agents.rag.agentic.document_grader

Document Grading Agent for Agentic RAG.

This agent evaluates retrieved documents for relevance using existing models from common.

Functions

create_document_grader_agent([name, temperature])

Create a document grader agent using direct SimpleAgent instantiation.

grade_documents(agent, query, documents)

Grade documents for relevance to a query.

Module Contents

agents.rag.agentic.document_grader.create_document_grader_agent(name='document_grader', temperature=0.0, **kwargs)

Create a document grader agent using direct SimpleAgent instantiation.

Parameters:

  • name (str) – Agent name (default: “document_grader”)

  • temperature (float) – LLM temperature (default: 0.0 for consistency)

  • **kwargs – Additional configuration options

Returns:

SimpleAgent configured for document grading

Return type:

haive.agents.simple.SimpleAgent

Example

# Create grader agent
grader = create_document_grader_agent(
name="doc_grader",
temperature=0.0
)

# Grade documents
result = await grader.arun({
"query": "What is quantum computing?",
"documents": [
{"content": "Quantum computing uses quantum mechanics...", "id": "doc1"},
{"content": "Classical computing uses binary digits...", "id": "doc2"}
]
})

# Access results
for decision in result.document_decisions:
print(f"Document {decision.document_id}: {decision.decision}")
print(f"Reason: {decision.justification}")
async agents.rag.agentic.document_grader.grade_documents(agent, query, documents)

Grade documents for relevance to a query.

Parameters:

  • agent (haive.agents.simple.SimpleAgent) – The document grader agent

  • query (str) – The user query

  • documents (list[dict[str, Any]]) – List of documents with ‘content’ and ‘id’ fields

Returns:

DocumentBinaryResponse with grading results

Return type:

haive.agents.rag.common.document_graders.models.DocumentBinaryResponse