Config Module

Feature configuration classes for agent-centric API

Module praisonaiagents.config

The config module provides dataclass-based configuration for all agent features. These classes offer type safety, sensible defaults, and IDE autocompletion.

Quick Start

python

from praisonaiagents import (
    MemoryConfig,
    OutputConfig,
    ExecutionConfig,
    PlanningConfig,
)

Or use the namespace style:

python

import praisonaiagents as pa

config = pa.config.MemoryConfig(user_id="user123")

Configuration Classes

MemoryConfig

Memory and session management

KnowledgeConfig

RAG and knowledge retrieval

OutputConfig

Output formatting and verbosity

ExecutionConfig

Agent execution parameters

PlanningConfig

Planning mode settings

ReflectionConfig

Self-reflection settings

GuardrailConfig

Safety and validation

WebConfig

Web search configuration

MemoryConfig

Configuration for agent memory and session management.

python

from praisonaiagents import MemoryConfig, MemoryBackend

# Simple usage
config = MemoryConfig(user_id="user123", session_id="session456")

# With backend selection
config = MemoryConfig(
    backend=MemoryBackend.SQLITE,
    user_id="user123",
    session_id="session456",
)

# Usage in Agent
from praisonaiagents import Agent
agent = Agent(
    instructions="...",
    memory=config,
)

Parameters

Parameter	Type	Default	Description
`backend`	`MemoryBackend`	`FILE`	Storage backend (FILE, SQLITE, CHROMA, MEM0, MONGODB, DAKERA, IN_MEMORY)
`user_id`	`str`	`None`	User identifier for memory isolation
`session_id`	`str`	`None`	Session identifier
`auto_memory`	`bool`	`False`	Enable automatic memory extraction
`claude_memory`	`bool`	`False`	Use Claude memory format
`learn`	`LearnConfig`	`None`	Continuous learning configuration

OutputConfig

Configuration for output formatting and verbosity.

python

from praisonaiagents import OutputConfig

# Verbose output with markdown
config = OutputConfig(
    output="verbose",
    markdown=True,
    show_reasoning=True,
)

# Using presets (simpler)
from praisonaiagents import Agent
agent = Agent(instructions="...", output="verbose")  # Preset string

Parameters

Parameter	Type	Default	Description
`verbose`	`bool`	`False`	Enable verbose output
`markdown`	`bool`	`True`	Format output as markdown
`show_reasoning`	`bool`	`False`	Show agent reasoning steps
`show_tool_calls`	`bool`	`True`	Display tool call details
`streaming`	`bool`	`False`	Enable streaming output

Presets

Preset	Description
`"silent"`	Minimal output
`"minimal"`	Basic output only
`"actions"`	Show actions (default)
`"verbose"`	Detailed output

ExecutionConfig

Configuration for agent execution parameters.

python

from praisonaiagents import ExecutionConfig

config = ExecutionConfig(
    max_iter=20,
    max_retry_limit=5,
    timeout=300,
)

# Using presets
from praisonaiagents import Agent
agent = Agent(instructions="...", execution="thorough")

Parameters

Parameter	Type	Default	Description
`max_iter`	`int`	`20`	Maximum iterations
`max_retry_limit`	`int`	`2`	Max retries for retryable tool and guardrail failures
`retry_initial_delay`	`float`	`1.0`	First retry delay (seconds)
`retry_backoff_factor`	`float`	`2.0`	Exponential backoff multiplier
`retry_jitter`	`float`	`0.1`	Jitter fraction of base delay
`timeout`	`int`	`None`	Execution timeout in seconds
`code_execution`	`bool`	`False`	Enable code execution
`code_mode`	`str`	`"safe"`	Code execution mode (“safe” or “unsafe”)
`code_tools`	`bool`	`False`	When `True`, model-generated code may call the agent’s registered tools via injected proxies
`code_tools_allow`	`Optional[List[str]]`	`None`	Explicit per-run allow-list of tool names callable from code. `None`/empty = no tools exposed (safe default)
`rate_limiter`	`Any`	`None`	Rate limiter instance
`allow_code_execution`	`bool`	`False`	⚠️ Deprecated — use `code_execution`

