feat(experimental): FunctionsAgent, uses new function payload for tooling

api/app/langchain/ChatAgent.js

@@ -6,11 +6,11 @@ const {
 } = require('@dqbd/tiktoken');
 const { fetchEventSource } = require('@waylaidwanderer/fetch-event-source');
 const { Agent, ProxyAgent } = require('undici');
-const TextStream = require('../stream');
+// const TextStream = require('../stream');
 const { ChatOpenAI } = require('langchain/chat_models/openai');
 const { CallbackManager } = require('langchain/callbacks');
 const { HumanChatMessage, AIChatMessage } = require('langchain/schema');
-const { initializeCustomAgent } = require('./agents/CustomAgent/initializeCustomAgent');
+const { initializeCustomAgent, initializeFunctionsAgent } = require('./agents/');
 const { getMessages, saveMessage, saveConvo } = require('../../models');
 const { loadTools, SelfReflectionTool } = require('./tools');
 const {
@@ -106,10 +106,10 @@ class ChatAgent {
     const preliminaryAnswer =
       result.output?.length > 0 ? `Preliminary Answer: "${result.output.trim()}"` : '';
     const prefix = preliminaryAnswer
-      ? `review and improve the answer you generated using plugins in response to the User Message below. The answer hasn't been sent to the user yet.`
+      ? `review and improve the answer you generated using plugins in response to the User Message below. The user hasn't seen your answer or thoughts yet.`
       : 'respond to the User Message below based on your preliminary thoughts & actions.';
 
-    return `As ChatGPT, ${prefix}${errorMessage}\n${internalActions}
+    return `As a helpful AI Assistant, ${prefix}${errorMessage}\n${internalActions}
 ${preliminaryAnswer}
 Reply conversationally to the User based on your ${
   preliminaryAnswer ? 'preliminary answer, ' : ''
@@ -145,8 +145,7 @@ Only respond with your conversational reply to the following User Message:
       this.options = options;
     }
 
-    this.agentOptions = this.options.agentOptions || {};
-    this.agentIsGpt3 = this.agentOptions.model.startsWith('gpt-3');
+    
     const modelOptions = this.options.modelOptions || {};
     this.modelOptions = {
       ...modelOptions,
@@ -160,10 +159,27 @@ Only respond with your conversational reply to the following User Message:
       stop: modelOptions.stop
     };
 
+    this.agentOptions = this.options.agentOptions || {};
+    this.functionsAgent = this.agentOptions.agent === 'functions';
+    this.agentIsGpt3 = this.agentOptions.model.startsWith('gpt-3');
+    if (this.functionsAgent) {
+      this.agentOptions.model = this.getFunctionModelName(this.agentOptions.model);
+    }
+
     this.isChatGptModel = this.modelOptions.model.startsWith('gpt-');
     this.isGpt3 = this.modelOptions.model.startsWith('gpt-3');
-    this.maxContextTokens = this.modelOptions.model === 'gpt-4-32k' ? 32767 : this.modelOptions.model.startsWith('gpt-4') ? 8191 : 4095,
-
+    const maxTokensMap = {
+      'gpt-4': 8191,
+      'gpt-4-0613': 8191,
+      'gpt-4-32k': 32767,
+      'gpt-4-32k-0613': 32767,
+      'gpt-3.5-turbo': 4095,
+      'gpt-3.5-turbo-0613': 4095,
+      'gpt-3.5-turbo-0301': 4095,
+      'gpt-3.5-turbo-16k': 15999,
+    };
+  
+    this.maxContextTokens = maxTokensMap[this.modelOptions.model] ?? 4095; // 1 less than maximum
     // Reserve 1024 tokens for the response.
     // The max prompt tokens is determined by the max context tokens minus the max response tokens.
     // Earlier messages will be dropped until the prompt is within the limit.
@@ -180,7 +196,7 @@ Only respond with your conversational reply to the following User Message:
     }
 
     this.userLabel = this.options.userLabel || 'User';
-    this.chatGptLabel = this.options.chatGptLabel || 'ChatGPT';
+    this.chatGptLabel = this.options.chatGptLabel || 'Assistant';
 
