0x15e9a4||_0x2f2eae;},'GVCuO':function(_0x3c2cef,_0x362541){return _0x3c2cef(_0x362541);},'jwEZy':_0x1f35dc(0xe3),'PCZrp':function(_0x37b28c,_0x46392c){return _0x37b28c===_0x46392c;},'GeWDw':'aAuDB','xkeqP':'[llm-agent-runtime]\x20tracing.enabled\x20but\x20no\x20LANGSMITH_API_KEY\x20/\x20LANGCHAIN_API_KEY\x20set','BEBvA':function(_0x2e180a,_0x141cd){return _0x2e180a===_0x141cd;},'vUhaU':_0x1f35dc(0x13a),'bKhuW':function(_0x75c083,_0x2bd95b){return _0x75c083(_0x2bd95b);},'fwBMf':function(_0xeeb14d,_0x23cb24){return _0xeeb14d>_0x23cb24;},'vhEuv':function(_0x52ba7c,_0x2e35af){return _0x52ba7c(_0x2e35af);},'CDPBe':function(_0x3d4fb4,_0x5e2c23,_0x3de703){return _0x3d4fb4(_0x5e2c23,_0x3de703);},'nOhCN':function(_0x5ef1d0,_0x293ffc){return _0x5ef1d0(_0x293ffc);},'GLhOT':_0x1f35dc(0xd1),'iheci':function(_0x1028bc,_0x48f59f){return _0x1028bc(_0x48f59f);},'ivcyd':'TApWO','xFldO':'KxKKv','QlqeD':_0x1f35dc(0x139),'PjBUa':_0x1f35dc(0xef),'cNHEV':function(_0x263404){return _0x263404();},'UIwOX':function(_0x393dfa,_0x22d21a){return _0x393dfa(_0x22d21a);},'VmghZ':_0x1f35dc(0xda),'OCMTj':function(_0x2f154a,_0x2fba17){return _0x2f154a(_0x2fba17);},'enjVU':function(_0x1c06b6,_0x58a4b5){return _0x1c06b6(_0x58a4b5);},'MueHD':function(_0x13992c,_0x534e9d){return _0x13992c*_0x534e9d;},'cMRTi':function(_0x3bb74b,_0x16fbe0){return _0x3bb74b*_0x16fbe0;}},_0x4e521a=process[_0x1f35dc(0xf5)]['LANGSMITH_API_KEY']?.[_0x1f35dc(0xc6)]()||process[_0x1f35dc(0xf5)][_0x1f35dc(0xe8)]?.[_0x1f35dc(0xc6)]()||undefined,_0x110fe0=_0x870a0d['tracing']?.[_0x1f35dc(0xf2)]??_0x27c671[_0x1f35dc(0x11c)](Boolean,_0x4e521a);if(_0x110fe0){if(!_0x4e521a){if(_0x27c671[_0x1f35dc(0x114)](_0x27c671[_0x1f35dc(0x12d)],_0x1f35dc(0xbc)))console['warn'](_0x27c671['xkeqP']);else{const _0x278108=_0x516d35[_0x1f35dc(0x11f)][_0x43d1f6[_0x1f35dc(0xf3)]];if(!_0x278108){const _0x1f2ba5=_0x3793d4['values'](_0x2c978b['intentRegistry'])[0x0];return _0x1f2ba5?.[_0x1f35dc(0x101)](_0x474dc4)??'';}return _0x278108[_0x1f35dc(0x101)](_0x408eae);}}else{const _0x29fa14=_0x870a0d['tracing']?.[_0x1f35dc(0xc2)]?.[_0x1f35dc(0xc6)]()||process[_0x1f35dc(0xf5)][_0x1f35dc(0x115)]?.[_0x1f35dc(0xc6)]()||process[_0x1f35dc(0xf5)][_0x1f35dc(0xba)]?.[_0x1f35dc(0xc6)]()||_0x27c671[_0x1f35dc(0x135)];process[_0x1f35dc(0xf5)]['LANGCHAIN_TRACING_V2']=_0x27c671[_0x1f35dc(0xd5)],process[_0x1f35dc(0xf5)]['LANGSMITH_TRACING']=_0x27c671[_0x1f35dc(0xd5)],process['env'][_0x1f35dc(0xe8)]=_0x4e521a,process['env']['LANGSMITH_API_KEY']=_0x4e521a,process[_0x1f35dc(0xf5)][_0x1f35dc(0xba)]=_0x29fa14,process['env'][_0x1f35dc(0x115)]=_0x29fa14,console[_0x1f35dc(0xe1)](_0x27c671['PjBUa'],{'project':_0x29fa14});}}const _0x418274=_0x870a0d[_0x1f35dc(0x12a)]??_0x870a0d['intents'][0x0]?.[_0x1f35dc(0xc8)]??'greetings',_0xe8a64d=new Set(_0x870a0d[_0x1f35dc(0x131)]['map'](_0x58a94e=>_0x58a94e[_0x1f35dc(0xc8)])),_0x54f882=_0x870a0d['intentAliases']??{},_0x213fb9=_0x2b6594=>{const _0x39559a=_0x1f35dc,_0x24a50d=_0x2b6594[_0x39559a(0xc6)]()[_0x39559a(0x12f)]()[_0x39559a(0xec)](/[\s-]+/g,'_');if(_0xe8a64d[_0x39559a(0xc4)](_0x24a50d))return _0x24a50d;return _0x54f882[_0x24a50d]??