Azure OpenAI Batch API Parallel Prompt Processing - n8n Workflow

• Data scientists requiring large-scale, asynchronous AI completions.


• Developers building scalable integrations with Azure OpenAI.


• Users looking for advanced n8n templates showcasing complex flow control (polling).


• n8n automation specialists optimizing token usage and throughput.

The Azure OpenAI Batch API is essential for processing large volumes of AI requests efficiently and asynchronously. This robust n8n workflow solves the complexity of managing the entire batch job lifecycle: creating the input file in JSONL format, uploading it, initiating the parallel batch job, continuously polling the job status, downloading the resulting output file, and correctly parsing the final responses.

This specialized n8n workflow demonstrates how to leverage this high-throughput API directly within your automation environment, significantly increasing the scalability of your generative AI tasks. The included example flow shows how to combine simple prompts with those generated from a Langchain chat memory component into a single batch submission, proving the versatility of this n8n template.

The n8n workflow operates in two main phases: preparation/submission and monitoring/retrieval.

1. Preparation: The process starts either via the primary n8n trigger (When Executed by Another Workflow) or the built-in manual example. Input JSON requests are converted to the JSON Lines (JSONL) format required by the Batch API using a dedicated Code n8n node (JSON requests to JSONL). This file is then packaged using the Convert to File n8n node and uploaded to Azure OpenAI using the Upload batch file HTTP Request n8n node.


2. File Polling: An If n8n node (If upload processed) checks the file status. If the file is not yet ready, a Wait n8n node (File Upload Poll Interval) pauses execution before the Track file upload status HTTP Request n8n node polls the API again.


3. Batch Creation & Polling: Once the file is processed, the Create batch job HTTP Request n8n node initiates the parallel processing job. The Track batch job progress node then polls the API status via a loop controlled by an If n8n node and a Batch Status Poll Interval n8n node. This flow control ensures the n8n workflow waits for completion without tying up resources.


4. Result Retrieval: When the job is marked as completed, the n8n workflow uses the Retrieve batch job output file HTTP Request n8n node to download the final results, which are delivered in JSONL format.


5. Parsing & Filtering: The Parse response Code n8n node transforms the raw JSONL output into readable JSON objects. The Split Out Parsed Results node separates the batch responses. Finally, Filter n8n nodes use the pre-defined custom_id fields to correctly associate the parallel AI responses back to their original input prompts, completing the end-to-end n8n workflow execution.

1. Importing the n8n template: Copy the provided JSON code. In your n8n instance, go to the Workflows section, click "+ New", and select "Import from JSON". This will deploy the entire n8n workflow structure.


2. Azure OpenAI Credentials: This n8n workflow uses the generic HTTP Request n8n node for API interaction. You must configure the necessary Azure OpenAI authentication (Endpoint and API Key) as headers or query parameters (e.g., api-key) for all relevant HTTP Request nodes, ensuring the correct base URL for your deployment is used.


3. Review Endpoints: Check the resource field in the HTTP Request n8n nodes (Upload batch file, Create batch job, etc.) to ensure they match your specific Azure OpenAI instance endpoint. The correct api-version is managed by a dedicated Set n8n node, but verification is recommended.


4. Testing: Use the Run example manual n8n trigger to execute the full batch pipeline functionality. The example uses Langchain Memory nodes; ensure the Langchain integration is installed if you wish to run this specific example part of the n8n workflow.

When Executed by Another Workflow (n8n trigger): The primary trigger for production use, allowing this complex batch process to be efficiently called from other n8n templates.
Code (JSON requests to JSONL, Parse response): Critical n8n node used for data transformation, converting standard JSON lists into the required JSONL format for batch submission and parsing the JSONL response file upon retrieval.
Convert To File (Convert requests jsonl to File): Transforms the JSONL output into a file object necessary for the multipart-form data upload via the HTTP Request n8n node.
HTTP Request (Upload batch file, Create batch job, Track status): Core n8n node utilized for communicating with the Azure OpenAI API, handling file uploads, batch job creation, and status polling.
Wait (Batch Status Poll Interval): An essential n8n node for implementing efficient polling, pausing the n8n workflow execution dynamically until the next required status check.
If (If ended processing, If upload processed): Manages the flow control loops, checking the status returned by the Azure API before proceeding to the next step or looping back to the Wait n8n node for further polling.

• Set (Set desired 'api-version'): Ensures the modern, correct API version is consistently included in the batch job creation payload, guaranteeing the entire n8n workflow uses compatible endpoints.