Using Chat History in Ragged

One of the key features of Ragged is low-level control over conversation history. This tutorial will guide you through the process of using the history feature in Ragged to create a persistent conversation history.

Prerequisites

Before getting started, make sure you have the following:

• Node.js installed on your machine
• Ragged library installed in your project
• API credentials for the desired LLM provider (e.g., OpenAI)

Step 1: Initialize Ragged

To begin, import the necessary modules from the Ragged library and create a new instance of the Ragged class with your desired provider and configuration.

import { Ragged } from "ragged";
import type { RaggedHistoryItem } from "ragged";

const r = new Ragged({
  provider: "openai",
  config: {
    apiKey: process.env.OPENAI_CREDS,
  },
});

Step 2: Initialize History Array

Create an empty array called history to store the conversation history. This array will be used to accumulate the responses from each chat interaction.

import { Ragged } from "ragged";
import type { RaggedHistoryItem } from "ragged";

const r = new Ragged({
  provider: "openai",
  config: {
    apiKey: process.env.OPENAI_CREDS,
  },
});

let history: RaggedHistoryItem[] = [];

Step 3: Send Initial Prompt

Define your initial prompt and send it to the LLM using the r.chat() method. Use the asHistory() method to obtain the response as a RaggedHistoryItem array.

import { Ragged } from "ragged";
import type { RaggedHistoryItem } from "ragged";

const r = new Ragged({
  provider: "openai",
  config: {
    apiKey: process.env.OPENAI_CREDS,
  },
});

let history: RaggedHistoryItem[] = [];

const PROMPT_1 = "say 'Hello World!'";
const firstResponseHistory = await r.chat(PROMPT_1).asHistory();

Step 4: Update History Array

Update the history array by concatenating the response history from the previous step.

import { Ragged } from "ragged";
import type { RaggedHistoryItem } from "ragged";

const r = new Ragged({
  provider: "openai",
  config: {
    apiKey: process.env.OPENAI_CREDS,
  },
});

let history: RaggedHistoryItem[] = [];

const PROMPT_1 = "say 'Hello World!'";
const firstResponseHistory = await r.chat(PROMPT_1).asHistory();

history = [...history, ...firstResponseHistory];

Step 5: Send Subsequent Prompts

Define your next prompt and send it to the LLM along with the updated history array. This allows the LLM to have context from the previous interactions.

import { Ragged } from "ragged";
import type { RaggedHistoryItem } from "ragged";

const r = new Ragged({
  provider: "openai",
  config: {
    apiKey: process.env.OPENAI_CREDS,
  },
});

let history: RaggedHistoryItem[] = [];

const PROMPT_1 = "say 'Hello World!'";
const firstResponseHistory = await r.chat(PROMPT_1).asHistory();

history = [...history, ...firstResponseHistory];

const PROMPT_2 = "say 'Hello World!' again";
const secondHistoryResponse = await r
  .chat([
    ...history,
    {
      type: "history.text",
      role: "human",
      data: {
        text: PROMPT_2,
      },
    },
  ])
  .asHistory();

Step 6: Repeat Steps 4-5

Repeat steps 4-5 for each subsequent prompt, updating the history array and sending the next prompt along with the updated history.

import { Ragged } from "ragged";
import type { RaggedHistoryItem } from "ragged";

const r = new Ragged({
  provider: "openai",
  config: {
    apiKey: process.env.OPENAI_CREDS,
  },
});

let history: RaggedHistoryItem[] = [];

const PROMPT_1 = "say 'Hello World!'";
const firstResponseHistory = await r.chat(PROMPT_1).asHistory();

history = [...history, ...firstResponseHistory];

const PROMPT_2 = "say 'Hello World!' again";
const secondHistoryResponse = await r
  .chat([
    ...history,
    {
      type: "history.text",
      role: "human",
      data: {
        text: PROMPT_2,
      },
    },
  ])
  .asHistory();

history = [...history, ...secondHistoryResponse];

Step 7: Access Complete History

After all the chat interactions are complete, you can access the complete conversation history by using the history array for further processing.

import { Ragged } from "ragged";
import type { RaggedHistoryItem } from "ragged";

const r = new Ragged({
  provider: "openai",
  config: {
    apiKey: process.env.OPENAI_CREDS,
  },
});

let history: RaggedHistoryItem[] = [];

const PROMPT_1 = "say 'Hello World!'";
const firstResponseHistory = await r.chat(PROMPT_1).asHistory();

history = [...history, ...firstResponseHistory];

const PROMPT_2 = "say 'Hello World!' again";
const secondHistoryResponse = await r
  .chat([
    ...history,
    {
      type: "history.text",
      role: "human",
      data: {
        text: PROMPT_2,
      },
    },
  ])
  .asHistory();

history = [...history, ...secondHistoryResponse];

// Access and process the complete conversation history
console.log(history);

Conclusion

By following this tutorial, you have learned how to use the history feature in Ragged to maintain a persistent conversation history across multiple chat interactions. This allows the LLM to provide more contextually relevant responses based on the previous interactions.

Remember to update the history array accordingly after each chat interaction. You can also customize the behavior by handling different item types in the response history.

With the power of Ragged and the history feature, you can build sophisticated conversational AI applications that maintain context and provide engaging user experiences.