_0x418274;},_0x2a3af8=_0x27c671['uxFQj'](stickySetFromIntents,_0x870a0d[_0x1f35dc(0x131)]),_0x2ad26c=_0x870a0d[_0x1f35dc(0xbe)]??{},_0x23c863=_0x870a0d[_0x1f35dc(0xd0)]??_0x27c671[_0x1f35dc(0xde)](llmConfigFromEnv),_0x316926=_0x27c671[_0x1f35dc(0xf9)](createModels,_0x23c863),_0x4260cf=_0x870a0d[_0x1f35dc(0xc1)]?.[_0x1f35dc(0x100)]??_0x316926[_0x1f35dc(0x100)],_0x202291=_0x870a0d['models']?.[_0x1f35dc(0xf4)]??_0x316926[_0x1f35dc(0xf4)],_0x51997c=_0x870a0d[_0x1f35dc(0xc1)]?.[_0x1f35dc(0xd6)]??_0x316926[_0x1f35dc(0xd6)];console[_0x1f35dc(0xe1)](_0x27c671[_0x1f35dc(0x12e)],{'provider':_0x316926[_0x1f35dc(0x106)][_0x1f35dc(0x109)],'chat':_0x316926[_0x1f35dc(0x106)][_0x1f35dc(0xdc)],'intentRouter':_0x316926['config'][_0x1f35dc(0x124)],'embeddingsProvider':_0x316926[_0x1f35dc(0x106)][_0x1f35dc(0x136)],'embeddings':_0x316926['config'][_0x1f35dc(0x12c)],'baseUrl':_0x23c863[_0x1f35dc(0x118)],'embeddingsBaseUrl':_0x23c863['embeddingsBaseUrl']});const _0x317fe7=_0x870a0d['toolOutputMap']?toolOutputMapToExtractors(_0x870a0d[_0x1f35dc(0x119)]):null,_0x2e2e82=_0x870a0d[_0x1f35dc(0xe2)]??_0x317fe7?.[_0x1f35dc(0xe2)],_0x2459b0=_0x870a0d[_0x1f35dc(0x103)]??_0x317fe7?.[_0x1f35dc(0x103)],_0x11a3ba=_0x870a0d[_0x1f35dc(0xe9)]??(_0x870a0d[_0x1f35dc(0x119)]?Object[_0x1f35dc(0x110)](_0x870a0d[_0x1f35dc(0x119)]):undefined),_0x1bbc8a=_0x870a0d['finalizeFlow']??(_0x11a3ba?.[_0x1f35dc(0x138)]?_0x27c671['OCMTj'](createCompleteOnFinalize,{'completeOn':_0x11a3ba,'toolOutputMap':_0x870a0d[_0x1f35dc(0x119)]}):undefined),_0x30e2de=_0x870a0d[_0x1f35dc(0xb5)]?createRedis(_0x870a0d['redisUrl']):undefined,_0x30f2b6=_0x27c671[_0x1f35dc(0xb6)](createChatCache,{'redis':_0x30e2de?.[_0x1f35dc(0xe7)],'config':{'enabled':_0x870a0d['cache']?.['enabled']??![],'ttlSeconds':_0x870a0d[_0x1f35dc(0xe5)]?.[_0x1f35dc(0xe0)],'flowTtlSeconds':_0x870a0d['cache']?.[_0x1f35dc(0xca)],'scope':_0x870a0d[_0x1f35dc(0xe5)]?.[_0x1f35dc(0x111)]}}),_0xe8ef51=createUsageStore({'redis':_0x30e2de?.[_0x1f35dc(0xe7)],'ttlSeconds':_0x870a0d['cache']?.[_0x1f35dc(0xe0)]??_0x27c671['MueHD'](_0x27c671['cMRTi'](_0x27c671['MueHD'](0x3c,0x3c),0x18),0x7)}),_0x366a80=await _0x27c671[_0x1f35dc(0x11c)](createAgentFactory,{'llm':_0x4260cf,'intentRegistry':_0x870a0d[_0x1f35dc(0x11f)],'redisUrl':_0x870a0d[_0x1f35dc(0xb5)],'checkpointer':_0x870a0d[_0x1f35dc(0xcc)]}),_0x600e73=_0x27c671[_0x1f35dc(0xd9)](createConversationService,_0x366a80),{runAgent:_0x265f80}=createAgentRunner({'agentFactory':_0x366a80,'recursionLimit':_0x870a0d[_0x1f35dc(0xc5)]}),_0x3b0e8b=createIntentEmbeddings({'embeddings':_0x51997c,'intents':_0x870a0d[_0x1f35dc(0x131)],'normalizeIntent':_0x213fb9}),_0x3911e2=_0x27c671[_0x1f35dc(0x11c)](createIntentRouter,{'model':_0x202291,'intents':_0x870a0d[_0x1f35dc(0x131)],'normalizeIntent':_0x213fb9,'defaultIntent':_0x418274,'routerSystemPrompt':_0x870a0d['routerSystemPrompt'],'llmConfidence':_0x870a0d[_0x1f35dc(0xb8)]?.[_0x1f35dc(0x11d)],'embeddingConfidence':_0x870a0d['thresholds']?.