Hello everyone,
I'm working on an n8n workflow (v1.92.2, self-hosted on Windows) to automate PDF to HTML conversion using the Adobe PDF Services API. I'm stuck on a persistent "404 Not Found" error and would appreciate any help.
The Goal: Implement the asynchronous process described in the Adobe API documentation:
Create Asset (POST /assets): Get an assetID and uploadUri.
Upload File (PUT {uploadUri}): Upload the binary PDF data.
Start Conversion (POST /pdf/export): Send the assetID to start the conversion.
Poll Status (GET {locationHeader}): Check the job status until it's "done".
Download Result: Download the final HTML file.
The Problem: My workflow consistently fails at Step 3 (Start Conversion). The node Adobe - Iniciar Conversão receives a 404 Not Found error from the Adobe API.
My Hypothesis: This is a classic race condition. The request to start the conversion is arriving at Adobe's servers before the uploaded file from Step 2 is fully processed and available, even though the upload itself returns a success code. The API then tries to find an assetID that, for its purposes, doesn't exist yet.
What I've Tried (Unsuccessfully): I've exhausted several architectures to solve this, all of which have failed:
Simple Linear Flow: A straight line of nodes fails because a node like "Upload File" does not pass the necessary data (assetID, access_token) to the next step.
Linear Flow with "Include Input Data": Attempting to use the "Include Input Data in Output" option on the HTTP Request nodes also failed to preserve the binary data correctly.
Parallel Flow with Merge Node: This approach caused an internal n8n error (TypeError: Cannot read properties of undefined (reading 'execute')), suggesting it's incompatible with my n8n version or this specific use case.
Parallel Flow with Code Node: I used a Code node to manually merge the binary file from one input with the Adobe API data from another. While this fixed the data loss, it introduced a race condition, as the Upload and Start Conversion steps were triggered in parallel.
Adding Wait Nodes: Adding a Wait node between the Upload and Start Conversion steps (even up to 30 seconds) did not resolve the 404 error.
Built-in "Retry on Fail": Configuring the "Start Conversion" node to retry on fail (5 attempts, 5-second interval) also resulted in a final 404 error.
New Adobe Credentials: I generated a completely new set of API credentials in the Adobe Developer Console, which did not solve the issue.
The Ask: What is the correct and robust n8n architecture to handle this specific asynchronous API pattern, where one step (Upload) must be fully completed and processed on the server before the next step (Start Conversion) can be initiated?
How can I reliably pass both the binary data and the JSON data through the entire workflow without one overwriting the other, and without triggering internal n8n errors?
I would be grateful for any insights or examples of similar workflows. Thank you.
AdChoices
