docmd-assistant is a universal, UI-agnostic headless AI Assistant engine designed for software documentation tools, web applications, and developer platforms. Powered directly by the aiplug multi-provider runtime, docmd-assistant enables developers to embed sophisticated conversational AI capabilities without locking themselves into a specific user interface or cloud platform.
docmd-assistant operates as a pure logic engine. It manages message history, system prompt reinforcement, multi-provider connections, and client tool execution pipelines while leaving user interface rendering entirely to your application.
Key Features
1. Multi-Provider & Multi-Model
Connect seamlessly to leading artificial intelligence providers including OpenAI, Anthropic, Gemini, DeepSeek, Groq, MiniMax, and Ollama. Easily switch models or providers dynamically at runtime.
2. Extensible Tool Execution
Register custom client-side or server-side tools with structured schema definitions. When requested by the language model, docmd-assistant executes registered tool handlers automatically and appends results to the context stream.
3. Headless & Framework Agnostic
Designed to work across React, Vue, Svelte, Angular, Vanilla JavaScript, Node.js backend services, CLI utilities, and desktop applications.
4. Dual Connection Modes
Operate in Direct Mode using local provider API keys via aiplug, or in Relay Mode by routing requests securely through Cloud AI Relays.
5. Event-Driven Architecture
Subscribe to real-time engine events for message updates, tool call invocations, execution results, and error notifications.
Ecosystem Integration
docmd-assistant functions as an independent library. It also powers conversational AI capabilities across the docmd ecosystem:
| Package | Purpose |
|---|---|
docmd-assistant |
Universal headless AI engine and tool runtime |
@docmd/plugin-ai |
Official docmd plugin providing embedded chat drawers and documentation search integration |
docmd-search |
Offline semantic search engine for documentation indexing |
Next Steps
- Getting Started: Install the library and run your first conversational turn.
- Engine Architecture: Learn how the turn execution pipeline and event bus operate.
- Tool System: Create and register custom client-side tools.
- Programmatic API: Explore full class definitions and method signatures.