Buddhi AI Changelog

• 0.4.0

  Dec 5, 2025

  Introducing Private In-Browser RAG

  

  We recognize that data privacy and confidentiality are paramount in today’s landscape. This major release introduces In-Browser RAG (Retrieval-Augmented Generation), a powerful feature that transforms Buddhi AI into the definitive tool for interacting with confidential and private documents.

  The core value proposition is simple: you can now upload PDF text documents directly to the chat interface, and then ask complex questions or extract information from them, all without ever sending the data to a remote server. This is a game-changer for establishing privacy-first workflows.

  ---

  Core Feature: Private Document Q&A

  Users gain the power of RAG, enabling them to upload their own PDF text documents. Buddhi AI will then index and process this content locally in the browser, creating a contextual, private knowledge base. Once indexed, you can ask detailed questions about the documents, allowing for instant and private retrieval of specific answers, summaries, or data points.

  ---

  Strategic and Technical Advantage

  This update centers on the integrity of our client-side processing.

  • 100% Client-Side Processing: The entire RAG pipeline, including the resource-intensive tasks of vector embedding generation, storage, and retrieval, runs exclusively in the user’s browser.
  • Zero Server Risk: Documents and their corresponding vector embeddings never leave the user’s hardware. This maintains our foundational privacy promise and establishes Buddhi AI as the ideal choice for sensitive tasks.
  • Advanced Technologies: This breakthrough was made possible by using the LlamaIndex TypeScript library for robust RAG functionality and PGlite with pgvector for high-performance, on-device vector storage.

  ---

  Unified Messaging for Value Delivery

  This release directly addresses the main drawback of traditional server-based AI: the inherent privacy risk and data exposure.

  1. Confidentiality Guaranteed: Your Documents Never Leave Your Device. Analyze Private PDFs Instantly.
  2. Eliminate server-side data exposure: Buddhi AI’s new RAG feature allows you to interrogate sensitive internal documents (e.g., quarterly reports, client data) with zero risk of a third-party breach.
• 0.3.0

  Nov 25, 2025

  Introducing Image-to-Text: The Multimodal, Private AI

  

  We are excited to announce a monumental upgrade in Buddhi AI, moving beyond text to introduce multimodal capabilities with the launch of the Image-to-Text feature. This release is powered by a completely new, high-performance, and privacy-focused engine.

  🚀 Next-Generation Engine for Vision and Text

  This version involved a major technological transition to unlock true multimodal processing while maintaining our core privacy promise.

  We have migrated our core framework from the WebLLM library to the Google MediaPipe Library. This provides a robust, cross-platform architecture optimized for efficient on-device machine learning, replacing the previous engine for a more stable and powerful foundation.

  The second critical change is the integration of the Google Gemma 3n E2B Instruction Model. This state-of-the-art model is inherently multimodal, which is what makes our new vision features possible. This shift from a text-only model to this powerful new architecture enables us to deliver cutting-edge AI features right in your browser.

  ✨ New Feature Spotlight: Image-to-Text

  With the power of the new Gemma 3n E2B model, the unified Chat interface is now capable of processing visual data:

  • Visual Analysis: You can now upload or paste an image directly into the Chat interface.
  • Intelligent Queries: Ask the AI to analyze, describe, or summarize the content of the image. You can ask sophisticated questions like, “What is the primary object in this photo?” or “Write a caption for this landscape.”
  • Privacy-First Multimodality: This is a groundbreaking step for client-side AI. Just like your text prompts, your images and visual data never leave your device for processing, ensuring complete privacy even for complex visual tasks.

  🔒 Core Commitment: Privacy Reaffirmed

  Despite this massive increase in capability, our commitment to privacy remains absolute: All multimodal and text processing is executed entirely within your browser using your local hardware. Your data, prompts, and chat history remain private and secure on your local machine.

• 0.2.0

  Nov 16, 2025

  Go further with 0.2.0

  

  🌟 Core Focus: Uncompromised Performance & Unified Experience

  Version 0.2.0 represents a pivotal upgrade focused on delivering superior stability and consolidating our tools into a single, high-performance interface. We have reinforced our commitment to a 100% client-side experience, guaranteeing that privacy remains the foundation of Buddhi AI.

