Upload app.py with huggingface_hub
Browse files
app.py
ADDED
|
@@ -0,0 +1,242 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from fastapi import FastAPI, Request, BackgroundTasks
|
| 2 |
+
from fastapi.responses import HTMLResponse
|
| 3 |
+
from fastapi.staticfiles import StaticFiles
|
| 4 |
+
from fastapi.templating import Jinja2Templates
|
| 5 |
+
import requests
|
| 6 |
+
from bs4 import BeautifulSoup
|
| 7 |
+
import asyncio
|
| 8 |
+
import aiohttp
|
| 9 |
+
from datetime import datetime, timezone
|
| 10 |
+
from typing import List, Dict, Optional
|
| 11 |
+
import uvicorn
|
| 12 |
+
import os
|
| 13 |
+
import pandas as pd
|
| 14 |
+
from datasets import Dataset, load_dataset
|
| 15 |
+
from huggingface_hub import HfApi
|
| 16 |
+
import logging
|
| 17 |
+
from contextlib import asynccontextmanager
|
| 18 |
+
|
| 19 |
+
# Configure logging
|
| 20 |
+
logging.basicConfig(level=logging.INFO)
|
| 21 |
+
logger = logging.getLogger(__name__)
|
| 22 |
+
|
| 23 |
+
# Global variables for dataset management
|
| 24 |
+
DATASET_REPO_NAME = os.getenv("DATASET_REPO_NAME", "nbroad/hf-inference-providers-data")
|
| 25 |
+
HF_TOKEN = os.getenv("HF_TOKEN")
|
| 26 |
+
|
| 27 |
+
# Time to wait between data collection runs in seconds
|
| 28 |
+
DATA_COLLECTION_INTERVAL = 1800
|
| 29 |
+
|
| 30 |
+
# Background task state
|
| 31 |
+
data_collection_task = None
|
| 32 |
+
|
| 33 |
+
@asynccontextmanager
|
| 34 |
+
async def lifespan(app: FastAPI):
|
| 35 |
+
"""Manage application lifecycle"""
|
| 36 |
+
# Start background task
|
| 37 |
+
global data_collection_task
|
| 38 |
+
data_collection_task = asyncio.create_task(timed_data_collection())
|
| 39 |
+
logger.info("Started hourly data collection task")
|
| 40 |
+
yield
|
| 41 |
+
# Cleanup
|
| 42 |
+
if data_collection_task:
|
| 43 |
+
data_collection_task.cancel()
|
| 44 |
+
logger.info("Stopped hourly data collection task")
|
| 45 |
+
|
| 46 |
+
app = FastAPI(title="Inference Provider Dashboard", lifespan=lifespan)
|
| 47 |
+
|
| 48 |
+
# List of providers to track
|
| 49 |
+
PROVIDERS = [
|
| 50 |
+
"togethercomputer",
|
| 51 |
+
"fireworks-ai",
|
| 52 |
+
"nebius",
|
| 53 |
+
"fal",
|
| 54 |
+
"groq",
|
| 55 |
+
"cerebras",
|
| 56 |
+
"sambanovasystems",
|
| 57 |
+
"replicate",
|
| 58 |
+
"novita",
|
| 59 |
+
"Hyperbolic",
|
| 60 |
+
"featherless-ai",
|
| 61 |
+
"CohereLabs",
|
| 62 |
+
"nscale",
|
| 63 |
+
]
|
| 64 |
+
|
| 65 |
+
templates = Jinja2Templates(directory="templates")
|
| 66 |
+
|
| 67 |
+
async def get_monthly_requests(session: aiohttp.ClientSession, provider: str) -> Dict[str, str]:
|
| 68 |
+
"""Get monthly requests for a provider from HuggingFace"""
|
| 69 |
+
url = f"https://huggingface.co/{provider}"
|
| 70 |
+
try:
|
| 71 |
+
async with session.get(url) as response:
|
| 72 |
+
html = await response.text()
|
| 73 |
+
soup = BeautifulSoup(html, 'html.parser')
|
| 74 |
+
request_div = soup.find('div', text=lambda t: t and 'monthly requests' in t.lower())
|
| 75 |
+
if request_div:
|
| 76 |
+
requests_text = request_div.text.split()[0].replace(',', '')
|
| 77 |
+
return {
|
| 78 |
+
"provider": provider,
|
| 79 |
+
"monthly_requests": requests_text,
|
| 80 |
+
"monthly_requests_int": int(requests_text) if requests_text.isdigit() else 0
|
| 81 |
+
}
|
| 82 |
+
return {
|
| 83 |
+
"provider": provider,
|
| 84 |
+
"monthly_requests": "N/A",
|
| 85 |
+
"monthly_requests_int": 0
|
| 86 |
+
}
|
| 87 |
+
except Exception as e:
|
| 88 |
+
logger.error(f"Error fetching {provider}: {e}")
|
| 89 |
+
return {
|
| 90 |
+
"provider": provider,
|
| 91 |
+
"monthly_requests": "N/A",
|
| 92 |
+
"monthly_requests_int": 0
|
| 93 |
+
}
|
| 94 |
+
|
| 95 |
+
async def collect_and_store_data():
|
| 96 |
+
"""Collect current data and store it in the dataset"""
|
| 97 |
+
if not HF_TOKEN:
|
| 98 |
+
logger.warning("No HF_TOKEN found, skipping data storage")
|
| 99 |
+
return
|
| 100 |
+
|
| 101 |
+
try:
|
| 102 |
+
logger.info("Collecting data for storage...")
