Calvinist Parrot

Overview

The Calvinist Parrot project embarked on an ambitious journey to create an AI-powered chatbot to facilitate a deeper understanding and exploration of the Bible from a Reformed theology perspective. Serving as both a tool and a learning companion, the application continually evolves to enhance user interactions and deepen understanding of religious texts.

Links: Calvinist Parrot | Github Repo

Tools

  • Main Chat (v2.8 Parrot):

    • Agents: Interact with Parrot, Calvin, and the CCEL Librarian.

    • Brief Mode: Shorter responses and interaction with Parrot and Calvin only.

  • Quick QA Feature:

    • Chain-of-Reasoning Approach: A multi-step reasoning chain with three disting agents provides an initial response, then leaves the last step to an orchestrator to ensure better response accuracy.

  • CCEL Search:

    • Extensive Library Access: Search through over 900 indexed books in the CCEL.

  • Study Helper:

    • Commentary Access: Retrieve commentaries from Bible Hub.

    • Multi-Step Queries: Breaks down questions into sub-questions for thorough analysis.

  • Devotionals:

    • AI-Generated Content: Morning and evening devotionals based on current news.

  • Sermon Review:

    • Evaluation Framework: Analyze sermons using Bryan Chappell's framework.

Goal

  • Engineer an AI-driven chatbot capable of providing thoughtful, accurate, and insightful responses to queries about the Bible from a Reformed theology standpoint.

  • Continually enhance the AI's comprehension and response capabilities to foster a more profound exploration of Scriptures for users.

Solutions

Initial Development:

  • Dual AI Agents: Developed an AI chatbot employing two GPT agents, Parrot and Calvin, to provide insightful and accurate responses curated to align with the principles of Reformed theology.

  • Feedback Loop: Incorporated a continuous learning and improvement system based on user feedback.

Version 2.0 Enhancements:

  • Autonomous Agent with Multiple Query Engines:

    • 11 Distinct Query Engines: Utilized to search through the vast Christian Classics Ethereal Library (CCEL).

    • GPT Integration: Query engines are powered by GPT-3.5, while the primary agent leverages GPT-4.

  • Library Indexing and Categorization:

    • Novel Algorithm Development: Created a patent-pending algorithm for analyzing books, generating multiple categories, and forming numerous query engines.

    • Accessibility: Made the extensive religious texts accessible and searchable.

Further Developments:

  • Multi-Container and Multi-Build Architecture:

    • Docker Implementation: Each Docker container houses a distinct query engine.

    • Performance Optimization: Managed the extensive library effectively without resorting to larger instances, leading to faster response times and lower resource utilization.

  • Additional Features:

    • Main Chat (v2.8 Parrot): Logging in is required to save and retrieve conversations, and three AI agents—Parrot, Calvin, and the CCEL Librarian—are utilized.

      • Brief Mode: Accessible without login, allowing interactions with Parrot and Calvin.

    • CCEL Search: Users can search over 900 indexed books in the Christian Classics Ethereal Library.

    • Study Helper: Accesses commentaries from Bible Hub using a multi-step query engine for thorough analysis.

    • Devotionals: Provides AI-generated morning and evening devotionals based on the latest news.

    • Sermon Review: Allows users to review sermons using Bryan Chappell's framework from Christ-Centered Preaching.

Findings

  • Reduction in Hallucinations: Multiple query engines significantly reduced inaccuracies, enhancing the reliability of responses.

  • Improved Response Depth: Access to the extensive CCEL enabled more insightful and well-informed answers regarding Reformed theology and biblical texts.

  • Enhanced User Experience: The autonomous agent and innovative indexing algorithm provided quicker, more precise answers, enriching user interactions.

Takeaways

The evolution of the Calvinist Parrot project emphasized several technical and innovative breakthroughs:

  • Algorithm Development: The novel algorithm for indexing and categorizing the CCEL was instrumental in organizing vast religious texts into manageable categories.

  • Autonomous Query Execution: Engineering an agent capable of selecting appropriate query engines marked a significant technical advancement for contextually relevant responses.

  • Performance Optimization: The multi-container architecture optimized performance, leading to faster responses and efficient resource management.

VERSION HISTORY

  • September 2024 Updates:

    • Quick QA Feature: Added to the homepage for easy, no-login-required questions.

    • Brief Mode in Main Chat: Accessible without login; full features available upon login to save conversations.

  • August 2024 Update:

  • July 2024 Update:

  • June 2024 Update (v2.7):

    • CCEL Tool Reintroduced: The CCEL Librarian joined the main chat.

  • March 2024 Update (v2.4):

    • Sermon Review Tool Added: Allows sermon evaluation using Bryan Chappell's framework.

  • February 2024 Update:

    • Main Chat Deprecation: Due to funding constraints, the CCEL index in the main chat was deprecated.

    • New Tool Development: I began working on a new Bible study tool.

  • November 2023 Update (v2.3):

    • BRB Integration: Incorporated the Biblical Reference Browser.

  • October 2023 Updates:

    • v2.2: Introduced the Devotionals tool.

    • v2.1: Added the Study Helper.

    • v2.0: Transitioned to a Python app using Streamlit, moving away from JavaScript.

  • April 2023 (v1):

    • Initial Release: Launched as a simple JavaScript chatbot using adversarial agents to ensure biblically grounded responses.