Config Options

export const ragChat = new RAGChat({
  model: openai("gpt-4-turbo"),

  vector: new Index({
    url: "UPSTASH_VECTOR_REST_URL",
    token: "UPSTASH_VECTOR_REST_TOKEN",
  }),

  redis: new Redis({
    url: "UPSTASH_REDIS_REST_URL",
    token: "UPSTASH_REDIS_REST_TOKEN",
  }),

  ratelimit: new Ratelimit({
    redis: Redis.fromEnv(),
    limiter: Ratelimit.slidingWindow(10, "10 s"),
  }),
  
  debug: true,
  /** 
    * Set to `true` if working with web apps and you want to be interactive without stalling users.
    */
  streaming: true,
  /** Chat session ID of the user interacting with the application.
    * @default "upstash-rag-chat-session"
    */
  sessionId: "session_id",
  /**
    * Namespace of the index you wanted to query.
    */
  namespace: "namespace",
  /**
    * Metadata for your chat message. This could be used to store anything in the chat history. 
    * By default RAG Chat SDK uses this to persist used model name in the history
    */
  metadata: { key: "value"},
  
  /** Rate limit session ID of the user interacting with the application.
    * @default "upstash-rag-chat-ratelimit-session"
    */
  ratelimitSessionId: "RatelimitSessionId"

  promptFn: ({context, question, chatHistory}) =>
    `You are an AI assistant with access to an Upstash Vector Store.
      Use the provided context and chat history to answer the question.
      If the answer isn't available, politely inform the user.
      ------
      Chat history:
      ${chatHistory}
      ------
      Context:
      ${context}
      ------
      Question: ${question}
      Answer:`,
});

ragChat.chat(
  "Tell me about machine learning",
  {
    /** Length of the conversation history to include in your LLM query. Increasing this may lead to hallucinations. Retrieves the last N messages.
     * @default 5
     */
    historyLength : 5,
    /** Configuration to retain chat history. After the specified time, the history will be automatically cleared.
     * @default 86_400 // 1 day in seconds
     */
    historyTTL : 86_400,
    /** Configuration to adjust the accuracy of results.
     * @default 0.5
     */
    similarityThreshold : 0.5,
    /** Amount of data points to include in your LLM query.
     * @default 5
     */
    topK : 5,
    /**
     *  Details of applied rate limit.
     */
    ratelimitDetails : (response: Awaited<ReturnType<Ratelimit["limit"]>>) => {},
    /**
     * Hook to modify or get data and details of each chunk. Can be used to alter streamed content.
     */
    onChunk : ({content, inputTokens, chunkTokens, totalTokens, rawContent,}: {
        inputTokens: number;
        chunkTokens: number;
        totalTokens: number;
        content: string;
        rawContent: string;
    }) => {},
    /**
     * Hook to access the retrieved context and modify as you wish.
     */
    onContextFetched : (context: PrepareChatResult) => {
        console.log(context);
        return context
    },
    /**
     * Hook to access the retrieved history and modify as you wish.
     */
    onChatHistoryFetched : (messages: UpstashMessage[]) => {
        console.log(messages);
        return messages;
    };
    /**
     * Allows disabling RAG and use chat as LLM in combination with prompt. This will give you ability to build your own pipelines.
     */
    disableRAG : false
  }
);