API Reference#

This page provides API references for the nv-ingest-api, nv-ingest-client, and nv-ingest modules.

Basic Usage#

The following will provide a basic example of submitting a job using the Python API

To begin, get the pipeline running in library mode with PipelineCreationSchema(), which creates a config object for the pipeline and run_pipeline(), which starts the pipline

from nv_ingest.framework.orchestration.ray.util.pipeline.pipeline_runners import run_pipeline
from nv_ingest.framework.orchestration.ray.util.pipeline.pipeline_runners import PipelineCreationSchema

# Start the pipeline subprocess for library mode
config = PipelineCreationSchema()

run_pipeline(config, block=False, disable_dynamic_scaling=True, run_in_subprocess=True)

Then, connect the client to the pipeline with NvIngestClient()

from nv_ingest_client.client import NvIngestClient

client = NvIngestClient(
    message_client_allocator=SimpleClient,
    message_client_port=7671,
    message_client_hostname="localhost"
)

Next, use the Ingestor class to ingest the test PDF (click the Ingestor link to see descriptions of the available tasks)

import time
from nv_ingest_client.client import NvIngestClient

# gpu_cagra accelerated indexing is not available in milvus-lite
# Provide a filename for milvus_uri to use milvus-lite
milvus_uri = "milvus.db"
collection_name = "test"
sparse = False

# do content extraction from files
ingestor = (
    Ingestor(client=client)
    .files("data/multimodal_test.pdf")
    .extract(
        extract_text=True,
        extract_tables=True,
        extract_charts=True,
        extract_images=True,
        paddle_output_format="markdown",
        extract_infographics=True,
        # extract_method="nemoretriever_parse", #Slower, but maximally accurate, especially for PDFs with pages that are scanned images
        text_depth="page"
    ).embed()
    .vdb_upload(
        collection_name=collection_name,
        milvus_uri=milvus_uri,
        sparse=sparse,
        # for llama-3.2 embedder, use 1024 for e5-v5
        dense_dim=2048
    )
)

print("Starting ingestion..")
t0 = time.time()

# Return both successes and failures
# Use for large batches where you want successful chunks/pages to be committed, while collecting detailed diagnostics for failures.
results, failures = ingestor.ingest(show_progress=True, return_failures=True)

# Return only successes
# results = ingestor.ingest(show_progress=True)

t1 = time.time()
print(f"Total time: {t1 - t0} seconds")

To inspect the results, use ingest_json_results_to_blob()

from nv_ingest_client.util.process_json_files import ingest_json_results_to_blob

# results blob is directly inspectable
print(ingest_json_results_to_blob(results[0]))

To query the ingested results from the milvus database, use nvingest_retrieval()

from nv_ingest_client.util.milvus import nvingest_retrieval

milvus_uri = "milvus.db"
collection_name = "test"
sparse=False

queries = ["Which animal is responsible for the typos?"]

retrieved_docs = nvingest_retrieval(
    queries,
    collection_name,
    milvus_uri=milvus_uri,
    hybrid=sparse,
    top_k=1,
)

NV-Ingest Packages

On this page