Tag: llm-integration
Authoring Builtin Tools for Agent Runtimes
This guide details the architecture for creating agent-callable built-in tools, requiring the definition of a manifest, executor, and execution runtime. Developers must manage five distinct faces—Manifest, Executor, ExecutionRuntime, Client…
Guide for integrating new large language models
This skill provides a comprehensive, step-by-step guide for developers to integrate support for a novel LLM provider into the Dust framework. It details necessary modifications across configuration, pricing, central registries, and testing …
Debugging heterogeneous agent pipelines and integrations
This skill provides comprehensive guidance for developers debugging complex, multi-source agent pipelines, including integrations for Claude Code, Codex, and external CLI agents. It covers adapter event mapping, transport layer issues, and …
Framework for Building Agent Builtin Tools
This comprehensive framework enables developers to create structured, callable APIs (builtin tools) for an agent runtime. It mandates defining manifests, executors, and various UI surfaces to integrate custom, robust logic seamlessly.
Workflow for adding new AI provider documentation
This skill provides a comprehensive workflow for integrating a new AI provider, covering the creation of usage documentation (English and Chinese), updating environment variable definitions, modifying Docker configurations, and preparing al…
Setup mcp-rubber-duck for AI coding tools
This skill automates the integration of the mcp-rubber-duck MCP server across various AI coding environments, including VS Code, Cursor, and Claude. It manages platform-specific configuration files and environment variables, ensuring non-de…
Secure Proxy for Managed API Credentials
This proxy routes all external API requests through a secure local endpoint, preventing the agent from ever handling raw API keys. It manages credentials and enforces domain restrictions, ensuring all interactions are audited and securely r…
Client-side AI Model Integration for Web Apps
Integrate large language models directly into frontend web applications using the CloudBase JavaScript SDK. It supports text generation and streaming for models such as DeepSeek, Hunyuan, and Kimi without needing a Node.js proxy.