Spaces:
Running on Zero
Running on Zero
Use /respite/ prefix to avoid Gradio proxy conflict
Browse files
app.py
CHANGED
|
@@ -7,7 +7,6 @@ import torch
|
|
| 7 |
import torchaudio
|
| 8 |
import gradio as gr
|
| 9 |
import spaces
|
| 10 |
-
from fastapi import FastAPI
|
| 11 |
from fastapi.responses import JSONResponse
|
| 12 |
from einops import rearrange
|
| 13 |
from huggingface_hub import login
|
|
@@ -79,11 +78,11 @@ API_SPECS = {
|
|
| 79 |
"name": "Respite API",
|
| 80 |
"version": "1.0.0",
|
| 81 |
"description": "General-purpose AI API server with audio generation capabilities.",
|
| 82 |
-
"base_path": "/
|
| 83 |
"authentication": "none",
|
| 84 |
"content_types": ["application/json", "audio/wav"],
|
| 85 |
-
"resources_endpoint": "/
|
| 86 |
-
"specs_endpoint": "/
|
| 87 |
"limits": {
|
| 88 |
"max_concurrent_requests": 1,
|
| 89 |
"max_queue_size": 4,
|
|
@@ -105,23 +104,6 @@ def _get_runtime_specs():
|
|
| 105 |
}
|
| 106 |
|
| 107 |
|
| 108 |
-
# ---------------------------------------------------------------------------
|
| 109 |
-
# FastAPI application with discovery routes
|
| 110 |
-
# ---------------------------------------------------------------------------
|
| 111 |
-
|
| 112 |
-
api = FastAPI(title=API_SPECS["name"], version=API_SPECS["version"])
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
@api.get("/api/resources")
|
| 116 |
-
def resources():
|
| 117 |
-
return JSONResponse({"resources": API_RESOURCES})
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
@api.get("/api/specs")
|
| 121 |
-
def specs():
|
| 122 |
-
return JSONResponse({**API_SPECS, "runtime": _get_runtime_specs()})
|
| 123 |
-
|
| 124 |
-
|
| 125 |
# ---------------------------------------------------------------------------
|
| 126 |
# Model cache
|
| 127 |
# ---------------------------------------------------------------------------
|
|
@@ -219,21 +201,42 @@ with gr.Blocks(title="Respite API") as demo:
|
|
| 219 |
|
| 220 |
|
| 221 |
# ---------------------------------------------------------------------------
|
| 222 |
-
#
|
| 223 |
#
|
| 224 |
-
#
|
| 225 |
-
#
|
| 226 |
-
#
|
| 227 |
# ---------------------------------------------------------------------------
|
| 228 |
|
| 229 |
demo.queue(max_size=4, default_concurrency_limit=1)
|
| 230 |
|
| 231 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 232 |
|
|
|
|
|
|
|
|
|
|
| 233 |
|
| 234 |
-
|
| 235 |
-
|
| 236 |
-
|
| 237 |
|
|
|
|
| 238 |
|
| 239 |
-
|
|
|
|
| 7 |
import torchaudio
|
| 8 |
import gradio as gr
|
| 9 |
import spaces
|
|
|
|
| 10 |
from fastapi.responses import JSONResponse
|
| 11 |
from einops import rearrange
|
| 12 |
from huggingface_hub import login
|
|
|
|
| 78 |
"name": "Respite API",
|
| 79 |
"version": "1.0.0",
|
| 80 |
"description": "General-purpose AI API server with audio generation capabilities.",
|
| 81 |
+
"base_path": "/respite",
|
| 82 |
"authentication": "none",
|
| 83 |
"content_types": ["application/json", "audio/wav"],
|
| 84 |
+
"resources_endpoint": "/respite/resources",
|
| 85 |
+
"specs_endpoint": "/respite/specs",
|
| 86 |
"limits": {
|
| 87 |
"max_concurrent_requests": 1,
|
| 88 |
"max_queue_size": 4,
|
|
|
|
| 104 |
}
|
| 105 |
|
| 106 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 107 |
# ---------------------------------------------------------------------------
|
| 108 |
# Model cache
|
| 109 |
# ---------------------------------------------------------------------------
|
|
|
|
| 201 |
|
| 202 |
|
| 203 |
# ---------------------------------------------------------------------------
|
| 204 |
+
# API discovery endpoints
|
| 205 |
#
|
| 206 |
+
# Use /respite/ prefix instead of /api/ to avoid collision with Gradio's
|
| 207 |
+
# internal /api proxy which intercepts all /api/* requests in the
|
| 208 |
+
# sdk:gradio Spaces runtime.
|
| 209 |
# ---------------------------------------------------------------------------
|
| 210 |
|
| 211 |
demo.queue(max_size=4, default_concurrency_limit=1)
|
| 212 |
|
| 213 |
+
# These decorators run at import time. demo.app is the FastAPI instance
|
| 214 |
+
# underlying the Gradio Blocks. Routes registered here will be served
|
| 215 |
+
# alongside the Gradio UI once the server starts.
|
| 216 |
+
# NOTE: demo.app is created lazily by Gradio, so we use a startup hook
|
| 217 |
+
# to register routes after the server is ready.
|
| 218 |
+
import threading
|
| 219 |
+
|
| 220 |
+
def _register_routes():
|
| 221 |
+
# Wait for demo.app to be available (created during launch)
|
| 222 |
+
import time
|
| 223 |
+
for _ in range(30):
|
| 224 |
+
if hasattr(demo, "app") and demo.app is not None:
|
| 225 |
+
break
|
| 226 |
+
time.sleep(1)
|
| 227 |
+
|
| 228 |
+
if not hasattr(demo, "app") or demo.app is None:
|
| 229 |
+
_log("Warning: demo.app not available, API routes not registered")
|
| 230 |
+
return
|
| 231 |
|
| 232 |
+
@demo.app.get("/respite/resources")
|
| 233 |
+
def resources():
|
| 234 |
+
return JSONResponse({"resources": API_RESOURCES})
|
| 235 |
|
| 236 |
+
@demo.app.get("/respite/specs")
|
| 237 |
+
def specs():
|
| 238 |
+
return JSONResponse({**API_SPECS, "runtime": _get_runtime_specs()})
|
| 239 |
|
| 240 |
+
_log("API routes registered at /respite/resources and /respite/specs")
|
| 241 |
|
| 242 |
+
threading.Thread(target=_register_routes, daemon=True).start()
|