Use LlamaParse from the LlamaIndex Framework

The LlamaIndex Framework is our open-source Python toolkit for building agents and retrieval-augmented generation (RAG) over your own data. Its built-in readers handle clean text well. For scanned PDFs, forms, spreadsheets and slide decks, run a Parse job first and hand the result to the framework: the quality of what goes into an index decides the quality of every answer it gives.

Both products use the same API key. Set LLAMA_CLOUD_API_KEY in your environment and install the two packages:

pip install llama-index "llama-cloud>=2.8"
export LLAMA_CLOUD_API_KEY=llx-...

Build an index from Parse output

parsing.parse() uploads the file, waits for the job to finish, and returns the result. With expand=["markdown"] the result carries one Markdown string per page. Wrap each page in a framework Document and build an index as you would from any other text:

from llama_cloud import LlamaCloud
from llama_index.core import Document, VectorStoreIndex

client = LlamaCloud()  # reads LLAMA_CLOUD_API_KEY

file = client.files.create(file="data/report.pdf", purpose="parse")
result = client.parsing.parse(
    file_id=file.id, tier="agentic", version="latest", expand=["markdown"]
)
pages = result.markdown.pages
documents = [
    Document(text=p.markdown, metadata={"page": p.page_number})
    for p in pages
    if p.success
]

index = VectorStoreIndex.from_documents(documents)
query_engine = index.as_query_engine()
response = query_engine.query("What does this document say about pricing?")
print(response)

Pages that failed to parse come back with success set to False and no Markdown, which is why the comprehension filters on it. Keeping the page number in metadata lets retrieval results point back to the page they came from.

The same pattern works for a folder of files: run one Parse job per file, collect the documents, and build a single index. For larger collections, Index keeps an index in sync with your data sources and serves retrieval over the API, so nothing needs to be rebuilt locally.

Use Extract output in an agent

Extract returns JSON shaped by your schema. A framework agent can call it as a tool, so the agent decides when to pull structured data out of a document:

from llama_cloud import LlamaCloud
from llama_index.core.agent.workflow import FunctionAgent
from llama_index.llms.openai import OpenAI

client = LlamaCloud()


def extract_invoice(path: str) -> dict:
    """Extract vendor, invoice number, date and total from an invoice file."""
    file = client.files.create(file=path, purpose="extract")
    job = client.extract.run(  # creates the job and waits for it to finish
        file_input=file.id,
        configuration={
            "data_schema": {
                "type": "object",
                "properties": {
                    "vendor": {"type": "string"},
                    "invoice_number": {"type": "string"},
                    "date": {"type": "string"},
                    "total": {"type": "number"},
                },
            }
        },
    )
    return job.extract_result


agent = FunctionAgent(
    tools=[extract_invoice],
    llm=OpenAI(model="gpt-4.1-mini"),
    system_prompt="You answer questions about invoices by extracting their fields.",
)

See Getting started with Extract for schema options and Building agents for the agent side.

Choosing where parsing runs

Need	Use
Clean text files, Markdown, HTML	The framework’s built-in readers
Local parsing with no API key	LiteParse, the open-source parser
Scans, tables, charts, forms, 130+ formats	Parse, as above
Structured fields from documents	Extract, as above
A managed index that stays in sync with a data source	Index

See also

• Parse getting started
• Framework starter tutorial
• LlamaIndex Framework overview