Project Templates Guide

Navigation: Documentation Home → Guides → Project Templates

Learn how to use built-in templates and create custom project templates for AgenticGoKit projects.

Overview

AgenticGoKit's template system allows you to quickly scaffold projects with pre-configured settings for common use cases. Templates can be built-in (provided with AgenticGoKit) or custom (defined by you or your team).

Quick Start

Using Built-in Templates

# Create a research assistant with web search capabilities
agentcli create research-bot --template research-assistant

# Create a RAG-powered knowledge base
agentcli create knowledge-base --template rag-system

# Create a data processing pipeline
agentcli create data-pipeline --template data-pipeline

# List all available templates
agentcli template list

Creating Custom Templates

# Create a new template file
agentcli template create my-company-standard

# Edit the generated template file
# .agenticgokit/templates/my-company-standard.yaml

# Use your custom template
agentcli create my-project --template my-company-standard

Built-in Templates

basic

Simple Multi-Agent System

• 2 agents with sequential orchestration
• Basic error handling and responsible AI
• Good starting point for learning

agentcli create my-project --template basic

research-assistant

Multi-Agent Research System

• 3 collaborative agents (researcher, analyzer, synthesizer)
• Web search and summarization tools via MCP
• Ideal for information gathering and analysis

agentcli create research-bot --template research-assistant

rag-system

RAG Knowledge Base

• 3 collaborative agents for document processing
• PostgreSQL with vector search (pgvector)
• OpenAI embeddings and RAG capabilities
• Perfect for Q&A systems and knowledge bases

agentcli create knowledge-base --template rag-system

data-pipeline

Sequential Data Processing

• 4 agents in sequential pipeline (ingester, processor, validator, outputter)
• Error handling and workflow visualization
• Great for ETL and data transformation workflows

agentcli create data-flow --template data-pipeline

chat-system

Conversational System

• 2 agents with route-based orchestration
• Session-based memory for multi-user scenarios
• In-memory storage for fast responses
• Ideal for chatbots and conversational interfaces

agentcli create chat-bot --template chat-system

Template Customization

You can override template defaults with additional flags:

# Use research-assistant template but with 5 agents and production MCP
agentcli create my-research --template research-assistant --agents 5 --mcp production

# Use rag-system template but with Weaviate instead of PostgreSQL
agentcli create my-kb --template rag-system --memory weaviate

# Use basic template but add memory and visualization
agentcli create my-project --template basic --memory pgvector --visualize

Custom Templates

Template Locations

Custom templates are searched in the following locations (in priority order):

1. Current directory: .agenticgokit/templates/
2. User home: ~/.agenticgokit/templates/
3. System-wide:
   • Unix/Linux/macOS: /etc/agenticgokit/templates/
   • Windows: %PROGRAMDATA%/AgenticGoKit/templates/

Template Format

Templates can be defined in JSON or YAML format:

YAML Example

name: "E-commerce Assistant"
description: "Multi-agent system for e-commerce operations"
features:
  - "product-search"
  - "inventory-management"
  - "customer-support"

config:
  numAgents: 4
  provider: "openai"
  orchestrationMode: "collaborative"
  collaborativeAgents:
    - "product-searcher"
    - "inventory-manager"
    - "customer-support"
    - "order-processor"
  
  # Memory configuration
  memoryEnabled: true
  memoryProvider: "pgvector"
  embeddingProvider: "openai"
  
  # MCP configuration
  mcpEnabled: true
  mcpTools:
    - "product_catalog"
    - "inventory_api"
    - "customer_db"
  
  # Other options
  responsibleAI: true
  errorHandler: true
  visualize: true

JSON Example

{
  "name": "Trading Bot System",
  "description": "Multi-agent trading system with market analysis",
  "features": [
    "market-analysis",
    "risk-management",
    "automated-trading"
  ],
  "config": {
    "numAgents": 5,
    "provider": "openai",
    "orchestrationMode": "collaborative",
    "collaborativeAgents": [
      "market-analyzer",
      "sentiment-analyzer",
      "risk-manager",
      "strategy-executor",
      "portfolio-manager"
    ],
    "memoryEnabled": true,
    "memoryProvider": "pgvector",
    "ragEnabled": true,
    "mcpEnabled": true,
    "mcpProduction": true,
    "responsibleAI": true,
    "errorHandler": true
  }
}

Configuration Options

Basic Configuration

• numAgents (int): Number of agents to create
• provider (string): LLM provider ("openai", "azure", "ollama", "mock")
• orchestrationMode (string): Agent coordination ("sequential", "collaborative", "loop", "route")

Agent Configuration

• collaborativeAgents ([]string): Agent names for collaborative mode
• sequentialAgents ([]string): Agent names for sequential mode
• loopAgent (string): Agent name for loop mode