     // Use these faux tokens to help the AI understand the context since we are building the chat log ourselves.
     // Trying to use "<|im_start|>" causes the AI to still generate "<" or "<|" at the end sometimes for some reason,
@@ -388,6 +404,26 @@ Only respond with your conversational reply to the following User Message:
     this.actions.push(action);
   }
 
+  getFunctionModelName(input) {
+    const prefixMap = {
+      'gpt-4': 'gpt-4-0613',
+      'gpt-4-32k': 'gpt-4-32k-0613',
+      'gpt-3.5-turbo': 'gpt-3.5-turbo-0613'
+    };
+  
+    const prefix = Object.keys(prefixMap).find(key => input.startsWith(key));
+    return prefix ? prefixMap[prefix] : 'gpt-3.5-turbo-0613';
+  }
+
+  createLLM(modelOptions, configOptions) {
+    let credentials = { openAIApiKey: this.openAIApiKey };
+    if (this.azure) {
+      credentials = { ...this.azure };
+    }
+  
+    return new ChatOpenAI({ credentials, ...modelOptions }, configOptions);
+  }
+
   async initialize({ user, message, onAgentAction, onChainEnd, signal }) {
     const modelOptions = {
       modelName: this.agentOptions.model,
@@ -400,21 +436,7 @@ Only respond with your conversational reply to the following User Message:
       configOptions.basePath = this.langchainProxy;
     }
 
-    const model = this.azure
-      ? new ChatOpenAI({
-        ...this.azure,
-        ...modelOptions
-      })
-      : new ChatOpenAI(
-        {
-          openAIApiKey: this.openAIApiKey,
-          ...modelOptions
-        },
-        configOptions
-        // {
-        //   basePath: 'http://localhost:8080/v1'
-        // }
-      );
+    const model = this.createLLM(modelOptions, configOptions);
 
     if (this.options.debug) {
       console.debug(`<-----Agent Model: ${model.modelName} | Temp: ${model.temperature}----->`);
@@ -466,7 +488,8 @@ Only respond with your conversational reply to the following User Message:
     };
 
     // initialize agent
-    this.executor = await initializeCustomAgent({
+    const initializer = this.options.agentOptions?.agent === 'functions' ? initializeFunctionsAgent : initializeCustomAgent;
+    this.executor = await initializer({
       model,
       signal,
       tools: this.tools,
@@ -594,7 +617,7 @@ Only respond with your conversational reply to the following User Message:
     console.log('sendMessage', message, opts);
 
     const user = opts.user || null;
-    const { onAgentAction, onChainEnd, onProgress } = opts;
+    const { onAgentAction, onChainEnd } = opts;
     const conversationId = opts.conversationId || crypto.randomUUID();
     const parentMessageId = opts.parentMessageId || '00000000-0000-0000-0000-000000000000';
     const userMessageId = opts.overrideParentMessageId || crypto.randomUUID();
@@ -658,13 +681,13 @@ Only respond with your conversational reply to the following User Message:
       return { ...responseMessage, ...this.result };
     }
 
-    if (!this.agentIsGpt3 && this.result.output) {
-      responseMessage.text = this.result.output;
-      await this.saveMessageToDatabase(responseMessage, user);
-      const textStream = new TextStream(this.result.output);
-      await textStream.processTextStream(onProgress);
-      return { ...responseMessage, ...this.result };
-    }
+    // if (!this.agentIsGpt3 && this.result.output) {
+    //   responseMessage.text = this.result.output;
+    //   await this.saveMessageToDatabase(responseMessage, user);
+    //   const textStream = new TextStream(this.result.output);
+    //   await textStream.processTextStream(opts.onProgress);
+    //   return { ...responseMessage, ...this.result };
+    // }
 
     if (this.options.debug) {
       console.debug('this.result', this.result);
@@ -871,7 +894,7 @@ Only respond with your conversational reply to the following User Message:
     return orderedMessages.map((msg) => ({
       messageId: msg.messageId,
       parentMessageId: msg.parentMessageId,
-      role: msg.isCreatedByUser ? 'User' : 'ChatGPT',
+      role: msg.isCreatedByUser ? 'User' : 'Assistant',
       text: msg.text
     }));
   }

api/app/langchain/agents/CustomAgent/initializeCustomAgent.js

@@ -51,6 +51,4 @@ Query: {input}
   return AgentExecutor.fromAgentAndTools({ agent, tools, memory, ...rest });
 };
 
