Describe the bug

When using the Qdrant Upsert Document node, the JSON provided in the Additional Collection Configuration field is ignored when a new collection is created for the first time.

This issue specifically prevents the configuration of distributed features like replication_factor and shard_number. The node proceeds to create the collection using Qdrant's internal default values, rather than the user-specified parameters. This makes it impossible to leverage Qdrant's high-availability and sharding strategies directly from Flowise.

The key failure is that replication_factor remains at 1 (the default) instead of the specified value (e.g., 3). While shard_number might appear to work if set to a value that matches the default, it's likely that the entire configuration object is being disregarded.

To Reproduce

• Set up a distributed Qdrant cluster. For this test, a cluster of 4 nodes (1 main, 3 peers) was used, as replication_factor > 1 requires a cluster.
• In Flowise, create a new flow.
• Add a Qdrant Upsert Document node.
• Configure the node with the Qdrant server URL and a new, non-existent collection name to ensure the creation logic is triggered.
• In the Additional Collection Configuration field, enter the following JSON to specify sharding and replication:

{
  "shard_number": 4,
  "replication_factor": 3
}

• Connect document loaders and embeddings to the node to provide content for the upsert.
• Execute the flow.
• Connect to the Qdrant instance directly (via the web UI or API) and inspect the parameters of the newly created collection. You will see that replication_factor is 1, not 3.

Expected behavior

The newly created collection in Qdrant should have its parameters set according to the JSON provided in the Additional Collection Configuration field.

Specifically, after the flow runs, the collection's configuration in Qdrant should show:

• replication_factor: 3
• shard_number: 4

Instead, the collection is created with replication_factor: 1.

Screenshots

Flow

test-insert Chatflow.json

Use Method

Docker

Flowise Version

3.0.4

Operating System

Linux

Browser

Chrome

Additional context

This issue has been thoroughly troubleshooted to isolate the cause to the Flowise node.

• Qdrant Cluster is Healthy: The Qdrant cluster is confirmed to be healthy and fully operational. A GET /cluster API call shows all 4 nodes are connected and a leader is elected.
• Manual Creation Works: Crucially, creating a collection manually via the Qdrant UI or a direct API call with the exact same parameters (shard_number: 4, replication_factor: 3) works perfectly. The collection is created with the correct replication factor.

This confirms that the Qdrant cluster is capable of fulfilling the request, and the issue lies specifically with how the Flowise node processes (or fails to process) the Additional Collection Configuration parameter during the initial creation of a collection.