DOCSEARCHbyAlgolia
Submit Docs

Peki Chatbot Platform

website·Indexed 31 days ago

https://docs.peki.ai

An advanced solution for building, managing, and deploying intelligent chatbots. Features include Retrieval-Augmented Generation (RAG), custom task and function definitions, a JavaScript SDK for website integration, and a Developer API for managing messages, users, and sessions. Includes comprehensive analytics for tracking session sentiment, token usage, and conversation outcomes.

Tokens
10.5K
Snippets
32
Records
90
Agent score
52%

What's inside Peki

  1. Overview of Chatbot Functions

    Functions are tool calls that extend a chatbot's capabilities by integrating it with external services (such as APIs or SMTP). They allow the chatbot to perform dynamic actions, handle complex workflows, and interact with external infrastructure in real-time. Functions are triggered automatically based on user interactions or predefined tasks.

  2. Overview of Peki Chatbot Platform capabilities

    The Peki Chatbot Platform is a solution for building, managing, and deploying intelligent chatbots. It features a visual configuration interface for managing prompts (Tasks), tool calls (Functions), file-based knowledge, and messaging provider integrations.

  3. Monitor chatbot interactions using the Conversation Log

    The Conversation Log is accessible within a selected session and provides a complete history of messages exchanged between the user and the chatbot. It is used to diagnose communication issues, monitor effectiveness, and identify areas for improvement. Each message entry includes the Sender (user or chatbot), the Message Content, and a Timestamp.

  4. Understand and analyze chatbot sessions in Peki

    A session in the Peki Chatbot Platform represents a single conversation between a user and the chatbot, starting from the first message and ending upon completion or a period of inactivity. Sessions are used to track user behavior, chatbot performance, and conversation outcomes. You can view session data by navigating to the Messages Page.

  5. Understand how session language is detected and handled

    The platform automatically detects the language of each chat session using the active AI model based on the content of user messages. The chatbot dynamically adjusts and provides responses in the same detected language, enabling seamless multilingual support without manual intervention. If the AI model cannot recognize the language, the chatbot defaults to the primary language configured for the bot.

  6. Define chatbot behavior using Tasks

    Tasks in Peki.ai are prompts or directives (system messages) that guide the AI's responses, personality, tone, and purpose. By creating specific tasks, you can tailor the chatbot's interactions to meet specific business needs, such as acting as a technical expert or a friendly support agent.

  7. Use HTTP Actions to call external APIs

    The HTTP Action allows the chatbot to make API calls to external services to retrieve or send data. For example, if a user asks for details about a specific product ID, the chatbot can trigger an HTTP Action to fetch that data from an API endpoint and then display the formatted result to the user.

  8. Implement RAG using the Files section

    Retrieval-Augmented Generation (RAG) allows the chatbot to reference external knowledge sources (like PDFs, CSVs, and DOCX files) during conversations instead of relying solely on pre-trained AI models. This ensures more accurate, relevant, and document-backed responses for domain-specific queries.

  9. Integrate Peki chatbot for authenticated users

    For logged-in users, the Peki integration requires a userId and name in addition to the chatbotId. You can also provide a dataDictionary to pass additional user metadata like email, phone, and city to the chatbot.
    <script>
  window.pekiSettings = {
    chatbotId: "YOUR_CHATBOT_ID",
    userId: "unique_user_id",
    name: "User Name",
    dataDictionary: {
      email: "user@example.com",
      phone: "+1234567890",
      city: "New York",
    },
    welcomeMessage: "Hello, how can I assist you today?",
    popup: false,
    buttonPosition: "right",
    theme: "light",
  };
  !(function () {
    var script = document.createElement("script");
    script.src = "https://script.peki.ai/chatbox";
    script.async = true;
    document.body.appendChild(script);
  })();
</script>

  10. Interpret and use the Session Count chart

    The Session Count chart tracks the total number of chat sessions initiated by users over time (daily, weekly, or monthly). It is used to monitor user adoption, identify engagement patterns (such as seasonal or daily peaks), and correlate activity spikes with external events like marketing campaigns or feature releases.

    Key actions based on this data include:

    • Scaling support resources during identified peak traffic periods.
    • Investigating drops in activity to find potential causes.
    • Benchmarking current session counts against historical data to evaluate the impact of chatbot improvements.

  11. Interpret and use the Session Sentiment chart

    The Session Sentiment chart tracks sentiment trends in chat sessions over time using natural language processing to evaluate the tone of user messages.

    • Positive Sentiment: Indicates a good user experience.
    • Negative Sentiment: Signals user frustration or dissatisfaction.

    Common use cases include monitoring overall chatbot effectiveness, detecting recurring issues that trigger negative responses, and optimizing the chatbot's tone and clarity.

  12. Improve chatbot resolution rates based on Session Result analytics

    When analyzing the Session Result chart, take the following actions to optimize bot performance:

    • Increase Resolution Rates: Improve training data to help the bot handle more queries successfully.
    • Fix Unresolved Queries: Identify common patterns in 'Unresolved' sessions and update the chatbot's responses or functions.
    • Enhance Capabilities: Review 'Escalated' cases to determine which specific capabilities are missing from the chatbot that necessitate human intervention.