|
| 103 |
+
|
| 104 |
+
# Collect current data
|
| 105 |
+
async with aiohttp.ClientSession() as session:
|
| 106 |
+
tasks = [get_monthly_requests(session, provider) for provider in PROVIDERS]
|
| 107 |
+
results = await asyncio.gather(*tasks)
|
| 108 |
+
|
| 109 |
+
# Create DataFrame with timestamp
|
| 110 |
+
timestamp = datetime.now(timezone.utc).isoformat()
|
| 111 |
+
data_rows = []
|
| 112 |
+
|
| 113 |
+
for result in results:
|
| 114 |
+
data_rows.append({
|
| 115 |
+
"timestamp": timestamp,
|
| 116 |
+
"provider": result["provider"],
|
| 117 |
+
"monthly_requests": result["monthly_requests"],
|
| 118 |
+
"monthly_requests_int": result["monthly_requests_int"]
|
| 119 |
+
})
|
| 120 |
+
|
| 121 |
+
new_df = pd.DataFrame(data_rows)
|
| 122 |
+
|
| 123 |
+
# Try to load existing dataset and append
|
| 124 |
+
try:
|
| 125 |
+
existing_dataset = load_dataset(DATASET_REPO_NAME, split="train")
|
| 126 |
+
existing_df = existing_dataset.to_pandas()
|
| 127 |
+
combined_df = pd.concat([existing_df, new_df], ignore_index=True)
|
| 128 |
+
except Exception as e:
|
| 129 |
+
logger.info(f"Creating new dataset (existing not found): {e}")
|
| 130 |
+
combined_df = new_df
|
| 131 |
+
|
| 132 |
+
# Convert back to dataset and push
|
| 133 |
+
new_dataset = Dataset.from_pandas(combined_df)
|
| 134 |
+
new_dataset.push_to_hub(DATASET_REPO_NAME, token=HF_TOKEN, private=False)
|
| 135 |
+
|
| 136 |
+
logger.info(f"Successfully stored data for {len(results)} providers")
|
| 137 |
+
|
| 138 |
+
except Exception as e:
|
| 139 |
+
logger.error(f"Error collecting and storing data: {e}")
|
| 140 |
+
|
| 141 |
+
async def timed_data_collection():
|
| 142 |
+
"""Background task that runs every DATA_COLLECTION_INTERVAL seconds to collect data"""
|
| 143 |
+
while True:
|
| 144 |
+
try:
|
| 145 |
+
await collect_and_store_data()
|
| 146 |
+
await asyncio.sleep(DATA_COLLECTION_INTERVAL)
|
| 147 |
+
except asyncio.CancelledError:
|
| 148 |
+
logger.info("Data collection task cancelled")
|
| 149 |
+
break
|
| 150 |
+
except Exception as e:
|
| 151 |
+
logger.error(f"Error in hourly data collection: {e}")
|
| 152 |
+
# Wait 5 minutes before retrying on error
|
| 153 |
+
await asyncio.sleep(300)
|
| 154 |
+
|
| 155 |
+
@app.get("/")
|
| 156 |
+
async def dashboard(request: Request):
|
| 157 |
+
"""Serve the main dashboard page"""
|
| 158 |
+
return templates.TemplateResponse("dashboard.html", {"request": request})
|
| 159 |
+
|
| 160 |
+
@app.get("/api/providers")
|
| 161 |
+
async def get_providers_data():
|
| 162 |
+
"""API endpoint to get provider data"""
|
| 163 |
+
async with aiohttp.ClientSession() as session:
|
| 164 |
+
tasks = [get_monthly_requests(session, provider) for provider in PROVIDERS]
|
| 165 |
+
results = await asyncio.gather(*tasks)
|
| 166 |
+
|
| 167 |
+
# Sort by request count descending
|
| 168 |
+
results.sort(key=lambda x: x["monthly_requests_int"], reverse=True)
|
| 169 |
+
|
| 170 |
+
return {
|
| 171 |
+
"providers": results,
|
| 172 |