[_0x1f35dc(0xb7)]}),{resolveFlow:_0x48042f,finalizeAgentFlow:_0x3aeaee}=createFlowRouter({'stickyFlows':_0x2a3af8,'menuMap':_0x2ad26c,'normalizeIntent':_0x213fb9,'intents':_0x870a0d[_0x1f35dc(0x131)],'getEmbeddedIntents':_0x3b0e8b['getEmbeddedIntents'],'scoreIntentsForMessage':_0x3b0e8b[_0x1f35dc(0xc7)],'classifyIntentWithLlm':_0x3911e2[_0x1f35dc(0xfa)],'mergeIntentResults':_0x3911e2[_0x1f35dc(0x120)],'chatCache':_0x30f2b6,'finalizeFlow':_0x1bbc8a,'thresholds':_0x870a0d[_0x1f35dc(0xb8)]}),_0x2f8733=_0x870a0d[_0x1f35dc(0xff)]??(_0x2e2e82?createToolOutputExtractor(_0x2e2e82,_0x2459b0,{'includeUnknown':_0x870a0d['includeUnknownToolOutputs']??![]}):extractRawToolOutputs),_0x67fadb=_0x2890ff=>{const _0x1e682e=_0x1f35dc;if(_0x27c671[_0x1e682e(0x114)](_0x27c671['rPYWG'],_0x27c671[_0x1e682e(0xbd)]))return _0xa04e90['slice'](_0x599055);else{const _0x377b2f=_0x870a0d[_0x1e682e(0x128)];if(!_0x377b2f)return'';return _0x27c671[_0x1e682e(0xee)](typeof _0x377b2f,_0x1e682e(0xd4))?_0x27c671[_0x1e682e(0x105)](_0x377b2f,_0x2890ff):_0x377b2f;}},_0x21fd50=_0x23d23f=>{const _0x18a528=_0x1f35dc,_0x1d9b1e={'PCXYe':_0x27c671['nRfCR'],'cdvcu':_0x27c671[_0x18a528(0xd5)],'Fhucx':_0x18a528(0xef)};if(_0x27c671[_0x18a528(0x10a)](_0x27c671['QXagV'],_0x27c671[_0x18a528(0x108)])){const _0x25f7ca=_0x870a0d['intentRegistry'][_0x23d23f[_0x18a528(0xf3)]];if(!_0x25f7ca){if(_0x27c671['BwQvj'](_0x18a528(0x127),_0x18a528(0x137))){const _0xd2ac9b=_0x1ee463['sharedPrompt'];if(!_0xd2ac9b)return'';return typeof _0xd2ac9b===_0x27c671[_0x18a528(0xe6)]?_0x27c671[_0x18a528(0x105)](_0xd2ac9b,_0x1141a4):_0xd2ac9b;}else{const _0x5371d2=Object[_0x18a528(0xb4)](_0x870a0d['intentRegistry'])[0x0];return _0x5371d2?.[_0x18a528(0x101)](_0x23d23f)??'';}}return _0x25f7ca[_0x18a528(0x101)](_0x23d23f);}else{const _0x1dc4e5=_0x34b935['tracing']?.[_0x18a528(0xc2)]?.[_0x18a528(0xc6)]()||_0x2a642b['env'][_0x18a528(0x115)]?.['trim']()||_0x56f2f0[_0x18a528(0xf5)][_0x18a528(0xba)]?.[_0x18a528(0xc6)]()||_0x1d9b1e[_0x18a528(0xbb)];_0x4f61fd[_0x18a528(0xf5)][_0x18a528(0x11b)]=_0x1d9b1e[_0x18a528(0xb9)],_0x1e955b[_0x18a528(0xf5)][_0x18a528(0xea)]=_0x1d9b1e['cdvcu'],_0x56ee72['env'][_0x18a528(0xe8)]=_0x5b9e29,_0x35f8a3['env'][_0x18a528(0x132)]=_0xa5754,_0x25e463[_0x18a528(0xf5)]['LANGCHAIN_PROJECT']=_0x1dc4e5,_0x3f3fe8[_0x18a528(0xf5)][_0x18a528(0x115)]=_0x1dc4e5,_0x1f1fb1['log'](_0x1d9b1e[_0x18a528(0xd3)],{'project':_0x1dc4e5});}},_0x19fd33=_0x870a0d[_0x1f35dc(0xd8)]??(_0xd29617=>{const _0x256d8b=_0x1f35dc,_0x3576d5=_0x27c671[_0x256d8b(0x11c)](_0x67fadb,_0xd29617)[_0x256d8b(0xc6)](),_0xfcc4e=_0x27c671['uxFQj'](_0x21fd50,_0xd29617)[_0x256d8b(0xc6)]();if(_0x3576d5&&_0xfcc4e)return _0x3576d5+'\x0a\x0a'+_0xfcc4e;return _0x3576d5||_0xfcc4e;}),_0x5572e4=_0x870a0d['resolvePhoneKey']??(({threadId:_0x27a257,phoneKey:_0x124f33})=>_0x124f33??