squishy-wsi 0.1.1

Client SDK for WSI embedding backend

WSI Embed Client SDK

Pip-installable frontend/client package for sending a local slide to the backend and saving embedding results on the user's machine.

1) Install

cd frontend
python -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

Install from built wheel (recommended for remote client machines):

pip install squishy_wsi-0.1.0-py3-none-any.whl

2) Usage

Python API

from wsi_embed_client import embed

result = embed(
    slide_path="/path/to/slide.svs",
    model="titan-real-v1",
    key="team-key-123",
    server_url="http://your-gpu-server:8000",
)

print(result["embedding_id"])
print(result["saved_path"])

CLI

After install, use:

wsi-embed /path/to/slide.svs \
  --model titan-real-v1 \
  --key team-key-123 \
  --server-url http://your-gpu-server:8000

Optional flags:

• --output-dir /path/to/save
• --timeout-seconds 600
• --print-json

What embed(...) does

1. Zips the slide file locally.
2. Sends the zip to the backend with X-API-Key.
3. Receives embedding + metadata.
4. Saves a rich JSON object to ./embeddings/ by default.

Saved filename format:

• <slide>__<model>__<timestamp>__<shortid>.embedding.json

Supported model (validated)

The end-to-end path that is tested and validated is:

• titan-real-v1

This corresponds to backend processing using TRIDENT patching + CONCH v1.5 patch features + TITAN slide aggregation.

3) Test

pytest -q