Presets

Preset	Description
`"fast"`	Quick execution (max_iter=5)
`"balanced"`	Balanced (max_iter=20)
`"thorough"`	Thorough (max_iter=25)

KnowledgeConfig

Configuration for RAG and knowledge retrieval.

python

from praisonaiagents import KnowledgeConfig

config = KnowledgeConfig(
    sources=["docs/", "data.pdf"],
    retrieval_k=10,
    rerank=True,
)

Parameters

Parameter	Type	Default	Description
`sources`	`List[str]`	`[]`	Knowledge source paths
`retrieval_k`	`int`	`5`	Number of chunks to retrieve
`retrieval_threshold`	`float`	`0.0`	Minimum similarity threshold
`rerank`	`bool`	`False`	Enable reranking
`auto_retrieve`	`bool`	`True`	Automatic context retrieval

PlanningConfig

Configuration for planning mode.

python

from praisonaiagents import PlanningConfig

config = PlanningConfig(
    llm="gpt-4o",
    auto_approve=False,
    read_only=True,
)

Parameters

Parameter	Type	Default	Description
`llm`	`str`	`None`	LLM for planning (defaults to agent’s LLM)
`tools`	`List`	`None`	Tools available for planning
`reasoning`	`bool`	`False`	Enable reasoning in plans
`auto_approve`	`bool`	`False`	Auto-approve plans
`read_only`	`bool`	`False`	Only read-only operations

ReflectionConfig

Configuration for self-reflection.

python

from praisonaiagents import ReflectionConfig

config = ReflectionConfig(
    min_iterations=1,
    max_iterations=5,
    prompt="Evaluate your response for accuracy...",
)

Parameters

Parameter	Type	Default	Description
`min_iterations`	`int`	`1`	Minimum reflection iterations
`max_iterations`	`int`	`3`	Maximum reflection iterations
`llm`	`str`	`None`	LLM for reflection
`prompt`	`str`	`None`	Custom reflection prompt

GuardrailConfig

Configuration for guardrails and safety validation.

python

from praisonaiagents import GuardrailConfig, GuardrailAction

config = GuardrailConfig(
    validator=my_validator_fn,
    max_retries=5,
    on_fail=GuardrailAction.RAISE,
)

Parameters

Parameter	Type	Default	Description
`validator`	`Callable`	`None`	Validation function
`description`	`str`	`None`	Guardrail description for LLM-based validation
`max_retries`	`int`	`3`	Maximum retries on failure
`on_fail`	`GuardrailAction`	`RETRY`	Action on failure (RETRY, SKIP, RAISE)
`policies`	`List[str]`	`[]`	Policy strings

WebConfig

Configuration for web search capabilities.

python

from praisonaiagents import WebConfig, WebSearchProvider

config = WebConfig(
    provider=WebSearchProvider.TAVILY,
    max_results=10,
)

Parameters

Parameter	Type	Default	Description
`provider`	`WebSearchProvider`	`DUCKDUCKGO`	Search provider
`max_results`	`int`	`5`	Maximum search results
`fetch_content`	`bool`	`True`	Fetch page content

Bool/String/Config Pattern

All config classes follow the “progressive disclosure” pattern:

python

from praisonaiagents import Agent

# Level 1: Boolean toggle (simplest)
Agent(memory=True)

# Level 2: String preset
Agent(output="verbose")

# Level 3: Config object (full control)
Agent(memory=MemoryConfig(user_id="user123", backend=MemoryBackend.SQLITE))

This pattern applies to: memory, knowledge, planning, reflection, guardrails, web, output, execution, caching, hooks, skills, autonomy.

Config Module

Module praisonaiagents.config

Quick Start

Configuration Classes

MemoryConfig

Parameters

OutputConfig

Parameters

Presets

ExecutionConfig

Parameters

Presets

KnowledgeConfig

Parameters

PlanningConfig

Parameters

ReflectionConfig

Parameters

GuardrailConfig

Parameters

WebConfig

Parameters

Bool/String/Config Pattern

See Also