_0x27a257),_0x3d61e6=async _0x296a96=>{const _0x82b791=_0x1f35dc,_0x2fd116={'Gxuoi':function(_0x396c78,_0x45c789){return _0x27c671['GVCuO'](_0x396c78,_0x45c789);},'RhptV':function(_0x19e8a4,_0x3cfb1b){const _0x1297b6=a1_0x26c0;return _0x27c671[_0x1297b6(0x10a)](_0x19e8a4,_0x3cfb1b);}};if(_0x27c671['jwEZy']===_0x27c671['jwEZy']){const _0x9b4ed3=_0x366a80[_0x82b791(0xcc)];if(_0x27c671['PCZrp'](typeof _0x9b4ed3[_0x82b791(0xed)],_0x27c671[_0x82b791(0xe6)])){if(_0x82b791(0x11a)===_0x27c671[_0x82b791(0xd2)])await _0x9b4ed3['deleteThread'](_0x296a96);else{const _0x474fbf=_0x2fd116[_0x82b791(0x10d)](_0x402ccf,_0xc3c471[_0x2dc48b]?.[_0x82b791(0xd7)]??'')['trim']();if(_0x2fd116[_0x82b791(0x130)](_0x474fbf,_0x4d6e2d))return _0x24a118[_0x82b791(0x129)](_0x505274);}}await _0x30f2b6['clearChatCacheFlowState'](_0x296a96),await _0xe8ef51['clearThreadUsage'](_0x296a96);}else{const _0x40aaa3=_0x27c671[_0x82b791(0x11c)](_0x392684,_0x37c1b2)[_0x82b791(0xc6)](),_0x454f02=_0x4a3ac5(_0xa1535c)[_0x82b791(0xc6)]();if(_0x40aaa3&&_0x454f02)return _0x40aaa3+'\x0a\x0a'+_0x454f02;return _0x27c671[_0x82b791(0xe4)](_0x40aaa3,_0x454f02);}},_0xefbb6b=async _0x29f054=>{const _0x4a7c21=_0x1f35dc;_0x27c671['GVCuO'](assertInvokeInput,_0x29f054);if(_0x30f2b6[_0x4a7c21(0xf2)]){const _0x2026b6=await _0x30f2b6[_0x4a7c21(0xcb)](_0x29f054[_0x4a7c21(0xbf)]);if(_0x2026b6){if(_0x27c671[_0x4a7c21(0x10e)](_0x27c671['vUhaU'],'utjSj')){const _0x2f842c=await _0xe8ef51['getUsageReport'](_0x29f054[_0x4a7c21(0x133)]);return{'messages':[],'answer':_0x2026b6,'intent':{'intent':_0x418274,'score':0x1},'flow':{'active':![],'intent':null},'toolOutputs':{},'flowCompleted':![],'extras':{'cacheHit':!![]},'usage':_0x2f842c};}else _0x5a8939['warn'](_0x27c671[_0x4a7c21(0x121)]);}}const _0x32c514=await _0x30f2b6['getChatCacheFlowState'](_0x29f054[_0x4a7c21(0x133)]),[_0x4449ae,_0x1fed1d]=await Promise[_0x4a7c21(0x10f)]([_0x600e73[_0x4a7c21(0xb3)](_0x29f054['threadId']),_0x600e73[_0x4a7c21(0xc0)](_0x29f054[_0x4a7c21(0x133)])]),{messageIntent:_0x133fcc,flow:_0x33595b,promptIntent:_0x121dae,llmResult:_0x19ed74}=await _0x48042f({'message':_0x29f054[_0x4a7c21(0xbf)],'explicitIntent':_0x29f054['explicitIntent'],'cachedFlowState':_0x32c514,'conversationContext':_0x1fed1d}),_0xb646b6=_0x27c671['uxFQj'](_0x5572e4,{'threadId':_0x29f054[_0x4a7c21(0x133)],'phoneKey':_0x29f054[_0x4a7c21(0x104)],'context':_0x29f054[_0x4a7c21(0xf0)]}),_0x290034=_0x27c671['bKhuW'](_0x19fd33,{'hasConversationHistory':_0x27c671[_0x4a7c21(0xfe)](_0x4449ae[_0x4a7c21(0x138)],0x0),'intent':_0x121dae,'intentScore':_0x33595b[_0x4a7c21(0x113)]?0x1:_0x133fcc[_0x4a7c21(0x122)],'message':_0x29f054[_0x4a7c21(0xbf)],'context':_0x29f054['context']}),_0x538c35=await _0x27c671[_0x4a7c21(0xdd)](_0x265f80,{'message':_0x29f054[_0x4a7c21(0xbf)],'threadId':_0x29f054[_0x4a7c21(0x133)],'messages':_0x4449ae,'systemPrompt':_0x290034,'phoneKey':_0xb646b6,'intent':_0x121dae}),_0x3f14cc=_0x538c35[_0x4a7c21(0xce)],_0x144220=_0x27c671['CDPBe'](getTurnMessages,_0x3f14cc,_0x29f054[_0x4a7c21(0xbf)]),_0x3a6f63=_0x27c671[_0x4a7c21(0xf8)](_0x2f8733,_0x144220),_0x4a0ccc=await _0x3aeaee({'threadId':_0x29f054[_0x4a7c21(0x133)],'promptIntent':_0x121dae,'flow':_0x33595b,'toolOutputs':_0x3a6f63}),_0x59cdcb=getLastMessageContent(_0x3f14cc);if(_0x30f2b6[_0x4a7c21(0xf2)]&&_0x59cdcb){if(_0x27c671[_0x4a7c21(0x10b)]!