🔗 feat: Agent Chain (Mixture-of-Agents) (#6374)

* wip: first pass, dropdown for selecting sequential agents

* refactor: Improve agent selection logic and enhance performance in SequentialAgents component

* wip: seq. agents working ideas

* wip: sequential agents style change

* refactor: move agent form options/submission outside of AgentConfig

* refactor: prevent repeating code

* refactor: simplify current agent display in SequentialAgents component

* feat: persist  form value handling in AgentSelect component for agent_ids

* feat: first pass, sequential agnets agent update

* feat: enhance message display with agent updates and empty text handling

* chore: update Icon component to use EModelEndpoint for agent endpoints

* feat: update content type checks in BaseClient to use constants for better readability

* feat: adjust max context tokens calculation to use 90% of the model's max tokens

* feat: first pass, agent run message pruning

* chore: increase max listeners for abort controller to prevent memory leaks

* feat: enhance runAgent function to include current index count map for improved token tracking

* chore: update @librechat/agents dependency to version 2.2.5

* feat: update icons and style of SequentialAgents component for improved UI consistency

* feat: add AdvancedButton and AdvancedPanel components for enhanced agent settings navigation, update styling for agent form

* chore: adjust minimum height of AdvancedPanel component for better layout consistency

* chore: update @librechat/agents dependency to version 2.2.6

* feat: enhance message formatting by incorporating tool set into agent message processing, in order to allow better mix/matching of agents (as tool calls for tools not found in set will be stringified)

* refactor: reorder components in AgentConfig for improved readability and maintainability

* refactor: enhance layout of AgentUpdate component for improved visual structure

* feat: add DeepSeek provider to Bedrock settings and schemas

* feat: enhance link styling in mobile.css for better visibility and accessibility

* fix: update banner model import in update banner script; export Banner model

* refactor: `duplicateAgentHandler` to include tool_resources only for OCR context files

* feat: add 'qwen-vl' to visionModels for enhanced model support

* fix: change image format from JPEG to PNG in DALLE3 response

* feat: reorganize Advanced components and add localizations

* refactor: simplify JSX structure in AgentChain component to defer container styling to parent

* feat: add FormInput component for reusable input handling

* feat: make agent recursion limit configurable from builder

* feat: add support for agent capabilities chain in AdvancedPanel and update data-provider version

* feat: add maxRecursionLimit configuration for agents and update related documentation

* fix: update CONFIG_VERSION to 1.2.3 in data provider configuration

* feat: replace recursion limit input with MaxAgentSteps component and enhance input handling

* feat: enhance AgentChain component with hover card for additional information and update related labels

* fix: pass request and response objects to `createActionTool` when using assistant actions to prevent auth error

* feat: update AgentChain component layout to include agent count display

* feat: increase default max listeners and implement capability check function for agent chain

* fix: update link styles in mobile.css for better visibility in dark mode

* chore: temp. remove agents package while bumping shared packages

* chore: update @langchain/google-genai package to version 0.1.11

* chore: update @langchain/google-vertexai package to version 0.2.2

* chore: add @librechat/agents package at version 2.2.8

* feat: add deepseek.r1 model with token rate and context values for bedrock

api/server/controllers/agents/client.js

@@ -7,7 +7,16 @@
 // validateVisionModel,
 // mapModelToAzureConfig,
 // } = require('librechat-data-provider');
-const { Callback, createMetadataAggregator } = require('@librechat/agents');
+require('events').EventEmitter.defaultMaxListeners = 100;
+const {
+  Callback,
+  GraphEvents,
+  formatMessage,
+  formatAgentMessages,
+  formatContentStrings,
+  getTokenCountForMessage,
+  createMetadataAggregator,
+} = require('@librechat/agents');
 const {
   Constants,
   VisionModes,
@@ -17,24 +26,19 @@ const {
   KnownEndpoints,
   anthropicSchema,
   isAgentsEndpoint,
+  AgentCapabilities,
   bedrockInputSchema,
   removeNullishValues,
 } = require('librechat-data-provider');
-const {
-  formatMessage,
-  addCacheControl,
-  formatAgentMessages,
-  formatContentStrings,
-  createContextHandlers,
-} = require('~/app/clients/prompts');
+const { getCustomEndpointConfig, checkCapability } = require('~/server/services/Config');
+const { addCacheControl, createContextHandlers } = require('~/app/clients/prompts');
 const { spendTokens, spendStructuredTokens } = require('~/models/spendTokens');
 const { getBufferString, HumanMessage } = require('@langchain/core/messages');
 const { encodeAndFormat } = require('~/server/services/Files/images/encode');
-const { getCustomEndpointConfig } = require('~/server/services/Config');
 const Tokenizer = require('~/server/services/Tokenizer');
 const BaseClient = require('~/app/clients/BaseClient');
+const { logger, sendEvent } = require('~/config');
 const { createRun } = require('./run');
-const { logger } = require('~/config');
 
