A single, powerful desktop application to discover, install, manage, and interact with any AI tool, model, or script.
The world of open-source AI is exploding with innovation. Every day, new models, agentic frameworks, and powerful scripts are released on GitHub and Hugging Face. But this rapid progress has created a chaotic and fragmented ecosystem for the end-user.
- Dependency Hell: Trying to run two different AI tools often leads to a nightmare of conflicting Python versions, CUDA driver incompatibilities, and broken dependencies.
- Cryptic Installations: READMEs are often incomplete, assuming expert-level knowledge of command-line tools, virtual environments, and manual configuration file editing.
- Discovery Overload: Finding the right tool for a specific task is a full-time job of monitoring Twitter, Discord, and GitHub.
- Credential Sprawl: Every tool and service requires its own API key and configuration, leading to insecure practices like pasting secrets into multiple config files scattered across your hard drive.
The result? The most powerful technology of our time is often inaccessible to those who are not expert developers.
SakeAI's mission is to democratize access to AI by abstracting away the complexity of installation, configuration, and management.
We believe that installing and running a complex tool like RVC or a standalone Windows application should be as simple as installing an app from an app store. SakeAI is a single, robust desktop application that serves as the command center for your entire AI workflow.
The core of SakeAI is a sophisticated installation agent that can reliably install almost any tool from GitHub, whether it's a Python project or a pre-compiled binary.
- Manifest-Driven Installation: We define the
SakeAI.jsonstandard, a simple metadata file that provides SakeAI with a perfect, step-by-step recipe for installation. - LLM-Powered Heuristic Engine: For any GitHub repo without a manifest, the agent uses a Large Language Model (LLM) to intelligently scan the repository's README, code, and release assets. It then generates a probable installation plan for user approval, effectively creating a
SakeAI.jsonon the fly. - Human-in-the-Loop: For installers that require graphical interaction, the agent will pause and prompt you for input, seamlessly blending automated and manual steps.
- User-Choice LLM Integration: You control how the agent thinks. In the Settings panel, you can choose from:
- Cloud API: Use your OpenAI, Anthropic, or other cloud provider API key.
- Local API: Connect to a local LLM server like Ollama or LM Studio for privacy and cost savings.
- Manual Web Chat: SakeAI will generate a perfectly crafted prompt for you to paste into a web service like ChatGPT or Claude. You then paste the resulting plan back into SakeAI to proceed.
- System Validation: Before any installation, SakeAI checks your hardware (GPU, VRAM, RAM) and software (OS, NVIDIA drivers) against the tool's requirements to prevent failed installations.
- Isolated Environments: Every Python-based tool is installed into its own sandboxed Conda environment, guaranteeing zero dependency conflicts.
Stop scouring the web. The Discover page brings the tools to you.
- Dynamic HF Feed: An automatically updated feed of the most popular and trending tools from Hugging Face Spaces.
- Curated Catalog: Browse a list of vetted and pre-configured tools from a local
sakeai_catalog.jsonfile, ensuring high-quality, one-click installations. - Add Your Own Tools: Use the Add a Tool page to point SakeAI at any GitHub repository. The Heuristic Engine will generate an installation configuration, which you can then review and add to your personal catalog.
Manage all your installed applications from one clean interface.
- Unified View: See every tool you've installed, its status (installed, failed), and its installation path.
- One-Click Launch: Run any installed tool directly from the SakeAI UI.
- Easy Management: Uninstall tools and clean up their environments with a single click.
graph TD
A[User provides GitHub URL] --> B{Check for SakeAI.json manifest};
B -- Yes --> C[Parse Manifest];
C --> D{Validate System Requirements};
D -- Pass --> E[Execute Manifest Steps];
E --> S{Step requires user input?};
S -- No --> T[Execute Step Automatically];
T --> U{More steps?};
U -- Yes --> E;
S -- Yes --> V[Pause and Prompt User];
V --> W[User provides input];
W --> T;
U -- No --> F[Installation Success];
D -- Fail --> G[Error: System does not meet requirements];
B -- No --> H[Heuristic Analysis];
H --> I[Scan Repo, Releases & Prepare Context];
I --> J{Query LLM via User's Chosen Method};
J --> K[LLM Generates Plan];
K --> O{Present Plan for Approval};
O -- Approve --> P[Execute Generated Plan];
P --> E;
O -- Deny --> Q[Installation Cancelled];