==_0x4a7c21(0x123))await _0x30f2b6[_0x4a7c21(0xf7)](_0x29f054[_0x4a7c21(0xbf)],_0x59cdcb);else{if(!_0xd43cbd)_0x469bd4[_0x4a7c21(0xeb)](_0x27c671[_0x4a7c21(0x121)]);else{const _0x592a31=_0x108a96[_0x4a7c21(0xcf)]?.[_0x4a7c21(0xc2)]?.[_0x4a7c21(0xc6)]()||_0x448cdd['env'][_0x4a7c21(0x115)]?.[_0x4a7c21(0xc6)]()||_0x4ddd57[_0x4a7c21(0xf5)][_0x4a7c21(0xba)]?.[_0x4a7c21(0xc6)]()||_0x4a7c21(0xcd);_0x2592d5[_0x4a7c21(0xf5)][_0x4a7c21(0x11b)]=_0x27c671[_0x4a7c21(0xd5)],_0x2eaeb5['env'][_0x4a7c21(0xea)]=_0x27c671[_0x4a7c21(0xd5)],_0x1ea219[_0x4a7c21(0xf5)][_0x4a7c21(0xe8)]=_0x1e7498,_0x5cc36c[_0x4a7c21(0xf5)]['LANGSMITH_API_KEY']=_0x3afa23,_0x2f46dc['env']['LANGCHAIN_PROJECT']=_0x592a31,_0x20e6af[_0x4a7c21(0xf5)][_0x4a7c21(0x115)]=_0x592a31,_0x3273c1['log'](_0x4a7c21(0xef),{'project':_0x592a31});}}}const _0x523642=_0x27c671[_0x4a7c21(0xf1)](extractTokenUsageFromMessages,_0x144220),_0x35c8e5=_0x133fcc['intent']||_0x121dae||null,_0x33b048=_0x4a0ccc['flow'][_0x4a7c21(0x113)]&&_0x4a0ccc[_0x4a7c21(0xc3)][_0x4a7c21(0xf3)]?_0x4a0ccc['flow'][_0x4a7c21(0xf3)]:_0x2a3af8[_0x4a7c21(0xc4)](_0x121dae)?_0x121dae:null,{total:_0x20fa49,flow:_0x33201c,intents:_0x487877}=await _0xe8ef51['recordTurn']({'threadId':_0x29f054[_0x4a7c21(0x133)],'intent':_0x35c8e5,'flowIntent':_0x33b048,'turn':_0x523642});if(_0x4a0ccc[_0x4a7c21(0xfd)]&&_0x33b048){if(_0x27c671['ivcyd']===_0x27c671[_0x4a7c21(0x117)]){const _0x520e21=_0x373a34[_0x4a7c21(0xb4)](_0x3bf335['intentRegistry'])[0x0];return _0x520e21?.[_0x4a7c21(0x101)](_0x568617)??'';}else await _0xe8ef51['clearFlowUsage'](_0x29f054[_0x4a7c21(0x133)],_0x33b048);}const _0x42feee={'turn':_0x523642,'flow':{..._0x33201c,'intent':_0x33b048},'total':_0x20fa49,'intents':_0x487877};return{'messages':_0x3f14cc,'answer':_0x59cdcb,'intent':_0x133fcc,'flow':_0x4a0ccc[_0x4a7c21(0xc3)],'toolOutputs':_0x3a6f63,'flowCompleted':_0x4a0ccc[_0x4a7c21(0xfd)],'extras':_0x4a0ccc[_0x4a7c21(0xdf)],'usage':_0x42feee,..._0x19ed74?{'router':{'action':_0x19ed74['action'],'reason':_0x19ed74[_0x4a7c21(0x112)]}}:{}};};return{'invoke':_0xefbb6b,'connect':async()=>{await _0x30e2de?.['connect']();},'disconnect':async()=>{await _0x30e2de?.['disconnect']();},'clearFlow':_0x30f2b6['clearChatCacheFlowState'],'clearThread':_0x3d61e6,'clearSession':_0x3d61e6,'getUsage':_0x56ba1d=>_0xe8ef51[_0x1f35dc(0x10c)](_0x56ba1d),'getCachedAnswer':_0x30f2b6[_0x1f35dc(0xcb)],'setCachedAnswer':_0x30f2b6[_0x1f35dc(0xf7)],'invalidateIntentEmbeddings':_0x3b0e8b['invalidate']};}

package/dist/embeddings/xenova.d.ts

@@ -16,4 +16,3 @@ export declare class XenovaEmbeddings extends Embeddings {
     embedDocuments(documents: string[]): Promise<number[][]>;
     embedQuery(document: string): Promise<number[]>;
 }
-//# sourceMappingURL=xenova.d.ts.map

package/dist/embeddings/xenova.js

@@ -1,69 +1 @@
-import { Embeddings } from "@langchain/core/embeddings";
-import { createRequire } from "node:module";
-import path from "node:path";
-import { pathToFileURL } from "node:url";
-async function loadXenovaTransformers() {
