Save To Bucket¶
Collects outputs from upstream nodes, organizes them into a structured folder layout, and uploads the result set to a cloud bucket location. Returns a signed URL to a zip that represents the uploaded folder contents, suitable for downloading the full run output.
Usage¶
Use this node at the end of a workflow to persist results to a cloud bucket. Connect one or more outputs (text, structured data, or domain-specific types) to results_1, results_2, etc. Provide a destination folder path within the bucket. The node will group files by producing node and metadata when available, generate seeds/configs manifests if provided by inputs, upload everything, and return a signed URL for convenient download.
Inputs¶
| Field | Required | Type | Description | Example |
|---|---|---|---|---|
| results_1 | True | WILDCARD | First result to save. Accepts any connected output. Additional inputs results_2..results_101 become available as you connect more results. | Dictionary of PDBs, a FASTA string, or any text-like output |
| results_2..results_101 | False | WILDCARD | Additional results to save. These appear progressively after connecting the previous result input. | Additional outputs such as JSON-like configs, A3M dictionaries, etc. |
| folder_path | True | STRING | Path to the destination folder in the cloud bucket. Results may be merged with other runs by a unique workflow execution identifier. Supports nested paths using forward slashes. | biotech/experiments/run_outputs |
Outputs¶
| Field | Type | Description | Example |
|---|---|---|---|
| url | STRING | Signed URL to a zip representing the current content of the uploaded folder. Use this to download the results. | https://storage.example.com/path/to/results.zip?X-Goog-Signature= |
Important Notes¶
- Folder path is required: The node rejects empty folder paths and normalizes leading/trailing slashes.
- Input organization: Files are grouped into folders based on upstream metadata when available (including per-input folder/file names); otherwise they are grouped by producing node and output index.
- Manifests: If upstream metadata includes seeds and/or configuration values, seeds.json and configs.json are added at the root of the uploaded folder.
- Content handling: Inputs are saved as text files. Complex objects are stringified, and certain domain types (e.g., PDB, A3M) are written as individual files per item.
- Multiple results: You can chain up to 101 results inputs (results_1 through results_101). Inputs appear progressively as you connect the previous one.
- Signed link validity: The returned URL is a signed link. Its availability and expiration are managed by the platform configuration.
- Execution namespacing: Results are associated with the current workflow execution identifier to distinguish multiple runs in the same destination folder.
Troubleshooting¶
- Error: Please specify non empty folder path: Provide a non-empty folder_path. Do not rely on the bucket root.
- Upload timeout or no URL returned: Ensure the platform’s eventing and storage services are available and you have permission to write to the bucket. Retry the run.
- Unexpected folder structure: Some inputs may lack metadata. In that case, files are grouped by producing node class and output index. Add or propagate metadata upstream if you need specific names.
- Binary or large data not saved as expected: The node writes results as text. Ensure inputs are text-based or converted to a text representation before saving.
- Signed URL not downloadable: The link may have expired or been revoked. Re-run the node to generate a fresh URL or check platform policies.
Example Pipelines¶
