OSSS.ai.workflows.prompt_composer

OSSS.ai.workflows.prompt_composer

Prompt Composer System for Dynamic Agent Behavior Configuration

This module provides dynamic prompt composition capabilities, allowing agents to modify their behavior based on YAML configuration without code changes. The composer integrates with the existing prompt system while enabling runtime customization.

Architecture: - Template-based prompt composition with variable substitution - Integration with existing prompt_loader.py system - Configuration-driven behavioral modifications - Fallback to default prompts for backward compatibility

TemplateVariables

Bases: TypedDict

Strictly typed template variables for prompt composition.

Uses TypedDict with total=False to allow partial dictionaries while maintaining strict typing for each key that is present. This ensures type safety while allowing different agents to use different subsets of variables.

ComposedPrompt

Bases: BaseModel

Container for a composed prompt with metadata.

Provides structured prompt composition with template management, variable substitution, and comprehensive metadata tracking for agent behavior configuration.

validate_system_prompt_not_empty(v) classmethod

Ensure system prompt is not empty.

validate_template_names(v) classmethod

Ensure template names are valid identifiers.

get_template(template_name)

Get a specific template by name.

substitute_variables(template)

Substitute variables in a template string.

PromptComposer

Dynamic prompt composition system for configurable agent behavior.

The PromptComposer enables runtime modification of agent prompts based on configuration while maintaining backward compatibility with existing systems.

__init__()

Initialize the prompt composer.

compose_refiner_prompt(config)

Compose refiner agent prompt based on configuration.

Parameters:

Name	Type	Description	Default
config	RefinerConfig	RefinerConfig with behavioral and prompt settings	required

Returns:

Type	Description
ComposedPrompt	ComposedPrompt with customized system prompt and templates

compose_critic_prompt(config)

Compose critic agent prompt based on configuration.

Parameters:

Name	Type	Description	Default
config	CriticConfig	CriticConfig with analysis and evaluation settings	required

Returns:

Type	Description
ComposedPrompt	ComposedPrompt with customized critical analysis prompt

compose_historian_prompt(config)

Compose historian agent prompt based on configuration.

Parameters:

Name	Type	Description	Default
config	HistorianConfig	HistorianConfig with search and context settings	required

Returns:

Type	Description
ComposedPrompt	ComposedPrompt with customized context retrieval prompt

compose_synthesis_prompt(config)

Compose synthesis agent prompt based on configuration.

Parameters:

Name	Type	Description	Default
config	SynthesisConfig	SynthesisConfig with integration and analysis settings	required

Returns:

Type	Description
ComposedPrompt	ComposedPrompt with customized synthesis prompt

compose_final_prompt(config)

Compose FINAL agent prompt based on configuration.

Parameters:

Name	Type	Description	Default
config	FinalConfig	FinalConfig with behavioral and prompt settings	required

Returns:

Type	Description
ComposedPrompt	ComposedPrompt with customized system prompt and templates

compose_prompt(agent_type, config)

Universal prompt composition method for any agent type.

Parameters:

Name	Type	Description	Default
agent_type	str	Type of agent ("refiner", "critic", "historian", "synthesis")	required
config	AgentConfigType	Agent-specific configuration object	required

Returns:

Type	Description
ComposedPrompt	ComposedPrompt with customized prompt for the agent

Raises:

Type	Description
ValueError	If agent_type is not supported

get_default_prompt(agent_type)

Get the default prompt for an agent type (fallback behavior).

validate_composition(composed_prompt)

Validate that a composed prompt is well-formed.

Parameters:

Name	Type	Description	Default
composed_prompt	ComposedPrompt	ComposedPrompt to validate	required

Returns:

Type	Description
bool	True if prompt is valid, False otherwise