 /** @typedef {import('@librechat/agents').MessageContentComplex} MessageContentComplex */
 /** @typedef {import('@langchain/core/runnables').RunnableConfig} RunnableConfig */
@@ -99,6 +103,8 @@ class AgentClient extends BaseClient {
     this.outputTokensKey = 'output_tokens';
     /** @type {UsageMetadata} */
     this.usage;
+    /** @type {Record<string, number>} */
+    this.indexTokenCountMap = {};
   }
 
   /**
@@ -377,6 +383,10 @@ class AgentClient extends BaseClient {
       }));
     }
 
+    for (let i = 0; i < messages.length; i++) {
+      this.indexTokenCountMap[i] = messages[i].tokenCount;
+    }
+
     const result = {
       tokenCountMap,
       prompt: payload,
@@ -622,6 +632,9 @@ class AgentClient extends BaseClient {
       //   });
       // }
 
+      /** @type {TCustomConfig['endpoints']['agents']} */
+      const agentsEConfig = this.options.req.app.locals[EModelEndpoint.agents];
+
       /** @type {Partial<RunnableConfig> & { version: 'v1' | 'v2'; run_id?: string; streamMode: string }} */
       const config = {
         configurable: {
@@ -629,19 +642,30 @@ class AgentClient extends BaseClient {
           last_agent_index: this.agentConfigs?.size ?? 0,
           hide_sequential_outputs: this.options.agent.hide_sequential_outputs,
         },
-        recursionLimit: this.options.req.app.locals[EModelEndpoint.agents]?.recursionLimit,
+        recursionLimit: agentsEConfig?.recursionLimit,
         signal: abortController.signal,
         streamMode: 'values',
         version: 'v2',
       };
 
-      const initialMessages = formatAgentMessages(payload);
+      const toolSet = new Set((this.options.agent.tools ?? []).map((tool) => tool && tool.name));
+      let { messages: initialMessages, indexTokenCountMap } = formatAgentMessages(
+        payload,
+        this.indexTokenCountMap,
+        toolSet,
+      );
       if (legacyContentEndpoints.has(this.options.agent.endpoint)) {
-        formatContentStrings(initialMessages);
+        initialMessages = formatContentStrings(initialMessages);
       }
 
       /** @type {ReturnType<createRun>} */
       let run;
+      const countTokens = ((text) => this.getTokenCount(text)).bind(this);
+
+      /** @type {(message: BaseMessage) => number} */
+      const tokenCounter = (message) => {
+        return getTokenCountForMessage(message, countTokens);
+      };
 
       /**
        *
@@ -649,12 +673,23 @@ class AgentClient extends BaseClient {
        * @param {BaseMessage[]} messages
        * @param {number} [i]
        * @param {TMessageContentParts[]} [contentData]
+       * @param {Record<string, number>} [currentIndexCountMap]
        */
-      const runAgent = async (agent, _messages, i = 0, contentData = []) => {
+      const runAgent = async (agent, _messages, i = 0, contentData = [], _currentIndexCountMap) => {
         config.configurable.model = agent.model_parameters.model;
+        const currentIndexCountMap = _currentIndexCountMap ?? indexTokenCountMap;
         if (i > 0) {
           this.model = agent.model_parameters.model;
         }
+        if (agent.recursion_limit && typeof agent.recursion_limit === 'number') {
+          config.recursionLimit = agent.recursion_limit;
+        }
+        if (
+          agentsEConfig?.maxRecursionLimit &&
+          config.recursionLimit > agentsEConfig?.maxRecursionLimit
+        ) {
+          config.recursionLimit = agentsEConfig?.maxRecursionLimit;
+        }
         config.configurable.agent_id = agent.id;
         config.configurable.name = agent.name;
         config.configurable.agent_index = i;
@@ -717,11 +752,29 @@ class AgentClient extends BaseClient {
         }
 
