|
import gradio as gr |
|
from huggingface_hub import InferenceClient |
|
import weaviate.classes as wvc |
|
import weaviate |
|
from weaviate.auth import AuthApiKey |
|
import logging |
|
import os |
|
import requests |
|
import json |
|
|
|
logging.basicConfig(level=logging.INFO) |
|
|
|
client = weaviate.connect_to_embedded( |
|
headers={ |
|
"X-Huggingface-Api-Key": os.environ["HUGGINGFACE_API_KEY"] |
|
} |
|
) |
|
|
|
if client.is_ready(): |
|
logging.info('') |
|
logging.info(f'Found {len(client.cluster.nodes())} Weaviate nodes.') |
|
logging.info('') |
|
for node in client.cluster.nodes(): |
|
logging.info(node) |
|
logging.info('') |
|
|
|
client.collections.delete_all() |
|
|
|
questions = client.collections.create( |
|
name="Question", |
|
vectorizer_config=wvc.config.Configure.Vectorizer.text2vec_huggingface(wait_for_model=True), |
|
generative_config=wvc.config.Configure.Generative.openai() |
|
) |
|
resp = requests.get('https://raw.githubusercontent.com/databyjp/wv_demo_uploader/main/weaviate_datasets/data/jeopardy_1k.json') |
|
data = json.loads(resp.text) |
|
|
|
question_objs = list() |
|
for i, d in enumerate(data): |
|
question_objs.append({ |
|
"answer": d["Answer"], |
|
"question": d["Question"], |
|
"category": d["Category"], |
|
"air_date": d["Air Date"], |
|
"round": d["Round"], |
|
"value": d["Value"] |
|
}) |
|
|
|
logging.info('Importing 1000 Questions...') |
|
questions = client.collections.get("Question") |
|
questions.data.insert_many(question_objs) |
|
logging.info('Finished Importing Questions') |
|
|
|
def respond(query): |
|
|
|
response = questions.query.near_text( |
|
query=query, |
|
limit=1 |
|
) |
|
|
|
return response.objects[0].properties |
|
|
|
with gr.Blocks(title="Search the Jeopardy Vector Database powered by Weaviate") as demo: |
|
gr.Markdown("""# Search the Jeopardy Vector Database powered by Weaviate""") |
|
semantic_examples = [ |
|
["Nature"], |
|
["Music"], |
|
["Wine"], |
|
["Consumer Products"], |
|
["Sports"] |
|
] |
|
semantic_input_text = gr.Textbox(label="Enter a search concept or choose an example below:", value=semantic_examples[0][0]) |
|
gr.Examples(semantic_examples, |
|
|
|
inputs=semantic_input_text, label="Example search concepts:" |
|
) |
|
vdb_button = gr.Button(value="Search the Jeopardy Vector Database.") |
|
vdb_button.click(fn=respond, inputs=[semantic_input_text], outputs=gr.Textbox(label="Search Results")) |
|
|
|
|
|
if __name__ == "__main__": |
|
demo.launch() |
|
|