-    const errors = [];
-    try {
-        return (await import("@xenova/transformers"));
-    }
-    catch (err) {
-        errors.push(`package import: ${err instanceof Error ? err.message : String(err)}`);
-    }
-    // pnpm nests this package; resolve from the app's node_modules / cwd.
-    try {
-        const requireFromCwd = createRequire(path.join(process.cwd(), "package.json"));
-        const resolved = requireFromCwd.resolve("@xenova/transformers");
-        return (await import(pathToFileURL(resolved).href));
-    }
-    catch (err) {
-        errors.push(`cwd resolve: ${err instanceof Error ? err.message : String(err)}`);
-    }
-    throw new Error([
-        'embeddingsProvider "xenova" could not load @xenova/transformers.',
-        "Install in your app: pnpm add @xenova/transformers",
-        "If you see a sharp error, rebuild native deps: pnpm rebuild sharp --config.ignore-scripts=false",
-        ...errors.map((e) => `  - ${e.split("\n")[0]}`),
-    ].join("\n"));
-}
-/**
- * Free, in-process embeddings via `@xenova/transformers` (no Ollama / OpenAI).
- * Install: `pnpm add @xenova/transformers`
- */
-export class XenovaEmbeddings extends Embeddings {
-    model;
-    pipelinePromise = null;
-    constructor(fields) {
-        super(fields ?? {});
-        this.model = fields?.model ?? "Xenova/all-MiniLM-L6-v2";
-    }
-    async getPipeline() {
-        if (!this.pipelinePromise) {
-            this.pipelinePromise = (async () => {
-                const mod = await loadXenovaTransformers();
-                return mod.pipeline("feature-extraction", this.model);
-            })();
-        }
-        return this.pipelinePromise;
-    }
-    toVector(result) {
-        if (typeof result.tolist === "function") {
-            const list = result.tolist();
-            return Array.isArray(list[0]) ? list[0] : list;
-        }
-        return Array.from(result.data);
-    }
-    async embedDocuments(documents) {
-        const pipe = await this.getPipeline();
-        const vectors = [];
-        for (const doc of documents) {
-            const result = await pipe(doc, { pooling: "mean", normalize: true });
-            vectors.push(this.toVector(result));
-        }
-        return vectors;
-    }
-    async embedQuery(document) {
-        const [vector] = await this.embedDocuments([document]);
-        return vector ?? [];
-    }
-}
+const a2_0x418009=a2_0x1bd0;(function(_0xef6c1e,_0x58eba6){const _0x49f50e=a2_0x1bd0,_0x4f8f04=_0xef6c1e();while(!![]){try{const _0x15eae3=parseInt(_0x49f50e(0x17b))/0x1+-parseInt(_0x49f50e(0x167))/0x2+parseInt(_0x49f50e(0x166))/0x3+parseInt(_0x49f50e(0x177))/0x4*(parseInt(_0x49f50e(0x165))/0x5)+parseInt(_0x49f50e(0x16f))/0x6*(parseInt(_0x49f50e(0x181))/0x7)+-parseInt(_0x49f50e(0x15d))/0x8+-parseInt(_0x49f50e(0x16b))/0x9*(-parseInt(_0x49f50e(0x17c))/0xa);if(_0x15eae3===_0x58eba6)break;else _0x4f8f04['push'](_0x4f8f04['shift']());}catch(_0x50c3be){_0x4f8f04['push'](_0x4f8f04['shift']());}}}(a2_0x556a,0x3cc49));function