a la https://x.com/DigThatData/status/1800562625442623741

3d-umap-interactive-dataviz.py

import numpy as np
from openai import OpenAI
import plotly
import plotly.graph_objs as go
import umap


url = "http://localhost:80"

client = OpenAI(
    # This is the default and can be omitted
    #api_key=os.environ.get("OPENAI_API_KEY"),
    api_key="123",
    base_url=url + "/v1"
)

def get_model_name():
    response = client.models.list()
    return response.to_dict()['data'][0]['id']

MODEL_NAME = get_model_name()


def generate(prompt,
    model=MODEL_NAME,
    max_tokens=1024,
    temperature=0.1,
             **kargs
):
    if not model:
        model = get_model_name()
        
    completion = client.completions.create(
        prompt=prompt,
        model=model,
        max_tokens=max_tokens,
        temperature=temperature,
        **kargs
    )

    response = completion.choices[0].text
    response = response.strip()
    return response

def embed(content,
    model=MODEL_NAME,
    **kargs
):
    if not model:
        model = get_model_name()
        
    response = client.embeddings.create(
        input=content,
        model=model,
        encoding_format='float',
        **kargs
    )

    return response

for i, a in enumerate(articles):
    #a['vect'] = model.encode([a['content']])
    a['content']
    if len(content) > 32768:
        content = content[:32768] # would be nice if we could get the model's character limit from the API
    a['vect'] = embed(content).data[0].embedding
    if (i % 50) == 0:
        print(f"{i}\t{a['metadata']['inferred_article_title']}")
        

X = np.array([np.array(a['vect']).ravel() for a in articles])
trans = umap.UMAP(n_neighbors=10, metric='cosine', n_components=3, random_state=42).fit(X)


xs = np.array([a['umap'][:,0] for a in articles]).ravel()
ys = np.array([a['umap'][:,1] for a in articles]).ravel()
zs = np.array([a['umap'][:,2] for a in articles]).ravel()
ts = [a['metadata'].get('inferred_article_title', '') for a in articles]
cs = [a['metadata']['create_time'] for a in articles]

scattered = go.Scatter3d(
    x=xs,
    y=ys,
    z=zs,
    text=ts,
    hoverinfo='text',
    marker={'size':2, 'color':cs, 'colorscale':'Spectral'},
    line={'width':.5, 'color':cs, 'colorscale':'Spectral'},
)

fig = go.Figure(data=scattered)
fig.update_layout(showlegend=False, height=int(700), 
                  scene=dict(
                      xaxis=dict(showbackground=False, visible=False),
                      yaxis=dict(showbackground=False, visible=False),
                      zaxis=dict(showbackground=False, visible=False),
                  ))
fig.show()