|
|
import base64 |
|
|
import mimetypes |
|
|
import os |
|
|
import PIL.Image |
|
|
import requests |
|
|
import uuid |
|
|
|
|
|
from smolagents import tool |
|
|
|
|
|
|
|
|
|
|
|
def encode_image(image_path): |
|
|
if image_path.startswith("http"): |
|
|
user_agent = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/119.0.0.0 Safari/537.36 Edg/119.0.0.0" |
|
|
request_kwargs = { |
|
|
"headers": {"User-Agent": user_agent}, |
|
|
"stream": True, |
|
|
} |
|
|
|
|
|
|
|
|
response = requests.get(image_path, **request_kwargs) |
|
|
response.raise_for_status() |
|
|
content_type = response.headers.get("content-type", "") |
|
|
|
|
|
extension = mimetypes.guess_extension(content_type) |
|
|
if extension is None: |
|
|
extension = ".download" |
|
|
|
|
|
fname = str(uuid.uuid4()) + extension |
|
|
download_path = os.path.abspath(os.path.join("downloads", fname)) |
|
|
|
|
|
with open(download_path, "wb") as fh: |
|
|
for chunk in response.iter_content(chunk_size=512): |
|
|
fh.write(chunk) |
|
|
|
|
|
image_path = download_path |
|
|
|
|
|
with open(image_path, "rb") as image_file: |
|
|
return base64.b64encode(image_file.read()).decode("utf-8") |
|
|
|
|
|
|
|
|
def resize_image(image_path): |
|
|
img = PIL.Image.open(image_path) |
|
|
width, height = img.size |
|
|
img = img.resize((int(width / 2), int(height / 2))) |
|
|
new_image_path = f"resized_{image_path}" |
|
|
img.save(new_image_path) |
|
|
return new_image_path |
|
|
|
|
|
|
|
|
@tool |
|
|
def visualizer(image_path: str, question: str | None = None) -> str: |
|
|
"""A tool that can answer questions about attached images. |
|
|
|
|
|
Args: |
|
|
image_path: The path to the image on which to answer the question. This should be a local path to downloaded image. |
|
|
question: The question to answer. |
|
|
""" |
|
|
add_note = False |
|
|
if not question: |
|
|
add_note = True |
|
|
question = "Please write a detailed caption for this image." |
|
|
if not isinstance(image_path, str): |
|
|
raise Exception("You should provide at least `image_path` string argument to this tool!") |
|
|
|
|
|
mime_type, _ = mimetypes.guess_type(image_path) |
|
|
base64_image = encode_image(image_path) |
|
|
|
|
|
payload = { |
|
|
"model": "gpt-4o", |
|
|
"messages": [ |
|
|
{ |
|
|
"role": "user", |
|
|
"content": [ |
|
|
{"type": "text", "text": question}, |
|
|
{"type": "image_url", "image_url": {"url": f"data:{mime_type};base64,{base64_image}"}}, |
|
|
], |
|
|
} |
|
|
], |
|
|
"max_tokens": 1000, |
|
|
} |
|
|
headers = {"Content-Type": "application/json", "Authorization": f"Bearer {os.getenv('OPENAI_API_KEY')}"} |
|
|
response = requests.post("https://api.openai.com/v1/chat/completions", headers=headers, json=payload) |
|
|
try: |
|
|
output = response.json()["choices"][0]["message"]["content"] |
|
|
except Exception: |
|
|
raise Exception(f"Response format unexpected: {response.json()}") |
|
|
|
|
|
if add_note: |
|
|
output = f"You did not provide a particular question, so here is a detailed caption for the image: {output}" |
|
|
|
|
|
return output |
|
|
|