-module.exports = {
-  initializeCustomAgent
-};
+module.exports = initializeCustomAgent;

api/app/langchain/agents/Functions/FunctionsAgent.js

@@ -0,0 +1,122 @@
+const { Agent } = require('langchain/agents');
+const { LLMChain } = require('langchain/chains');
+const { FunctionChatMessage, AIChatMessage } = require('langchain/schema');
+const {
+  ChatPromptTemplate,
+  MessagesPlaceholder,
+  SystemMessagePromptTemplate,
+  HumanMessagePromptTemplate
+} = require('langchain/prompts');
+const PREFIX = `You are a helpful AI assistant. Objective: Understand the human's query with available functions.
+The user is expecting a function response to the query; if only part of the query involves a function, prioritize the function response.`;
+
+function parseOutput(message) {
+  if (message.additional_kwargs.function_call) {
+    const function_call = message.additional_kwargs.function_call;
+    return {
+      tool: function_call.name,
+      toolInput: function_call.arguments ? JSON.parse(function_call.arguments) : {},
+      log: message.text
+    };
+  } else {
+    return { returnValues: { output: message.text }, log: message.text };
+  }
+}
+
+class FunctionsAgent extends Agent {
+  constructor(input) {
+    super({ ...input, outputParser: undefined });
+    this.tools = input.tools;
+  }
+
+  lc_namespace = ['langchain', 'agents', 'openai'];
+
+  _agentType() {
+    return 'openai-functions';
+  }
+
+  observationPrefix() {
+    return 'Observation: ';
+  }
+
+  llmPrefix() {
+    return 'Thought:';
+  }
+
+  _stop() {
+    return ['Observation:'];
+  }
+
+  static createPrompt(_tools, fields) {
+    const { prefix = PREFIX, currentDateString } = fields || {};
+
+    return ChatPromptTemplate.fromPromptMessages([
+      SystemMessagePromptTemplate.fromTemplate(`Date: ${currentDateString}\n${prefix}`),
+      HumanMessagePromptTemplate.fromTemplate(`{chat_history}
+Query: {input}
+{agent_scratchpad}`),
+      new MessagesPlaceholder('agent_scratchpad')
+    ]);
+  }
+
+  static fromLLMAndTools(llm, tools, args) {
+    FunctionsAgent.validateTools(tools);
+    const prompt = FunctionsAgent.createPrompt(tools, args);
+    const chain = new LLMChain({
+      prompt,
+      llm,
+      callbacks: args?.callbacks
+    });
+    return new FunctionsAgent({
+      llmChain: chain,
+      allowedTools: tools.map((t) => t.name),
+      tools
+    });
+  }
+
+  async constructScratchPad(steps) {
+    return steps.flatMap(({ action, observation }) => [
+      new AIChatMessage('', {
+        function_call: {
+          name: action.tool,
+          arguments: JSON.stringify(action.toolInput)
+        }
+      }),
+      new FunctionChatMessage(observation, action.tool)
+    ]);
+  }
+
+  async plan(steps, inputs, callbackManager) {
+    // Add scratchpad and stop to inputs
+    var thoughts = await this.constructScratchPad(steps);
+    var newInputs = Object.assign({}, inputs, { agent_scratchpad: thoughts });
+    if (this._stop().length !== 0) {
+      newInputs.stop = this._stop();
+    }
+
+    // Split inputs between prompt and llm
+    var llm = this.llmChain.llm;
+    var valuesForPrompt = Object.assign({}, newInputs);
+    var valuesForLLM = {
+      tools: this.tools
+    };
+    for (var i = 0; i < this.llmChain.llm.callKeys.length; i++) {
+      var key = this.llmChain.llm.callKeys[i];
+      if (key in inputs) {
+        valuesForLLM[key] = inputs[key];
+        delete valuesForPrompt[key];
+      }
+    }
+
+    var promptValue = await this.llmChain.prompt.formatPromptValue(valuesForPrompt);
+    var message = await llm.predictMessages(
+      promptValue.toChatMessages(),
+      valuesForLLM,
+      callbackManager
+    );
+    console.log('message', message);
+    return parseOutput(message);
+  }
+}
+
+module.exports = FunctionsAgent;