Memory & RAG Configuration

• memoryEnabled (bool): Enable memory system
• memoryProvider (string): Memory provider ("memory", "pgvector", "weaviate")
• embeddingProvider (string): Embedding provider ("openai", "ollama", "dummy")
• embeddingModel (string): Specific embedding model
• ragEnabled (bool): Enable RAG functionality
• ragChunkSize (int): Document chunk size in tokens
• ragOverlap (int): Overlap between chunks
• ragTopK (int): Number of results to retrieve
• ragScoreThreshold (float): Minimum similarity score
• hybridSearch (bool): Enable hybrid search
• sessionMemory (bool): Enable session-based memory isolation

MCP Configuration

• mcpEnabled (bool): Enable MCP tool integration
• mcpProduction (bool): Enable production MCP features
• mcpTools ([]string): List of MCP tools to include
• withCache (bool): Enable MCP result caching
• withMetrics (bool): Enable Prometheus metrics

Other Options

• responsibleAI (bool): Include responsible AI features
• errorHandler (bool): Include error handling
• visualize (bool): Generate workflow diagrams

Template Management

Creating Templates

# Create a YAML template (default)
agentcli template create my-template

# Create a JSON template
agentcli template create my-template --format json

# Create template in specific location
agentcli template create my-template --output /path/to/template.yaml

Validating Templates

# Validate template syntax and configuration
agentcli template validate my-template.yaml

# Example output:
# Template validation successful!
# Name: My Custom Template
# Description: A custom project template
# Features: [custom-feature, example]
# Agents: 3
# Provider: openai
# Memory: pgvector
# RAG: enabled (chunk size: 1000)

Listing Templates

# List all available templates
agentcli template list

# Show template search paths
agentcli template paths

Common Use Cases

Company Standard Template

Create a template that matches your organization's standards:

name: "Company Standard"
description: "Standard multi-agent setup for our company"
config:
  numAgents: 3
  provider: "azure"  # Company uses Azure
  memoryEnabled: true
  memoryProvider: "pgvector"  # Company standard
  embeddingProvider: "openai"
  mcpEnabled: true
  mcpProduction: true
  responsibleAI: true
  errorHandler: true

Environment-Specific Templates

Different templates for different environments:

# development-template.yaml
name: "Development Setup"
description: "Fast setup for development"
config:
  numAgents: 2
  provider: "mock"  # Fast for testing
  memoryProvider: "memory"  # In-memory for speed
  mcpEnabled: false  # Simplified for dev

# production-template.yaml
name: "Production Setup"
description: "Production-ready configuration"
config:
  numAgents: 5
  provider: "openai"
  memoryEnabled: true
  memoryProvider: "pgvector"
  ragEnabled: true
  mcpEnabled: true
  mcpProduction: true
  withCache: true
  withMetrics: true

Domain-Specific Templates

Templates for specific industries or use cases:

name: "Healthcare Assistant"
description: "HIPAA-compliant healthcare agent system"
config:
  numAgents: 3
  provider: "azure"  # For compliance
  orchestrationMode: "sequential"
  memoryEnabled: true
  memoryProvider: "pgvector"
  ragEnabled: true
  mcpEnabled: true
  mcpTools:
    - "medical_database"
    - "patient_records"
  responsibleAI: true  # Critical for healthcare
  errorHandler: true

Best Practices

Template Design

1. Use descriptive names: Choose template names that clearly indicate their purpose
2. Include comprehensive features list: Help users understand what the template provides
3. Set sensible defaults: Configure options that work well together
4. Document your templates: Include clear descriptions and use cases

Template Organization

1. Version your templates: Consider including version information in descriptions
2. Share templates: Place commonly used templates in shared locations for team use
3. Validate templates: Always test your templates before sharing
4. Use consistent naming: Follow naming conventions for your organization

Template Usage

1. Start with templates: Use templates as starting points, then customize as needed
2. Override when necessary: Use flags to override template defaults for specific needs
3. Create project-specific templates: For recurring project patterns, create custom templates

Troubleshooting

Template Not Found

If your template isn't being found:

1. Check template search paths: agentcli template paths
2. Ensure the file has the correct extension (.json, .yaml, or .yml)
3. Validate the template syntax: agentcli template validate your-template.yaml

Template Validation Errors

Common validation errors:

• Invalid JSON/YAML syntax: Check for missing commas, quotes, or indentation
• Missing required fields: Ensure name, description, and config are present
• Invalid configuration values: Check that provider names, orchestration modes, etc. are valid

Template Override Issues

If template overrides aren't working:

• External templates take priority over built-in templates with the same name
• Check that the template is in the correct search path
• Use agentcli template list to see which templates are loaded

Related Guides

• CLI Reference - Complete CLI command reference
• Configuration - Project configuration options
• Getting Started - Basic project creation
• Best Practices - Development best practices