APIChains and Agents

[KeenanFernandes2000]

I wanted to know since the APIChain is a legacy chain is there a way to get agents to interact with an API endpoint by just proving the documentation.

[dosubot[bot]]

🤖

Sure, I can guide you on how to use agents in the LangChain Javascript framework to interact with an API endpoint using just the documentation. Here's a step-by-step guide on how to do this:

1. Import the necessary modules from the LangChain framework.

import { FaissStore } from "@langchain/community/vectorstores/faiss";
import { OpenAIEmbeddings } from "langchain/embeddings/openai";
import { TextLoader } from "langchain/document_loaders/fs/text";
import { RecursiveCharacterTextSplitter } from "langchain/text_splitter";
import {
  createRetrieverTool,
  createConversationalRetrievalAgent,
} from "langchain/agents/toolkits";
import { ChatOpenAI } from "langchain/chat_models/openai";

2. Load the documents you want to use for the conversation.

const loader = new TextLoader("state_of_the_union.txt");
const docs = await loader.load();

3. Split the documents into chunks.

const splitter = new RecursiveCharacterTextSplitter({
  chunkSize: 1000,
  chunkOverlap: 0,
});
const texts = await splitter.splitDocuments(docs);

4. Create a vector store from the documents.

const vectorStore = await FaissStore.fromDocuments(
  texts,
  new OpenAIEmbeddings()
);

5. Create a retriever from the vector store.

const retriever = vectorStore.asRetriever();

6. Create a retriever tool with the retriever.

const tool = createRetrieverTool(retriever, {
  name: "search_state_of_union",
  description:
    "Searches and returns documents regarding the state-of-the-union.",
});

7. Create a model for the conversation.

const model = new ChatOpenAI({});

8. Create a conversational retrieval agent with the model and the tool.

const executor = await createConversationalRetrievalAgent(model, [tool], {
  verbose: true,
});

9. Invoke a conversation with the user and retrieve information based on the user's input.

const result = await executor.invoke({
  input: "Hi, I'm Bob!",
});
console.log(result);

Please note that the above example is based on the context provided and may need to be adjusted based on the specific requirements of your application and the API endpoint you are interacting with. You can find more details in the examples/src/guides/conversational_retrieval/agent.ts file in the repository.

Regarding your question about the FaissStore module, it is responsible for managing and manipulating vectors associated with language models. It provides a set of methods for creating, adding, searching, merging, and deleting vectors. However, it does not directly interact with API endpoints. Instead, it provides the underlying functionality for managing vectors that can be used by other parts of the LangChain framework, which may expose this functionality through API endpoints. You can find more details in the libs/langchain-community/src/vectorstores/tests/faiss.test.ts file in the repository.

I hope this helps! If you have any other questions, feel free to ask.

Sources

• examples/src/guides/conversational_retrieval/agent.ts
• libs/langchain-community/src/vectorstores/tests/faiss.test.ts

About Dosu

This response is meant to be useful and save you time. It is not meant to be a precise solution, but rather a starting point for your own research.

Help me be more useful! Please leave a 👍 if this is helpful and 👎 if it is irrelevant.

If you want to continue the conversation, start your reply with @dosu-bot.