+
"last_updated": datetime.now().strftime('%Y-%m-%d %H:%M:%S'),
|
| 173 |
+
"total_providers": len(results)
|
| 174 |
+
}
|
| 175 |
+
|
| 176 |
+
@app.get("/api/providers/{provider}")
|
| 177 |
+
async def get_provider_data(provider: str):
|
| 178 |
+
"""API endpoint to get data for a specific provider"""
|
| 179 |
+
if provider not in PROVIDERS:
|
| 180 |
+
return {"error": "Provider not found"}
|
| 181 |
+
|
| 182 |
+
async with aiohttp.ClientSession() as session:
|
| 183 |
+
result = await get_monthly_requests(session, provider)
|
| 184 |
+
|
| 185 |
+
return {
|
| 186 |
+
"provider_data": result,
|
| 187 |
+
"last_updated": datetime.now().strftime('%Y-%m-%d %H:%M:%S')
|
| 188 |
+
}
|
| 189 |
+
|
| 190 |
+
@app.get("/api/historical")
|
| 191 |
+
async def get_historical_data():
|
| 192 |
+
"""API endpoint to get historical data for line chart"""
|
| 193 |
+
if not HF_TOKEN:
|
| 194 |
+
return {"error": "Historical data not available", "data": []}
|
| 195 |
+
|
| 196 |
+
try:
|
| 197 |
+
# Load historical dataset
|
| 198 |
+
dataset = load_dataset(DATASET_REPO_NAME, split="train")
|
| 199 |
+
df = dataset.to_pandas()
|
| 200 |
+
|
| 201 |
+
# Group by timestamp and provider, get the latest entry for each timestamp-provider combo
|
| 202 |
+
df['timestamp'] = pd.to_datetime(df['timestamp'])
|
| 203 |
+
df = df.sort_values('timestamp')
|
| 204 |
+
|
| 205 |
+
# Get last 48 hours of data (48 data points max for performance)
|
| 206 |
+
cutoff_time = datetime.now(timezone.utc) - pd.Timedelta(hours=48)
|
| 207 |
+
df = df[df['timestamp'] >= cutoff_time]
|
| 208 |
+
|
| 209 |
+
# Prepare data for Chart.js line chart
|
| 210 |
+
historical_data = {}
|
| 211 |
+
|
| 212 |
+
for provider in PROVIDERS:
|
| 213 |
+
provider_data = df[df['provider'] == provider].copy()
|
| 214 |
+
if not provider_data.empty:
|
| 215 |
+
# Format for Chart.js: {x: timestamp, y: value}
|
| 216 |
+
historical_data[provider] = [
|
| 217 |
+
{
|
| 218 |
+
"x": row['timestamp'].isoformat(),
|
| 219 |
+
"y": row['monthly_requests_int']
|
| 220 |
+
}
|
| 221 |
+
for _, row in provider_data.iterrows()
|
| 222 |
+
]
|
| 223 |
+
else:
|
| 224 |
+
historical_data[provider] = []
|
| 225 |
+
|
| 226 |
+
return {
|
| 227 |
+
"historical_data": historical_data,
|
| 228 |
+
"last_updated": datetime.now().strftime('%Y-%m-%d %H:%M:%S')
|
| 229 |
+
}
|
| 230 |
+
|
| 231 |
+
except Exception as e:
|
| 232 |
+
logger.error(f"Error fetching historical data: {e}")
|
| 233 |
+
return {"error": "Failed to fetch historical data", "data": []}
|
| 234 |
+
|
| 235 |
+
@app.post("/api/collect-now")
|
| 236 |
+
async def trigger_data_collection(background_tasks: BackgroundTasks):
|
| 237 |
+
"""Manual trigger for data collection"""
|
| 238 |
+
background_tasks.add_task(collect_and_store_data)
|
| 239 |
+
return {"message": "Data collection triggered", "timestamp": datetime.now().isoformat()}
|
| 240 |
+
|
| 241 |
+
if __name__ == "__main__":
|
| 242 |
+
uvicorn.run(app, host="0.0.0.0", port=7860)
|