a2_0x556a(){const _0x26ba90=['UgmyZ','mean','pasgn','HFWMn','onLnK','resolve','cwd\x20resolve:\x20','175OTQSvV','122283KvQmIY','945954tZudxN','package.json','map','push','1017zhReIA','\x20\x20-\x20','VSGBt','model','214836uIJLKh','pgwUy','feature-extraction','RsZRK','message','WNHLM','embedDocuments','embedQuery','37624DzsbbC','UqqHr','pipelinePromise','cwd','395121kuuECz','26830hUlRja','toVector','from','getPipeline','pipeline','7UGfRCB','@xenova/transformers','embeddingsProvider\x20\x22xenova\x22\x20could\x20not\x20load\x20@xenova/transformers.','LnHSo','data','join','tolist','3057560QoDMlJ'];a2_0x556a=function(){return _0x26ba90;};return a2_0x556a();}function a2_0x1bd0(_0x436b2c,_0x469d56){_0x436b2c=_0x436b2c-0x15a;const _0x556af0=a2_0x556a();let _0x1bd087=_0x556af0[_0x436b2c];return _0x1bd087;}import{Embeddings}from'@langchain/core/embeddings';import{createRequire}from'node:module';import a2_0x45d84b from'node:path';import{pathToFileURL}from'node:url';async function loadXenovaTransformers(){const _0x4ada97=a2_0x1bd0,_0x3c7e30={'UqqHr':function(_0xb981c4,_0x4be3b6){return _0xb981c4(_0x4be3b6);},'xJsFb':function(_0x39f74b,_0x2556e0){return _0x39f74b(_0x2556e0);},'pgwUy':_0x4ada97(0x168),'VSGBt':function(_0x4c0af3,_0x398cf2){return _0x4c0af3 instanceof _0x398cf2;},'EdEsv':function(_0x4242fb,_0x302326){return _0x4242fb(_0x302326);},'onLnK':_0x4ada97(0x183),'zNARI':'Install\x20in\x20your\x20app:\x20pnpm\x20add\x20@xenova/transformers','LnHSo':'If\x20you\x20see\x20a\x20sharp\x20error,\x20rebuild\x20native\x20deps:\x20pnpm\x20rebuild\x20sharp\x20--config.ignore-scripts=false'},_0x3d1820=[];try{return await import('@xenova/transformers');}catch(_0x3bf9de){_0x3d1820[_0x4ada97(0x16a)]('package\x20import:\x20'+(_0x3bf9de instanceof Error?_0x3bf9de[_0x4ada97(0x173)]:_0x3c7e30[_0x4ada97(0x178)](String,_0x3bf9de)));}try{const _0x4b52c3=_0x3c7e30['xJsFb'](createRequire,a2_0x45d84b['join'](process[_0x4ada97(0x17a)](),_0x3c7e30[_0x4ada97(0x170)])),_0x2b3984=_0x4b52c3[_0x4ada97(0x163)](_0x4ada97(0x182));return await import(pathToFileURL(_0x2b3984).href);}catch(_0x1880ac){_0x3d1820[_0x4ada97(0x16a)](_0x4ada97(0x164)+(_0x3c7e30[_0x4ada97(0x16d)](_0x1880ac,Error)?_0x1880ac[_0x4ada97(0x173)]:_0x3c7e30['EdEsv'](String,_0x1880ac)));}throw new Error([_0x3c7e30[_0x4ada97(0x162)],_0x3c7e30['zNARI'],_0x3c7e30[_0x4ada97(0x184)],..._0x3d1820[_0x4ada97(0x169)](_0x9da1f=>_0x4ada97(0x16c)+_0x9da1f['split']('\x0a')[0x0])][_0x4ada97(0x15b)]('\x0a'));}export class XenovaEmbeddings extends Embeddings{['model'];['pipelinePromise']=null;constructor(_0x1f14e7){const _0x3c88b5=a2_0x1bd0,_0x627dff={'IScun':function(_0x341864,_0xa81492){return _0x341864??_0xa81492;},'HFWMn':'Xenova/all-MiniLM-L6-v2'};super(_0x627dff['IScun'](_0x1f14e7,{})),this[_0x3c88b5(0x16e)]=_0x1f14e7?.[_0x3c88b5(0x16e)]??