agents.rag.utils.structured_output_enhancer¶

Structured Output Enhancer for RAG Agents.

from typing import Any, Dict This utility enables any agent to be enhanced with structured output by appending a SimpleAgent with the appropriate prompt template and Pydantic model. This follows the pattern of keeping prompts focused on generation while parsers handle structure.

Classes¶

`RAGEnhancementFactory`	Factory for creating enhanced RAG agents with structured output.
`StructuredOutputEnhancer`	Utility for enhancing any agent with structured output capabilities.

Functions¶

`create_fusion_enhancer`()	Create an enhancer for Fusion RAG structured output.
`create_hyde_enhancer`()	Create an enhancer for HyDE structured output.
`create_memory_enhancer`()	Create an enhancer for Memory-aware RAG structured output.
`create_speculative_enhancer`()	Create an enhancer for Speculative RAG structured output.
`demonstrate_enhancement_patterns`()	Demonstrate various enhancement patterns.

Module Contents¶

class agents.rag.utils.structured_output_enhancer.RAGEnhancementFactory¶

Factory for creating enhanced RAG agents with structured output.

static enhance_simple_rag(llm_config, enhancement_type='hyde')¶

Create a simple RAG with structured output enhancement.

Parameters:

llm_config (haive.core.models.llm.base.LLMConfig) – LLM configuration
enhancement_type (str) – Type of enhancement (hyde, fusion, speculative, memory)

Returns:

List of agents forming an enhanced RAG pipeline

Return type:

list[haive.agents.simple.agent.SimpleAgent]

class agents.rag.utils.structured_output_enhancer.StructuredOutputEnhancer(output_model, prompt_style=PromptStyle.DESCRIPTIVE, structured_output_version='v1')¶

Utility for enhancing any agent with structured output capabilities.

This class provides a clean pattern for appending structured output processing to any existing agent, following the principle that prompts focus on generation while parsers handle structure.

Example

>>> from haive.agents.rag.models import HyDEResult
>>>
>>> # Enhance any agent with HyDE structured output
>>> enhancer = StructuredOutputEnhancer(HyDEResult)
>>> structured_agent = enhancer.create_enhancement_agent(
...     llm_config=llm_config,
...     context_prompt="Based on the retrieved documents, generate a hypothetical document analysis"
... )
>>>
>>> # Or enhance an existing agent pipeline
>>> enhanced_pipeline = enhancer.enhance_agent_sequence([base_agent, retrieval_agent], llm_config)

Initialize the enhancer with a Pydantic output model.

Parameters:

output_model (type[pydantic.BaseModel]) – Pydantic model for structured output
prompt_style (haive.core.utils.pydantic_utils.base_model_to_prompt.PromptStyle) – Style for generating format instructions
structured_output_version (str) – v1 (parser-based) or v2 (tool-based)

create_enhancement_agent(llm_config, context_prompt, agent_name=None, include_state_context=True, **engine_kwargs)¶

Create a SimpleAgent for structured output enhancement.

Parameters:

llm_config (haive.core.models.llm.base.LLMConfig) – LLM configuration
context_prompt (str) – Main instruction for what to generate
agent_name (str | None) – Name for the enhancement agent
include_state_context (bool) – Whether to include previous state context
**engine_kwargs – Additional arguments for AugLLMConfig

Returns:

SimpleAgent configured for structured output

Return type:

haive.agents.simple.agent.SimpleAgent

create_enhancement_prompt(context_prompt, include_state_context=True)¶

Create a prompt template for structured output enhancement.

Parameters:

context_prompt (str) – The main instruction for what to generate
include_state_context (bool) – Whether to include previous state context

Returns:

ChatPromptTemplate configured for structured output

Return type:

langchain_core.prompts.ChatPromptTemplate

create_format_instructions()¶

Generate format instructions for the output model.

Return type:: str

enhance_agent_sequence(agents, llm_config, context_prompt=None, **kwargs)¶

Enhance a sequence of agents by appending structured output processing.

Parameters:

agents (list[Any]) – List of existing agents
llm_config (haive.core.models.llm.base.LLMConfig) – LLM configuration
context_prompt (str | None) – Custom context prompt (auto-generated if None)
**kwargs – Additional arguments for the enhancement agent

Returns: