Configuration Reference

Complete configuration guide for PraisonAI agents, tasks, memory, and more

Configuration Reference

This section provides comprehensive documentation for all configuration options in PraisonAI. Each component has its own detailed configuration guide with examples, best practices, and advanced options.

Configuration Categories

Agent Configuration

Complete guide to agent parameters including max_iter, max_retry_limit, context_length, and markdown options

Task Configuration

Task parameters including task_type, condition, next_tasks, and is_start options

Memory Configuration

Memory system configuration including graph store, quality scores, and embedder options

LLM Configuration

LLM settings including retry logic, timeouts, and custom headers

Tool Configuration

Tool timeout settings and performance tuning options

Handoff Configuration

Handoff filters and advanced delegation settings

Guardrail Configuration

Custom validation rules and guardrail settings

Best Practices

Configuration best practices and common patterns

CLI Configuration

Layered, project-aware defaults for the praisonai CLI (~/.praisonai/config.yaml)

Single-Source Config

Now also the single source of truth for mcp servers and permissions policy — model, MCP, and permissions in one file.

CLI config (~/.praisonai/config.yaml, project .praisonai/config.yaml) holds runtime defaults for the praisonai CLI itself — model, output, MCP servers, rules. See CLI Configuration and praisonai config.

Agents YAML (agents.yaml or similar) is the SDK-level definition of agents, tasks, and workflows. See Agent Configuration and Task Configuration.

See also: Centrally-Managed Config — distribute defaults and enforce policy (permissions, allowed models) across every developer machine from one source of truth.

Quick Reference

Essential Configuration Parameters

Component	Parameter	Type	Description
Agent	max_iter	int	Maximum iterations for task completion
Agent	max_retry_limit	int	Max retries for retryable tool and guardrail failures (1 + limit total attempts)
Agent	retry_initial_delay	float	First retry delay in seconds (exponential backoff)
Agent	retry_backoff_factor	float	Backoff multiplier per retry attempt
Agent	retry_jitter	float	Jitter fraction added to each retry delay
Task	task_type	str	Type of task execution
Memory	quality_threshold	float	Minimum quality score for memory retrieval
LLM	timeout	int	Request timeout in seconds
Tool	execution_timeout	int	Tool execution timeout

Environment Variables

PraisonAI supports configuration through environment variables for sensitive settings:

# LLM Configuration
export OPENAI_API_KEY="your-api-key"
export OPENAI_MODEL="gpt-4o"

# Memory Configuration
export PRAISONAI_MEMORY_PROVIDER="rag"
export PRAISONAI_DB_PATH="/path/to/db"

# Performance Settings
export PRAISONAI_MAX_WORKERS=4
export PRAISONAI_TIMEOUT=300

Configuration Files

PraisonAI supports YAML configuration files for complex setups:

# config.yaml
agents:
  default:
    max_iter: 15
    max_retry_limit: 3
    context_length: 128000
    markdown: true

memory:
  provider: rag
  quality_threshold: 0.7
  graph_enabled: true

llm:
  timeout: 60
  max_retries: 3
  temperature: 0.7

Getting Started

1. Start with the Agent Configuration to set up your agents
2. Configure Tasks for your workflow
3. Set up Memory for persistent storage
4. Fine-tune LLM settings for optimal performance

Need Help?

• Check our Best Practices guide
• See Examples for real-world configurations
• Visit our GitHub for issues and discussions