api/app/langchain/agents/Functions/initializeFunctionsAgent.js

@@ -0,0 +1,33 @@
+const FunctionsAgent = require('./FunctionsAgent');
+const { AgentExecutor } = require('langchain/agents');
+const { BufferMemory, ChatMessageHistory } = require('langchain/memory');
+
+const initializeFunctionsAgent = async ({
+  tools,
+  model,
+  pastMessages,
+  currentDateString,
+  ...rest
+}) => {
+  const agent = FunctionsAgent.fromLLMAndTools(
+    model,
+    tools, 
+    {
+      currentDateString,
+    });
+
+  const memory = new BufferMemory({
+    chatHistory: new ChatMessageHistory(pastMessages),
+    // returnMessages: true, // commenting this out retains memory
+    memoryKey: 'chat_history',
+    humanPrefix: 'User',
+    aiPrefix: 'Assistant',
+    inputKey: 'input',
+    outputKey: 'output'
+  });
+
+  return AgentExecutor.fromAgentAndTools({ agent, tools, memory, ...rest });
+};
+
+module.exports = initializeFunctionsAgent;
+

api/app/langchain/agents/index.js

@@ -0,0 +1,7 @@
+const initializeCustomAgent = require('./CustomAgent/initializeCustomAgent');
+const initializeFunctionsAgent = require('./Functions/initializeFunctionsAgent');
+
+module.exports = {
+  initializeCustomAgent,
+  initializeFunctionsAgent
+};

api/package.json

@@ -46,7 +46,7 @@
     "jsonwebtoken": "^9.0.0",
     "keyv": "^4.5.2",
     "keyv-file": "^0.2.0",
-    "langchain": "^0.0.92",
+    "langchain": "^0.0.94",
     "lodash": "^4.17.21",
     "meilisearch": "^0.33.0",
     "mongoose": "^7.1.1",

api/server/routes/ask/askGPTPlugins.js

@@ -39,8 +39,8 @@ router.post('/', requireJwtAuth, async (req, res) => {
   if (endpoint !== 'gptPlugins') return handleError(res, { text: 'Illegal request' });
 
   const agentOptions = req.body?.agentOptions ?? {
+    agent: 'classic',
     model: 'gpt-3.5-turbo',
-    // model: 'gpt-4', // for agent model
     temperature: 0,
     // top_p: 1,
     // presence_penalty: 0,
@@ -60,20 +60,12 @@ router.post('/', requireJwtAuth, async (req, res) => {
       presence_penalty: req.body?.presence_penalty ?? 0,
       frequency_penalty: req.body?.frequency_penalty ?? 0
     },
-    agentOptions
+    agentOptions: {
+      ...agentOptions,
+      // agent: 'functions'
+    }
   };
 
-  // const availableModels = getOpenAIModels();
-  // if (availableModels.find((model) => model === endpointOption.modelOptions.model) === undefined) {
-  //   return handleError(res, { text: `Illegal request: model` });
-  // }
-
-  // console.log('ask log', {
-  //   text,
-  //   conversationId,
-  //   endpointOption
-  // });
-
   console.log('ask log');
   console.dir({ text, conversationId, endpointOption }, { depth: null });
 

api/server/routes/endpoints.js

@@ -53,7 +53,7 @@ router.get('/', async function (req, res) {
     ? { availableModels: getOpenAIModels(), userProvide: apiKey === 'user_provided' }
     : false;
   const gptPlugins = apiKey
-    ? { availableModels: getPluginModels(), availableTools }
+    ? { availableModels: getPluginModels(), availableTools, availableAgents: ['classic', 'functions'] }
     : false;
   const bingAI = process.env.BINGAI_TOKEN
     ? { userProvide: process.env.BINGAI_TOKEN == 'user_provided' }