  ⚙️ Engine Upgrade: Transition to WebLLM

  The most critical change in this release is the migration of our core processing engine:

  Change	Old Engine	New Engine (V0.2.0)
  Technology	Experimental Chrome Built-in APIs	Robust WebLLM Library
  Benefit	Provided initial client-side proof-of-concept	Superior performance and stability across modern browsers.

  This transition to the robust WebLLM library ensures a more reliable and faster experience, solidifying our ability to run powerful AI models entirely on your device.

  💬 Unified Interface: The All-in-One Chat Hub

  We have consolidated our initial tools to deliver a streamlined and more powerful user experience:

  • Tools Removed: The standalone Summarizer and Writer tools have been retired.
  • New Consolidated Hub: All functionality has been integrated into a single, sophisticated Chat Interface.
  • Seamless Capabilities: The Chat interface now seamlessly handles all tasks:
    • Sophisticated Conversation
    • Advanced Summarization (by prompting the chat)
    • Writing Assistance (by prompting the chat)

  By focusing exclusively on this high-performance chat experience powered by WebLLM, Buddhi AI proves that you don’t have to sacrifice privacy for cutting-edge intelligence.

  🔒 Enhanced Privacy & Data Persistence

  We have maintained and improved upon our privacy guarantees:

  • Data Never Leaves Your Device: Your data and prompts remain entirely on your local machine.
  • Local History: Your Chat History is now securely saved locally using IndexedDB, ensuring your past conversations are private and accessible only to you.

  ---

  ✅ Requirements and Next Steps

  To experience the new performance gains, ensure your browser is up-to-date.

  We are confident that the move to WebLLM and the consolidated Chat interface will provide the fastest, most private, and most stable version of Buddhi AI yet.

• 0.1.0

  Nov 6, 2025

  Buddhi AI: Initial Launch Release Notes

  

  🚀 Introducing Buddhi AI: The Private, Browser-Based AI Suite

  We are thrilled to announce the initial launch of Buddhi AI, a project built to redefine how users interact with artificial intelligence by prioritizing privacy and performance.

  The core philosophy of Buddhi AI is simple: client-side processing. Unlike traditional server-side AI, Buddhi AI runs the model directly on your device and within your browser. This innovative approach ensures that your data never leaves your computer, making it inherently more secure and private.

  🛡️ Privacy-First Technology

  Buddhi AI is built upon the foundation of client-side AI, leveraging advanced capabilities to keep sensitive data local:

  • Data Stays Local: All processing occurs on the user’s machine. Your short-term context, sensitive documents, and queries are not logged on remote servers.
  • Core Technology: The platform utilizes Chrome’s Built-in AI APIs and supports local models like Gemini Nano for efficient on-device computation.
  • Reduced Costs: By shifting the computational heavy-lifting from our servers to the user’s device, we dramatically cut down on inference costs, allowing us to offer powerful tools affordably.

  ---

  🛠️ Initial Tool Collection (V0.1.0)

  For the initial launch, we focused on three essential, high-value use cases that perfectly leverage on-device AI capabilities:

  1. The Chat App: For quick queries, instant answers, and brainstorming sessions, ensuring casual questions are kept private.
  2. The Summarizer: Efficiently condenses long-form content—such as articles, meeting notes, or research papers—into key bullet points or short paragraphs instantly and privately.
  3. The Writer: Acts as a secure drafting assistant, helping users structure reports, draft emails, and improve existing text without compromising the security of their drafts.

  ---

  💡 Moving Forward

  This release marks the beginning of a larger vision. We believe that the best utility AI tools will run client-side, making AI faster, cheaper, and inherently more trustworthy. We plan to expand the Buddhi AI tool collection over the coming months to cover more high-value, client-side tasks, including translation and language detection.

  Note: Users may need to enable Chrome Built-in AI to access the full functionality of Buddhi AI.