         if (contentData.length) {
+          const agentUpdate = {
+            type: ContentTypes.AGENT_UPDATE,
+            [ContentTypes.AGENT_UPDATE]: {
+              index: contentData.length,
+              runId: this.responseMessageId,
+              agentId: agent.id,
+            },
+          };
+          const streamData = {
+            event: GraphEvents.ON_AGENT_UPDATE,
+            data: agentUpdate,
+          };
+          this.options.aggregateContent(streamData);
+          sendEvent(this.options.res, streamData);
+          contentData.push(agentUpdate);
           run.Graph.contentData = contentData;
         }
 
         await run.processStream({ messages }, config, {
           keepContent: i !== 0,
+          tokenCounter,
+          indexTokenCountMap: currentIndexCountMap,
+          maxContextTokens: agent.maxContextTokens,
           callbacks: {
             [Callback.TOOL_ERROR]: (graph, error, toolId) => {
               logger.error(
@@ -735,9 +788,13 @@ class AgentClient extends BaseClient {
       };
 
       await runAgent(this.options.agent, initialMessages);
-
       let finalContentStart = 0;
-      if (this.agentConfigs && this.agentConfigs.size > 0) {
+      if (
+        this.agentConfigs &&
+        this.agentConfigs.size > 0 &&
+        (await checkCapability(this.options.req, AgentCapabilities.chain))
+      ) {
+        const windowSize = 5;
         let latestMessage = initialMessages.pop().content;
         if (typeof latestMessage !== 'string') {
           latestMessage = latestMessage[0].text;
@@ -745,7 +802,16 @@ class AgentClient extends BaseClient {
         let i = 1;
         let runMessages = [];
 
-        const lastFiveMessages = initialMessages.slice(-5);
+        const windowIndexCountMap = {};
+        const windowMessages = initialMessages.slice(-windowSize);
+        let currentIndex = 4;
+        for (let i = initialMessages.length - 1; i >= 0; i--) {
+          windowIndexCountMap[currentIndex] = indexTokenCountMap[i];
+          currentIndex--;
+          if (currentIndex < 0) {
+            break;
+          }
+        }
         for (const [agentId, agent] of this.agentConfigs) {
           if (abortController.signal.aborted === true) {
             break;
@@ -780,7 +846,9 @@ class AgentClient extends BaseClient {
           }
           try {
             const contextMessages = [];
-            for (const message of lastFiveMessages) {
+            const runIndexCountMap = {};
+            for (let i = 0; i < windowMessages.length; i++) {
+              const message = windowMessages[i];
               const messageType = message._getType();
               if (
                 (!agent.tools || agent.tools.length === 0) &&
@@ -788,11 +856,13 @@ class AgentClient extends BaseClient {
               ) {
                 continue;
               }
-
+              runIndexCountMap[contextMessages.length] = windowIndexCountMap[i];
               contextMessages.push(message);
             }
-            const currentMessages = [...contextMessages, new HumanMessage(bufferString)];
-            await runAgent(agent, currentMessages, i, contentData);
+            const bufferMessage = new HumanMessage(bufferString);
+            runIndexCountMap[contextMessages.length] = tokenCounter(bufferMessage);
+            const currentMessages = [...contextMessages, bufferMessage];
+            await runAgent(agent, currentMessages, i, contentData, runIndexCountMap);
           } catch (err) {
             logger.error(
               `[api/server/controllers/agents/client.js #chatCompletion] Error running agent ${agentId} (${i})`,
@@ -803,6 +873,7 @@ class AgentClient extends BaseClient {
         }
       }
 
+      /** Note: not implemented */
       if (config.configurable.hide_sequential_outputs !== true) {
         finalContentStart = 0;
       }

api/server/controllers/agents/v1.js

@@ -1,10 +1,11 @@
 const fs = require('fs').promises;
 const { nanoid } = require('nanoid');
 const {
-  FileContext,
-  Constants,
   Tools,
+  Constants,
+  FileContext,
   SystemRoles,
+  EToolResources,
   actionDelimiter,
 } = require('librechat-data-provider');
 const {
@@ -203,14 +204,21 @@ const duplicateAgentHandler = async (req, res) => {
     }
 
     const {
-      _id: __id,
       id: _id,
+      _id: __id,
       author: _author,
       createdAt: _createdAt,
       updatedAt: _updatedAt,
+      tool_resources: _tool_resources = {},
       ...cloneData
     } = agent;
 
+    if (_tool_resources?.[EToolResources.ocr]) {
+      cloneData.tool_resources = {
+        [EToolResources.ocr]: _tool_resources[EToolResources.ocr],
+      };
+    }
+
     const newAgentId = `agent_${nanoid()}`;
     const newAgentData = Object.assign(cloneData, {
       id: newAgentId,