_0x627dff[_0x3c88b5(0x161)];}async[a2_0x418009(0x17f)](){const _0x33a176=a2_0x418009,_0x4edc4b={'pasgn':function(_0x5e5c45){return _0x5e5c45();}};return!this[_0x33a176(0x179)]&&(this[_0x33a176(0x179)]=((async()=>{const _0x10aa1f=_0x33a176,_0x39c720=await _0x4edc4b[_0x10aa1f(0x160)](loadXenovaTransformers);return _0x39c720[_0x10aa1f(0x180)](_0x10aa1f(0x171),this[_0x10aa1f(0x16e)]);})())),this[_0x33a176(0x179)];}[a2_0x418009(0x17d)](_0x24d61a){const _0x13861b=a2_0x418009,_0x242207={'EFJtL':function(_0x29c81a,_0x163bcb){return _0x29c81a===_0x163bcb;},'UgmyZ':'function'};if(_0x242207['EFJtL'](typeof _0x24d61a['tolist'],_0x242207[_0x13861b(0x15e)])){const _0x48f102=_0x24d61a[_0x13861b(0x15c)]();return Array['isArray'](_0x48f102[0x0])?_0x48f102[0x0]:_0x48f102;}return Array[_0x13861b(0x17e)](_0x24d61a[_0x13861b(0x15a)]);}async[a2_0x418009(0x175)](_0x5b80fc){const _0x827e1f=a2_0x418009,_0x4de9eb={'RsZRK':function(_0x107c71,_0x29ed2a,_0x3c6adb){return _0x107c71(_0x29ed2a,_0x3c6adb);},'WNHLM':_0x827e1f(0x15f)},_0x93075d=await this['getPipeline'](),_0x311ee9=[];for(const _0x3f939c of _0x5b80fc){const _0x443954=await _0x4de9eb[_0x827e1f(0x172)](_0x93075d,_0x3f939c,{'pooling':_0x4de9eb[_0x827e1f(0x174)],'normalize':!![]});_0x311ee9[_0x827e1f(0x16a)](this[_0x827e1f(0x17d)](_0x443954));}return _0x311ee9;}async[a2_0x418009(0x176)](_0xabf3d){const [_0x3c2aef]=await this['embedDocuments']([_0xabf3d]);return _0x3c2aef??[];}}

package/dist/env-llm.d.ts

@@ -1,11 +1,14 @@
 import type { EmbeddingsProvider, LlmConfig, LlmProvider } from "./types.js";
+/** Vercel / Lambda / Netlify — @xenova/transformers + sharp do not work here. */
+export declare function isServerlessRuntime(env?: NodeJS.ProcessEnv): boolean;
 /**
  * Resolve embeddings backend.
  *
- * - Explicit EMBEDDINGS_PROVIDER wins (ollama | openai | xenova).
- * - openrouter is not an embeddings backend → xenova (free).
- * - If unset: ollama chat → ollama embeddings; otherwise → xenova
- *   (so OpenRouter/OpenAI chat works in production without local Ollama).
+ * - Explicit EMBEDDINGS_PROVIDER wins (ollama | openai | xenova), except:
+ *   on serverless (Vercel/Lambda/…) with OPENAI_API_KEY, xenova → openai
+ *   (Transformers.js / sharp cannot load in those runtimes).
+ * - openrouter is not an embeddings backend → xenova locally, openai on serverless+key.
+ * - If unset: ollama chat → ollama; serverless+OPENAI_API_KEY → openai; else → xenova.
  */
 export declare function resolveEmbeddingsProvider(env: NodeJS.ProcessEnv, chatProvider: LlmProvider): EmbeddingsProvider;
 /**
@@ -14,10 +17,9 @@ export declare function resolveEmbeddingsProvider(env: NodeJS.ProcessEnv, chatPr
  * Env knobs:
  * - LLM_PROVIDER — ollama | openai | openrouter
  * - EMBEDDINGS_PROVIDER — ollama | openai | xenova
- *   (default: ollama if LLM is ollama, else xenova)
+ *   (default: ollama if LLM is ollama; openai on Vercel/Lambda when OPENAI_API_KEY set; else xenova)
  * - OLLAMA_MODEL / OPENAI_MODEL / OPENROUTER_MODEL
  * - XENOVA_EMBEDDINGS_MODEL / OLLAMA_EMBEDDINGS_MODEL / OPENAI_EMBEDDINGS_MODEL
  * - OLLAMA_BASE_URL / OPENAI_API_KEY / OPENROUTER_API_KEY / OPENROUTER_BASE_URL
  */
 export declare function llmConfigFromEnv(env?: NodeJS.ProcessEnv): LlmConfig;
-//# sourceMappingURL=env-llm.d.ts.map