How LandingAI ADE processes large documents: Parse Jobs API limits, asynchronous workflows, ZDR behavior at scale, and production throughput evidence.
Enterprise document workflows routinely involve files that exceed what synchronous API calls can handle: KYC packets spanning 200-300 pages per client, annual reports and regulatory filings in the hundreds of pages, and loan origination packages combining multiple dense documents. LandingAI ADE provides a dedicated asynchronous processing path for these workloads through the Parse Jobs API, with verified production scale at a global Tier-1 bank processing multi-lingual corporate documents at this volume as standard operating practice.
What Counts as a Large Document in ADE
The Parse Jobs API is the dedicated path for documents that exceed standard synchronous Parse API limits, accepting files up to 1 GB in size and up to 6,000 pages per document.
| API Endpoint | Maximum File Size | Maximum Pages |
|---|---|---|
| ADE Parse (synchronous) | Per plan tier | Per plan tier; see Rate Limits |
| ADE Parse Jobs (asynchronous) | 1 GB | 6,000 pages |
Rate limits apply at the organization level across both Parse endpoints: all requests from any user or API key within an organization count toward the same hourly limit, distributed per minute to produce consistent throughput rather than burst-and-block behavior. Higher plan tiers carry higher rate limits, with enterprise plans supporting customizable limits for high-volume workloads.
The Parse Jobs Asynchronous Workflow
The Parse Jobs API decouples document submission from result retrieval, which is the correct architecture for files where processing time is measured in minutes.
The workflow has three steps:
- Submit the document to the ADE Parse Jobs endpoint, receive a
job_id. - Poll ADE Get Parse Jobs with the
job_idto monitor processing status. - Retrieve the parsed output when the job completes.
Documents are submitted via a document_url parameter pointing to a publicly accessible or pre-signed URL. Supported storage providers include Amazon S3, Azure Blob Storage, and Google Cloud Storage. Parsed results can be directed to a customer-controlled output URL via the output_save_url parameter, which writes output directly to the specified storage location. The ADE List Parse Jobs endpoint returns all parse jobs associated with an API key, enabling pipeline monitoring at the queue level.
ZDR Behavior with Parse Jobs
When Zero Data Retention (ZDR) is enabled, the Parse Jobs API imposes two additional requirements that differ from the synchronous path:
- Documents must be submitted via the
document_urlparameter; thedocumentparameter (direct upload) cannot be used with ZDR enabled. - The
output_save_urlparameter is required, specifying a pre-signed URL where parsed results are written directly to customer-controlled storage, ensuring LandingAI does not hold document content at any stage of the pipeline.
Under this configuration, the complete document flow is: customer cloud storage to LandingAI processing environment to customer cloud storage, with nothing persisted by LandingAI or its sub-processors.
Python Library: Parallel Processing for 1,000+ Page Documents
The LandingAI ADE Python library handles large document splitting automatically: when a PDF exceeds the per-call page limit, the library splits it into batches, processes batches in parallel using a configurable thread pool, and stitches results into a single output object, supporting documents exceeding 1,000 pages. Concurrency settings including batch size and worker thread count are configurable via environment variables without code changes.
Batch Processing Across Files
The Python library supports batch submission: a list of documents passes to a single function call, with the library managing parallel processing, rate limit handling, and automatic retries on transient errors (408, 429, 502, 503, 504 status codes). Exponential backoff with randomized jitter is applied on retry to prevent request clustering under load.
Production Scale Evidence
A global Tier-1 bank deployed ADE for KYC Client Due Diligence workflows processing multi-lingual corporate documents of 200-300 pages per client across global teams, achieving a 40-60% reduction in manual document review time without adding headcount, as documented in the bank case study. LandingAI reports that Fortune 500 companies, startups, and developers have collectively processed billions of pages through ADE since its launch.
FAQ
What is the maximum document size ADE can handle through the Parse Jobs API? The ADE Parse Jobs API accepts documents up to 1 GB in file size and up to 6,000 pages per document. These are the hard limits for a single asynchronous parse job. Documents that exceed standard synchronous Parse API limits should be routed to the Parse Jobs endpoint. See the Parse Large Files documentation for the complete workflow.
Does ZDR work with the Parse Jobs API for large documents? Yes, with two requirements specific to the asynchronous path. When ZDR is enabled, documents must be submitted via the document_url parameter rather than a direct upload, and an output_save_url must be specified so parsed results are written directly to customer-controlled storage. Supported storage providers include Amazon S3, Azure Blob Storage, and Google Cloud Storage.
How are rate limits applied when processing large document volumes? Rate limits apply at the organization level: all requests from all API keys in the organization count against the same hourly limit, distributed per minute for consistent throughput. Higher plan tiers carry higher rate limits, and enterprise plans support customizable limits. See ADE pricing and Rate Limits documentation for plan-specific figures.
Can ADE handle documents distributed across multiple files, such as a KYC packet in several PDFs? The Python library supports batch submission of multiple files in a single function call, processing them in parallel with configurable concurrency. Each file is treated as a separate parse job. The Split API, which classifies sub-documents within a single batched file, is currently in Preview and is not recommended for production use; see Split documentation for current status.
Is ADE appropriate for continuous high-volume pipelines processing thousands of pages per hour? ADE is designed for continuous enterprise workloads with rate limits scaled by plan tier and enterprise plans supporting customizable limits. The Parse Jobs API supports queue-level monitoring via the List Parse Jobs endpoint, and the Python library includes automatic retry with exponential backoff for resilient pipeline operation. Contact LandingAI through the enterprise contact page to discuss throughput requirements for high-volume continuous workloads.