When deploying your app, you will be asked to select the input format of your data. This is the format of the data that you will be sending to the API. We currently support the following input options for custom input: DocumentArray, DocumentArrayUrl, local path, S3 bucket. In the following, we will explain each of these options in more detail.
If you have your data as a
DocumentArray you can use this option.
More information about how to create and push your own
DocumentArray can be found here.
Similar to the
DocumentArray option, but instead of pushing your
DocumentArray to the API, you can provide a URL to your data.
If you have your data stored locally, you can provide the path to the folder containing your data. The folder should contain all files that you want to index.
Example structure for a text-to-image search app:
usr ├── data │ ├── images │ │ ├── 1.jpg │ │ ├── 2.jpg │ │ ├── 3.jpg │ │ ├── 4.jpg
The local path you provide should be
AWS S3 bucket#
If you have your data stored in an AWS S3 bucket, you can provide the S3 URI, your
AWS access key ID and
AWS secret key.
Similar to the local folder option, the S3 bucket should contain all files that you want to index.
The only difference is that the S3 Uri should be in the following format:
Taking the example structure from above, the S3 URI would be
Supported File Formats#
Here is an overview of the supported file formats for each modality:
.txt(can also have a different extension, but has to be plain text)
.png, … (everything supported by
.mp3, … (everything supported by