AI Agent simplifies the implementation and use of generative AI with LangChain, you can add components such as vectorized search services (check options in "link"), conversation history (check options in "link"), custom databases (check options in "link") and API contracts (OpenAPI).
Use the package manager npm to install AI Agent.
npm install ai-agentLLM + Prompt Engineering
const agent = new Agent({
name: '<name>',
systemMesssage: '<a message that will specialize your agent>',
llmConfig: {
type: '<cloud-provider-llm-service>', // Check availability at <link>
model: '<llm-model>',
instance: '<instance-name>', // Optional
apiKey: '<key-your-llm-service>', // Optional
},
chatConfig: {
temperature: 0,
},
});
// If stream enabled, receiver on token
agent.on('onToken', async (token) => {
console.warn('token:', token);
});
agent.on('onMessage', async (message) => {
console.warn('MESSAGE:', message);
});
await agent.call({
question: 'What is the best way to get started with Azure?',
chatThreadID: '<chat-id>',
stream: true,
});When you use LLM + Chat history all message exchange is persisted and sent to LLM.
const agent = new Agent({
name: '<name>',
systemMesssage: '<a message that will specialize your agent>',
chatConfig: {
temperature: 0,
},
llmConfig: {
type: '<cloud-provider-llm-service>', // Check availability at <link>
model: '<llm-model>',
instance: '<instance-name>', // Optional
apiKey: '<key-your-llm-service>', // Optional
},
dbHistoryConfig: {
type: '<type-database>', // Check availability at <link>
host: '<host-database>', // Optional
port: "<port-database>", // Optional
sessionTTL: '<ttl-database>' // Optional. Time the conversation will be saved in the database
limit: '<limit-messages>' // Optional. Limit set for maximum messages included in conversation prompt
},
});
// If stream enabled, receiver on token
agent.on('onToken', async (token) => {
console.warn('token:', token);
});
agent.on('onMessage', async (message) => {
console.warn('MESSAGE:', message);
});
await agent.call({
question: 'What is the best way to get started with Azure?',
chatThreadID: '<chat-id>',
stream: true,
});When using LLM + Vector stores the Agent finds the documents relevant to the requested input. The documents found are used for the context of the Agent.
const agent = new Agent({
name: '<name>',
systemMesssage: '<a message that will specialize your agent>',
chatConfig: {
temperature: 0,
},
llmConfig: {
type: '<cloud-provider-llm-service>', // Check availability at <link>
model: '<llm-model>',
instance: '<instance-name>', // Optional
apiKey: '<key-your-llm-service>', // Optional
},
vectorStoreConfig: {
type: '<type-vector-service>', // Check availability at <link>
apiKey: '<your-api-key>', // Optional
indexes: ['<index-name>'], // Your indexes name. Optional
vectorFieldName: '<vector-base-field>', // Optional
name: '<vector-service-name>', // Optional
apiVersion: "<api-version>", // Optional
model: '<llm-model>' // Optional
customFilters: '<custom-filter>' // Optional. Example: 'field-vector-store=(userSessionId)' check at <link>
},
});
// If stream enabled, receiver on token
agent.on('onToken', async (token) => {
console.warn('token:', token);
});
agent.on('onMessage', async (message) => {
console.warn('MESSAGE:', message);
});
await agent.call({
question: 'What is the best way to get started with Azure?',
chatThreadID: '<chat-id>',
stream: true,
});SQL + LLM for prompt construction is a concept that involves using both Structured Query Language (SQL) and LLMs to create queries or prompts for data retrieval or interaction with databases. This approach leverages the power of SQL for database-specific commands and the capabilities of LLMs to generate natural language prompts, making it easier for users to interact with databases and retrieve information in a more user-friendly and intuitive manner.
const agent = new Agent({
name: '<name>',
systemMesssage: '<a message that will specialize your agent>',
chatConfig: {
temperature: 0,
},
llmConfig: {
type: '<cloud-provider-llm-service>', // Check availability at <link>
model: '<llm-model>',
instance: '<instance-name>', // Optional
apiKey: '<key-your-llm-service>', // Optional
},
dataSourceConfig: {
type: '<type-database>', // Check availability at <link>
username: '<username-database>', // Require
password: '<username-pass>', // Require
host: '<host-database>', // Require
name: '<connection-name>', // Require
includesTables: ['<table-name>'], // Optional
ssl: '<ssl-mode>', // Optional
maxResult: '<max-result-database>', // Optional. Limit set for maximum data included in conversation prompt.
customizeSystemMessage: '<custom-chain-prompt>', // Optional. Adds prompt specifications for custom database operations.
},
});
// If stream enabled, receiver on token
agent.on('onToken', async (token) => {
console.warn('token:', token);
});
agent.on('onMessage', async (message) => {
console.warn('MESSAGE:', message);
});
await agent.call({
question: 'What is the best way to get started with Azure?',
chatThreadID: '<chat-id>',
stream: true,
});OpenAPI + LLM for prompt construction is a concept that combines OpenAPI, a standard for documenting and describing RESTful APIs, with large language models (LLMs). This fusion allows for the automated generation of prompts or queries for interacting with APIs. By using LLMs to understand the OpenAPI specifications and generate natural language prompts, it simplifies and streamlines the process of interfacing with APIs, making it more user-friendly and accessible.
const agent = new Agent({
name: '<name>',
systemMesssage: '<a message that will specialize your agent>',
chatConfig: {
temperature: 0,
},
llmConfig: {
type: '<cloud-provider-llm-service>', // Check availability at <link>
model: '<llm-model>',
instance: '<instance-name>', // Optional
apiKey: '<key-your-llm-service>', // Optional
},
openAPIConfig: {
xApiKey: '<x-api-key>', // Optional. Using request API
data: '<data-contract>', // Require. OpenAPI contract
customizeSystemMessage: '<custom-chain-prompt>', // Optional. Adds prompt specifications for custom openAPI operations.
},
});
// If stream enabled, receiver on token
agent.on('onToken', async (token) => {
console.warn('token:', token);
});
agent.on('onMessage', async (message) => {
console.warn('MESSAGE:', message);
});
await agent.call({
question: 'What is the best way to get started with Azure?',
chatThreadID: '<chat-id>',
stream: true,
});If you've ever wanted to contribute to open source, and a great cause, now is your chance!
See the contributing docs for more information
 JP. Nobrega 💬 📖 👀 📢 |