Configuring S3 Storage Service

LobeHub supports multimodal AI sessions, including the ability to upload unstructured data such as images and files to large language models. To optimize storage and performance, we use S3-compatible file storage services to store image files and support file upload/knowledge base functionality.

Core Environment Variables

S3_ACCESS_KEY_ID

• Type: Required
• Description: Access key ID for S3 storage service
• Default: -
• Example: AKIAIOSFODNN7EXAMPLE

S3_SECRET_ACCESS_KEY

• Type: Required
• Description: Secret access key for S3 storage service
• Default: -
• Example: wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY

S3_ENDPOINT

• Type: Required
• Description: Request endpoint for the bucket; for virtual-host mode, should not include the bucket name prefix
• Default: -
• Example: https://cos.ap-chengdu.myqcloud.com

S3_BUCKET

• Type: Required
• Description: Name of the storage bucket
• Default: -
• Example: my-bucket

S3_REGION

• Type: Optional
• Description: Region of the storage bucket
• Default: -
• Example: ap-chengdu

S3_SET_ACL

• Type: Optional
• Description: Whether to set ACL to public-read when uploading files
• Default: 1
• Example: 0

S3_ENABLE_PATH_STYLE

• Type: Optional
• Description: Whether to enable S3 path-style access mode
• Default: 0
• Example: 1

<Callout type={'info'}> path-style and virtual-host are different methods of accessing buckets and objects in S3, with differences in URL structure and domain name resolution.

Assuming the S3 service provider's domain is s3.example.net, the bucket is mybucket, and the object is config.env, the specific differences are:

• path-style: s3.example.net/mybucket/config.env
• virtual-host: mybucket.s3.example.net/config.env </Callout>

LLM_VISION_IMAGE_USE_BASE64

• Type: Optional
• Description: Set to 1 to use base64 encoding for image upload.
• Default: undefined
• Example: 1

When set to 1, LobeHub will convert images to base64 encoding before uploading them to the LLM model. When encountering the following error, please consider configuring this environment variable to 1:

Route: [xai] ProviderBizError: Fetching images over plain http:// is not supported.

LLM_VISION_VIDEO_USE_BASE64

• Type: Optional
• Description: Set to 1 to use base64 encoding for video upload.
• Default: undefined
• Example: 1

When set to 1, LobeChat will convert videos to base64 encoding before uploading them to the LLM model.