Delta-9: An Integrated AI Assistant Platform Combining Multi-Agent Workflow and MPC

Delta-9 is an open-source integrated AI assistant platform that combines multi-agent workflow, MPC security mechanisms, and various mainstream AI services. It supports access to multiple platforms including Line, Gemini, Claude, and OpenAI, and is built using the Python/Node.js/React tech stack. Its core positioning is an AI capability aggregator, aiming to provide individuals and teams with a secure and efficient intelligent collaboration hub.

The vision of Delta-9 is to build an intelligent hub that integrates multiple AI capabilities, supports multi-platform access, and has enterprise-level security features—distinguishing itself from single-function chatbots or scenario-specific tools. In the project name, "Delta" symbolizes change and evolution, while "9" implies a design philosophy of approaching perfection. It aims to be an AI capability aggregator for individuals and teams, allowing users to call multiple AI services without switching between different applications, and ensuring data privacy through MPC.

Delta-9 adopts an end-to-end architecture: the backend is based on Python (for easy AI integration) and Node.js (for high-performance asynchronous processing), while the frontend uses React. The data layer employs a multi-database strategy (Redis for caching/message queues, SQLite for lightweight deployment, PostgreSQL as the production main database). It supports containerized deployment (Docker/Kubernetes), reducing operation and maintenance complexity, and enabling smooth scaling from a single machine to a cluster.

The core innovation of Delta-9 is its multi-agent workflow engine, which decomposes complex tasks into subtasks completed collaboratively by specialized agents (e.g., code generation, document processing). It supports multiple collaboration modes: serial (executed sequentially, with the output of the previous step as input for the next), parallel (multiple agents process subtasks simultaneously), and competitive (multiple agents generate candidates, and a judging agent selects the optimal one). Agents communicate via a standardized message protocol, and a visual workflow orchestration interface is provided.

Delta-9 introduces MPC (Secure Multi-Party Computation) to ensure data privacy: sensitive information is split into multiple shards and sent to different service providers. A single service cannot reconstruct the complete information; the final result is obtained only through the MPC protocol calculation, preventing data leakage and model poisoning. Its MPC implementation uses cryptographic verification protocols, supports additive homomorphism and comparison operations, and also provides a security audit function to record session metadata for compliance reviews.

Delta-9 supports multiple communication channels (Line, Slack, Tailscale) and AI services (Gemini, Claude, OpenAI, GitHub Copilot). It encapsulates different APIs into standardized interfaces through a unified abstraction layer, allowing free switching or combination of models. It also integrates enterprise tools like Google Workspace. Application scenarios include: personal daily assistant (schedule management, email composition), development teams (code review, document generation), enterprise customer service front desk, and workflow automation (e.g., email information extraction, code submission testing).

Delta-9 offers flexible deployment options: individual users can quickly start with Docker Compose, while enterprise users can achieve horizontal scaling and high availability using Kubernetes. Its modular design facilitates expansion—developers can use the SDK to build custom agents or integrate new services/channels. As an open-source project, it uses a permissive license, has an active community with contributions like plugin extensions and online seminars, and comprehensive documentation that lowers the barrier to participation.

In the future, Delta-9 will enhance multi-modal capabilities (processing images/audio/videos), optimize edge computing support, deepen integration with emerging AI technologies (multi-modal large models, embodied intelligence), and explore the combination of federated learning and MPC to enable cross-organizational collaborative model training. Summary: Delta-9 integrates scattered AI capabilities through architectural innovation, protects data with security technologies, and promotes evolution via an open ecosystem. It provides a solid foundation for individuals and teams to build intelligent workflows and is expected to become an important reference implementation in the open-source AI assistant field.