Reading
KIYO is a multilingual voice chatbot built on Streamlit. It deploys large language models locally via Ollama, enabling real-time language translation, voice input/output, and supporting dyslexia-friendly mode and personalized conversation styles, with a commitment to fostering inclusive communication.
Section 01
KIYO is a multilingual voice chatbot built on Streamlit and deploying large language models locally via Ollama. It enables real-time language translation, voice input/output, and supports dyslexia-friendly mode and personalized conversation styles, dedicated to promoting inclusive communication and making AI conversations truly accessible.
Section 02
In today's increasingly globalized world, language barriers remain a significant obstacle to people's equal access to information and services. The United Nations Sustainable Development Goal 10 (SDG 10) explicitly calls for reducing inequality, and linguistic inclusion is one of the key links to achieving this goal. Traditional AI chat tools are mostly English-centric, leaving non-English users facing many inconveniences. Born in this context, the KIYO project is not only a technical demonstration but also a practical attempt to make AI technology benefit everyone, aiming to realize the vision of 'technology without borders'.
Section 03
KIYO is developed based on Python 3.8+, using Streamlit as the web framework to ensure a simple interface and easy deployment. Its core highlight is running local LLMs like Llama 3 via the Ollama framework—all inferences are done locally, which not only protects user privacy but also improves response speed. Multilingual capabilities are achieved through a three-stage process: 'translation-inference-back-translation'—automatically detecting the input language, translating it into English to submit to the local LLM, then translating the response back to the original language. For voice interaction, speech-to-text is based on the SpeechRecognition library, and text-to-speech uses the pyttsx3 engine, supporting natural conversations.
Section 04
KIYO focuses on inclusive design: it supports dyslexia-friendly mode (enabling Lexend font to improve readability), location-based language suggestions (currently for Indian states, scalable), custom conversation styles (formal/casual), and personality types. Additionally, the project attempts an RLHF feedback mechanism—generating two candidate responses for users to choose from, recording preferences for model optimization, which is a lightweight and effective feedback collection solution.
Section 05
To deploy KIYO, follow these steps: 1. Install Python 3.8+ and Ollama; 2. Clone the repository and create a virtual environment; 3. Install dependencies (streamlit, ollama, google-cloud-translate, etc.); 4. Configure Google Cloud service account: enable the Cloud Translation API, download the JSON key file, set the GOOGLE_APPLICATION_CREDENTIALS environment variable to point to the key; 5. Start the Ollama service and pull the Llama 3 model; 6. Run streamlit run app.py to launch the application.
Section 06
The value of KIYO lies in demonstrating the possibility of building powerful and inclusive AI applications through reasonable architecture and combination of open-source components. It provides a reference implementation for developers and an accessible AI assistant experience for end users. In the future, with the maturity of multilingual large models and the advancement of local inference technology, similar applications are expected to become more popular—language will no longer be a barrier to accessing AI services, and everyone can converse with intelligent technology in their own way.
Keep going with more reads from the same topic.
Nornir MCP Server is an enterprise-level server based on the Model Context Protocol (MCP). It seamlessly integrates large language models (such as Claude) with the Nornir network automation framework, supporting natural language orchestration for multi-vendor network devices (Cisco, Arista, Juniper, etc.), and providing production-grade features like a dual-engine architecture (NAPALM + Netmiko), intelligent filtering, and a secure sandbox.
Bibliothèque Française LLM is a structured indexing and annotation project for French public domain literature designed specifically for large language models (LLMs). It integrates multiple authoritative sources such as DraCor, Common Corpus, and Wikisource, providing metadata indexing categorized by genre, author, and era, as well as in-depth annotations for dramatic texts (including characters, lines, stage directions, etc.). Its aim is to enable LLMs to efficiently read and understand classic French literary works.
Splinter is a minimalist, high-performance key-value (KV) and vector storage system enabling zero-latency inter-process communication via shared memory and atomic operations. With only 766 lines of core code, it supports millions of operations per second and 768-dimensional vector storage, offering a new architectural approach for local LLM inference and data-intensive applications.
Folkering OS is the world's first AI-native bare-metal operating system, entirely written in Rust no_std without relying on Linux, POSIX, or libc. It can generate commands from scratch, compile them into WASM, and run them in 10 seconds, achieving true self-evolution.