Juho Inkinen
Enhanced layout
0e83ca4
raw
history blame
3.58 kB
import gradio as gr
import requests
from annif_client import AnnifClient
import os
# Get VLM API base URL and API key from environment variables
VLM_API_BASE_URL = os.getenv("VLM_API_BASE_URL")
if not VLM_API_BASE_URL:
raise RuntimeError("VLM_API_BASE_URL environment variable must be set.")
VLM_API_KEY = os.getenv("VLM_API_KEY", "")
VLM_API_ENDPOINT = f"{VLM_API_BASE_URL}/v1/chat/completions"
# Initialize Annif client (no arguments)
annif = AnnifClient()
def get_caption(image):
# Convert image to base64 JPEG
import io
import base64
buf = io.BytesIO()
image.save(buf, format="JPEG")
img_bytes = buf.getvalue()
img_b64 = base64.b64encode(img_bytes).decode("utf-8")
# Prepare payload for VLM (OpenAI schema)
payload = {
"messages": [
{
"role": "user",
"content": [
{"type": "text", "text": "What is in this image?"},
{
"type": "image_url",
"image_url": {"url": f"data:image/jpeg;base64,{img_b64}"},
},
],
}
],
"max_tokens": 300,
}
headers = {"X-API-Key": VLM_API_KEY} if VLM_API_KEY else {}
try:
response = requests.post(VLM_API_ENDPOINT, json=payload, headers=headers)
response.raise_for_status()
data = response.json()
# Assume caption is in data['choices'][0]['message']['content']
caption = data["choices"][0]["message"]["content"]
except Exception as e:
print(f"VLM API error: {e}") # Detailed error for admin
raise gr.Error("Sorry, there was a problem generating a caption.")
return caption
PROJECT_ID = "yso-en" # Placeholder, update as needed
def get_subjects(caption):
try:
results = annif.suggest(project_id=PROJECT_ID, text=caption)
label_scores = {result["label"]: result["score"] for result in results}
if not label_scores:
return {}
return label_scores
except Exception as e:
print(f"Annif API error: {e}") # Detailed error for admin
raise gr.Error("Sorry, there was a problem getting subject suggestions.")
def process_image(image):
caption = get_caption(image)
subjects = get_subjects(caption)
return image, caption, subjects
with gr.Blocks(title="VLM Caption & Annif Subject Demo") as demo:
gr.Markdown("# VLM Caption & Annif Subject Demo")
gr.Markdown(
"""
**How it works:**
1. Upload or take a photo in the input section below.
2. The image is sent to a Visual Language Model to generate a caption.
3. Annif suggests subjects based on the caption.
"""
)
with gr.Row():
with gr.Column():
gr.Markdown("### Input")
image_input = gr.Image(
type="pil", label="Image Input (upload or take a photo)"
)
submit_btn = gr.Button("Submit")
clear_btn = gr.Button("Clear")
with gr.Column():
gr.Markdown("### Output")
caption_output = gr.Textbox(label="Caption", lines=10, interactive=False)
subjects_output = gr.Label(label="Subject Suggestions", show_heading=False)
def run_app(image):
caption, subjects = process_image(image)[1:]
return caption, subjects
submit_btn.click(
run_app, inputs=image_input, outputs=[caption_output, subjects_output]
)
clear_btn.click(lambda: ("", {}), outputs=[caption_output, subjects_output])
demo.launch()