simple_newapi

GoAgent Multi-Agent Workflow Example

A generic multi-agent task processing system with DAG-based workflow orchestration.

Tech Stack and Components

Technologies Used

• Language: Go 1.26+
• LLM Provider: Ollama (llama3.2) or other OpenAI API-compatible services
• Configuration Format: YAML
• Workflow Orchestration: DAG (Directed Acyclic Graph)
• Template Engine: Go template syntax
• Concurrency Control: errgroup

Core Components Used

Workflow Step Configuration

Key Feature Implementations

Code Location References:

• DAG construction: internal/workflow/engine/executor.go:80-120
• Topological sort: internal/workflow/engine/executor.go:150-200
• Parallel execution: internal/workflow/engine/executor.go:250-300
• Template variable parsing: internal/workflow/engine/template.go:50-100
• Step dependency management: internal/workflow/engine/types.go:30-80
• Result aggregation: examples/simple_newapi/main.go:100-140

Quick Start

1. Configure your LLM

Edit config/server.yaml:

llm:
  provider: "ollama"
  base_url: "http://localhost:11434"
  model: "llama3.2"

2. Configure your Agents

Edit config/server.yaml:

agents:
  sub:
    - id: "agent-researcher-a"
      type: "top"
      name: "Researcher A"

3. Define Your Workflow

Edit config/workflow.yaml:

id: "multi-agent-workflow"
steps:
  - id: "analyze-input"
    name: "Analyze Input"
    agent_type: "top"
    input: "Analyze user requirements from: {{.input}}"

  - id: "research-topic-a"
    name: "Research Topic A"
    agent_type: "top"
    depends_on: ["analyze-input"]
    input: "Based on analysis from {{.analyze-input}}, research and compile findings"

4. Run

cd examples/simple_newapi
go run main.go

Workflow Orchestration

The system supports DAG-based workflow orchestration:

Parallel Execution

steps:
  - id: "step1"
    name: "First Step"
    agent_type: "top"

  - id: "step2"
    name: "Parallel Step 1"
    depends_on: ["step1"]

  - id: "step3"
    name: "Parallel Step 2"
    depends_on: ["step1"]  # Runs in parallel with step2

Serial Execution

steps:
  - id: "step1"
    name: "First Step"
    agent_type: "top"

  - id: "step2"
    name: "Second Step"
    depends_on: ["step1"]

  - id: "step3"
    name: "Third Step"
    depends_on: ["step2"]  # Runs after step2

Complex DAG

steps:
  - id: "analyze"
    name: "Analyze"
    agent_type: "leader"

  - id: "code"
    name: "Generate Code"
    depends_on: ["analyze"]
    agent_type: "code"

  - id: "test"
    name: "Generate Tests"
    depends_on: ["code"]
    agent_type: "test"

  - id: "docs"
    name: "Generate Docs"
    depends_on: ["analyze"]
    agent_type: "docs"

  - id: "review"
    name: "Review"
    depends_on: ["code", "docs"]  # Waits for both
    agent_type: "review"

Workflow Features

Step Configuration

Each step supports:

• id: Unique identifier
• name: Display name
• agent_type: Agent type to use
• input: Task description with template variables
• depends_on: List of step IDs this step depends on
• timeout: Execution timeout
• retry_policy: Retry configuration
• metadata: Additional metadata

Template Variables

Use {{.step_id}} to reference output from previous steps:

steps:
  - id: "analyze-input"
    name: "Analyze Input"
    agent_type: "top"
    input: "Analyze from: {{.input}}"

  - id: "process"
    name: "Process"
    depends_on: ["analyze-input"]
    input: "Process based on: {{.analyze-input}}"

Retry Policy

Configure retry behavior:

retry_policy:
  max_attempts: 3
  initial_delay: 1s
  max_delay: 5s
  backoff_multiplier: 2.0

How It Works

1. Load Configuration - Load agents and workflow from YAML
2. Build DAG - Create directed acyclic graph from step dependencies
3. Topological Sort - Determine execution order
4. Execute in Parallel - Run independent steps concurrently
5. Collect Results - Gather outputs from all steps

Example Output

=== GoAgent Multi-Agent Workflow Example ===

=== Configured Agents ===
  - agent-researcher-a (top): Researcher A
  - agent-researcher-b (bottom): Researcher B
  - agent-analyzer (top): Data Analyzer

=== User Query ===
Analyze the latest tech trends in AI and cloud computing...

=== Executing Workflow ===

=== Workflow Execution Result ===
Execution ID: exec-xxx
Status: completed
Duration: 45s
Total Steps: 5

=== Task Results ===

Analyze Input:
  Status: completed
  Duration: 5s
  Output: {"domains": ["AI", "cloud computing"], "priority": "high"}

Research Topic A:
  Status: completed
  Duration: 12s
  Output: {"items": [{"name": "LLM Advances", "reason": "Key trend in AI"}]}

Research Topic B:
  Status: completed
  Duration: 11s  # Ran in parallel with Research Topic A

Research Topic C:
  Status: completed
  Duration: 10s  # Ran in parallel with Research Topic A

Aggregate Results:
  Status: completed
  Duration: 7s
  Output: Comprehensive analysis report...

=== Final Output ===
{
  "report": {
    "summary": "...",
    "key_findings": ["..."],
    "priorities": ["..."]
  }
}

Next Steps

• Add more agents to your config
• Create complex workflows with multiple dependencies
• Use retry policies for robustness
• Add metadata for tracking and debugging