Computer visionfor crops, pests, and livestock.

import os, requests
from pathlib import Path

# A new pest shows up mid-season — no labelled data. Zero-shot saves the day.
resp = requests.post(
    "https://api.msightflow.ai/v1/detect/zero-shot",
    headers={"Authorization": f"Bearer {os.environ['MSF_API_KEY']}"},
    files={"image": Path("drone_pass_0042.jpg").read_bytes()},
    data={
        "prompt": "leaf miner damage, aphid colony, powdery mildew, healthy leaf",
        "confidence_threshold": "0.25",
    },
).json()

# Per-detection: class label, bbox, confidence
healthy = [d for d in resp["detections"] if d["label"] == "healthy leaf"]
infected = [d for d in resp["detections"] if d["label"] != "healthy leaf"]
print(f"Healthy: {len(healthy)}  Infected: {len(infected)}")

# Fruit counting on a single tree image.
import requests

det = requests.post(
    "https://api.msightflow.ai/v1/detect",
    headers={"Authorization": f"Bearer {os.environ['MSF_API_KEY']}"},
    files={"image": open("orchard_tree.jpg", "rb")},
    data={"model": "yolo-orchard"},   # pre-trained orchard / fruit model
).json()

fruits_by_class = {}
for d in det["detections"]:
    fruits_by_class.setdefault(d["label"], 0)
    fruits_by_class[d["label"]] += 1
print(fruits_by_class)
# → {"apple_ripe": 142, "apple_unripe": 38, "apple_damaged": 7}

# Cluster a field into health zones using CLIP embeddings.
import requests, numpy as np
from pathlib import Path

def embed(p):
    r = requests.post("https://api.msightflow.ai/v1/embed",
        headers={"Authorization": f"Bearer {os.environ['MSF_API_KEY']}"},
        files={"image": p.read_bytes()},
    ).json()
    return np.array(r["vector"], dtype=np.float32)

vectors = {p.stem: embed(p) for p in Path("drone_tiles/").glob("*.jpg")}
# Then run k-means / DBSCAN on the vectors to find visually similar zones.
# Each cluster is a candidate health zone for ground-truth follow-up.

# Custom keypoint schema for cow lameness scoring.
# Define schema once (/v1/skeletons), then use the pose endpoint with it.
import requests

resp = requests.post(
    "https://api.msightflow.ai/v1/pose",
    headers={"Authorization": f"Bearer {os.environ['MSF_API_KEY']}"},
    files={"image": open("cow_gait_001.jpg", "rb")},
    data={"skeleton_id": "cow-12-point"},   # your custom schema
).json()

for cow in resp["animals"]:
    print(f"bbox={cow['box']}, head-tail angle={cow['keypoints']}")
# Combine across video frames for gait + lameness score.