TESS (Text Evaluation & Synthesis System)

TESS is a comprehensive local AI workspace built on top of Ollama. It provides a powerful, unified interface for managing, testing, and interacting with your local large language models.

Key Features

• Chat: A robust chat interface with history, model selection, parameter tuning, and dynamic context injection.
• Long-Term Memory: Persistent, tool-based memory system that allows models to remember user preferences, facts, and context across different conversations.
• Arena: Compare models side-by-side to evaluate performance and reasoning.
• Batch: Run prompts across multiple models simultaneously to compare outputs.
• Personas & System Variables: Manage custom system prompts (personas) and define custom system variables (e.g., loaded dynamically from local text files via @file(path)) for prompt templating.
• Python Workspace: An interactive local Python IDE to write, execute, and stop code, run shell commands, and export scripts directly into custom AI tools.
• Story Studio: High-fidelity, multi-speaker audio synthesis with voice cloning and dynamic character identification, using Omnivoice and Kokoro TTS.
• Voice Designer: Craft custom synthetic voices by adjusting parameters like gender, age, pitch, and accent.
• Visual Generation & Photopea Editor: Create stunning images using the Anima pipeline and edit them directly in the browser using an integrated Photopea workspace with layers support and save-back capabilities.
• Tools & Agents:
  • AI Tool Generator: Build custom tools using natural language; the system generates the schema and logic for you.
  • Integrated Debugger: Test and validate tools in a split-screen workspace before deployment.
• Web Search: Equip your local models with real-time web access via integrated DuckDuckGo search and URL extraction.
• Google Integration: Connect your Google Workspace to analyze and synthesize documents.
• Apps Ecosystem: A modular space for custom applications, including a dedicated Notes app with Google Drive synchronization and Routineer (a routine/habit tracker with calendar stats and interactive badges).
• Model Management: Easily pull, delete, and manage your local Ollama models, and create new model variants (Modelfiles) directly within the UI.
• GPU & VRAM Monitoring: Real-time VRAM usage and GPU activity monitoring in the header with one-click unload of all loaded models.

Getting Started

Prerequisites

1. Ollama: Install and run Ollama.
2. uv: Install uv, the fast Python package installer and manager:
   • Windows (PowerShell):

     powershell -c "irm https://astral.sh/uv/install.ps1 | iex"

   • macOS / Linux:

     curl -LsSf https://astral.sh/uv/install.sh | sh

3. Local LLM: Pull the default chat LLM in Ollama:

   ollama pull hf.co/unsloth/gemma-4-E4B-it-GGUF:Q4_K_M

4. Hardware (Recommended): An NVIDIA GPU with CUDA-compatible drivers (with at least 8GB VRAM required for running the Anima image generation model) is highly recommended for visual generation and voice synthesis (Kokoro/OmniVoice).

Installation & Launch

1. Clone the repository and enter the directory:

   git clone https://github.com/aole/TESS.git
   cd TESS

2. Run the application:
   • Windows: Run run.bat (which updates via git, syncs dependencies, and starts the server):

     .\run.bat

   • macOS / Linux: Run the main script with uv:

     uv run main.py

3. Open your browser to http://localhost:8080.

Google Workspace Integration (Optional)

To enable Google integration features (Gmail, YouTube, Drive indexing, and Google Drive Notes synchronization):

1. Go to the Google Cloud Console.
2. Create a project and enable the Gmail API, YouTube Data API v3, and Google Drive API.
3. Configure the OAuth Consent Screen and create credentials for an OAuth 2.0 Client ID (select Desktop app as the application type).
4. Download the JSON client secret, rename it to client_secret.json, and place it in the root of the TESS folder (see client_secret.json.example for reference).

Technology

Built with ❤️ using:

• NiceGUI - For the beautiful, responsive web interface.
• Ollama - For local LLM inference.
• uv - Fast Python package and project management.
• Omnivoice & Kokoro - For state-of-the-art TTS and voice cloning.