Overview

VeADK Go is an agent development framework for Go developers. It builds on ADK Go agents, runners, sessions, and event streams, then adds Volcengine model configuration, tools, memory, knowledge bases, observability, and app wrappers so Go applications can embed or serve agent capabilities. Preview documentation follows the veadk-go main branch. The current mainline module path is github.com/volcengine/veadk-go/v2 for new projects. github.com/volcengine/veadk-go V1 is a maintenance line for existing projects that cannot migrate yet.

Use Cases

Build Go agents that call models, tools, knowledge bases, or memory.
Invoke agents directly from server-side Go code through Runner, while controlling users, sessions, and event streams.
Compose sequential, parallel, or loop execution with Workflow Agents.
Package agents as Simple Apps, A2A Apps, or AgentKit Server Apps.
Add observability with OpenTelemetry, APMPlus, Cozeloop, and related exporters.

Version Lines

Version line	Module path	ADK dependency	Branch	Use case
V2	`github.com/volcengine/veadk-go/v2`	`google.golang.org/adk/v2`	`main`	Default for new projects. Preview documentation targets this line.
V1	`github.com/volcengine/veadk-go`	`google.golang.org/adk`	`release/v1`	Maintenance for existing V1 projects, mainly for compatibility, security, and critical fixes.

Preview documentation shows current V2 capabilities. Production projects should pin a published version or a reviewed commit.

Core Modules

Module	Purpose
`agent/llmagent`	Creates a default LLM agent with model, prompt, tool, skill, knowledge base, and long-term memory wiring.
`runner`	Runs agents from Go code, managing app name, user, session, event stream, short-term session, and long-term memory.
`agent/workflowagents`	Creates sequential, parallel, and loop workflow agents.
`tool/builtin_tools`	Provides built-in tools for Web Search, Web Fetch, Link Reader, MCP, image, video, speech, code execution, and Volcengine services.
`memory`	Provides short-term session and long-term memory wrappers with local and external storage backends.
`knowledgebase`	Builds knowledge bases and exposes retrieval to agents through tools.
`observability`	Integrates OpenTelemetry and exports traces or metrics to stdout, files, APMPlus, Cozeloop, TLS, or OTLP endpoints.
`apps`	Wraps agents as HTTP apps, A2A apps, or AgentKit Server apps.

Usage Paths

Goal	Documentation
Install and configure VeADK Go	Installation
Run your first agent	Quickstart
Review mainline changes	Changelog

Get Started

Core Components

Observability

Deploy

Overview

Use Cases

Version Lines

Core Modules

Usage Paths

​Use Cases

​Version Lines

​Core Modules

​Usage Paths

Use Cases

Version Lines

Core Modules

Usage Paths