Upload and Download Files with AI Agents

​

”Upload my training script and config to the training instance.”

1. list_instances — finds the target instance
2. file_upload — uploads train.py to /workspace/train.py
3. file_upload — uploads config.yaml to /workspace/config.yaml
4. ssh_execute — runs ls -lh /workspace/ to confirm

Uploaded 2 files to training-run:

File	Size	Remote Path
train.py	3.8 KB	/workspace/train.py
config.yaml	512 B	/workspace/config.yaml

Both files verified on the instance.

​

”Download the trained model weights and the training log.”

1. ssh_execute — runs ls -lh /workspace/output/ to see what’s there
2. file_download — downloads /workspace/output/model.safetensors
3. file_download — downloads /workspace/output/training.log

Files on the instance:

-rw-r--r-- 1 root root 1.2G model.safetensors
-rw-r--r-- 1 root root  48K training.log
-rw-r--r-- 1 root root 2.1K config.json

Downloaded model.safetensors (1.2 GB) and training.log (48 KB).

​

”Upload a requirements.txt and install everything on my dev-box.”

1. file_upload — uploads requirements.txt to /root/requirements.txt
2. ssh_execute — runs pip install -r /root/requirements.txt
3. ssh_execute — runs pip list | wc -l to confirm packages installed

Uploaded requirements.txt (28 packages) to dev-box. Installation output:

Successfully installed torch-2.4.0 transformers-4.44.0 ...

28 packages installed. Environment is ready.

​

Tools used in this workflow