Spaces:
Running
Running
Update xai_client.py
Browse files- xai_client.py +78 -12
xai_client.py
CHANGED
|
@@ -15,7 +15,8 @@ DEFAULT_IMAGE_MODEL = "grok-imagine-image"
|
|
| 15 |
DEFAULT_VIDEO_MODEL = "grok-imagine-video"
|
| 16 |
|
| 17 |
IMAGE_ASPECT_RATIOS = [
|
| 18 |
-
"auto", "1:1", "16:9", "9:16", "4:3", "3:4", "3:2", "2:3",
|
|
|
|
| 19 |
]
|
| 20 |
IMAGE_RESOLUTIONS = ["1k", "2k"]
|
| 21 |
VIDEO_ASPECT_RATIOS = ["16:9", "9:16", "1:1", "4:3", "3:4", "3:2", "2:3"]
|
|
@@ -64,12 +65,14 @@ def get_space_base_url(request: gr.Request | None) -> str | None:
|
|
| 64 |
space_host = os.getenv("SPACE_HOST")
|
| 65 |
if space_host:
|
| 66 |
return f"https://{space_host}".rstrip("/")
|
|
|
|
| 67 |
if request is not None:
|
| 68 |
headers = getattr(request, "headers", None) or {}
|
| 69 |
host = headers.get("x-forwarded-host") or headers.get("host")
|
| 70 |
proto = headers.get("x-forwarded-proto") or "https"
|
| 71 |
if host:
|
| 72 |
return f"{proto}://{host}".rstrip("/")
|
|
|
|
| 73 |
return None
|
| 74 |
|
| 75 |
|
|
@@ -84,22 +87,25 @@ def list_xai_models(api_key: str):
|
|
| 84 |
headers = auth_headers(api_key)
|
| 85 |
image_models = [DEFAULT_IMAGE_MODEL]
|
| 86 |
video_models = [DEFAULT_VIDEO_MODEL]
|
|
|
|
| 87 |
try:
|
| 88 |
r = requests.get(f"{API_BASE}/image-generation-models", headers=headers, timeout=60)
|
| 89 |
if r.ok:
|
| 90 |
image_models = [m["id"] for m in r.json().get("models", []) if m.get("id")] or image_models
|
| 91 |
except Exception:
|
| 92 |
pass
|
|
|
|
| 93 |
try:
|
| 94 |
r = requests.get(f"{API_BASE}/video-generation-models", headers=headers, timeout=60)
|
| 95 |
if r.ok:
|
| 96 |
video_models = [m["id"] for m in r.json().get("models", []) if m.get("id")] or video_models
|
| 97 |
except Exception:
|
| 98 |
pass
|
|
|
|
| 99 |
return (
|
| 100 |
gr.update(choices=image_models, value=image_models[0]),
|
| 101 |
gr.update(choices=video_models, value=video_models[0]),
|
| 102 |
-
f"Loaded {len(image_models)} image model(s) and {len(video_models)} video model(s)."
|
| 103 |
)
|
| 104 |
|
| 105 |
|
|
@@ -111,16 +117,19 @@ def generate_t2i(api_key, model, prompt, n, aspect_ratio, resolution, progress=g
|
|
| 111 |
"n": int(n),
|
| 112 |
"response_format": "b64_json",
|
| 113 |
}
|
|
|
|
| 114 |
if not payload["prompt"]:
|
| 115 |
raise gr.Error("Please enter a prompt.")
|
| 116 |
if aspect_ratio:
|
| 117 |
payload["aspect_ratio"] = aspect_ratio
|
| 118 |
if resolution:
|
| 119 |
payload["resolution"] = resolution
|
|
|
|
| 120 |
progress(0.2, desc="Generating images...")
|
| 121 |
resp = requests.post(f"{API_BASE}/images/generations", headers=headers, json=payload, timeout=300)
|
| 122 |
if not resp.ok:
|
| 123 |
raise gr.Error(f"xAI image generation failed:\n{safe_json_error(resp)}")
|
|
|
|
| 124 |
gallery, paths = [], []
|
| 125 |
for i, item in enumerate(resp.json().get("data", []), start=1):
|
| 126 |
if item.get("b64_json"):
|
|
@@ -129,35 +138,47 @@ def generate_t2i(api_key, model, prompt, n, aspect_ratio, resolution, progress=g
|
|
| 129 |
out = download_url_to_temp(item["url"], ".png")
|
| 130 |
else:
|
| 131 |
continue
|
|
|
|
| 132 |
paths.append(out)
|
| 133 |
gallery.append((out, f"Image {i}"))
|
|
|
|
| 134 |
if not paths:
|
| 135 |
raise gr.Error("xAI returned no images.")
|
|
|
|
| 136 |
progress(1.0, desc="Done")
|
| 137 |
return gallery, paths[0], paths, f"Generated {len(paths)} image(s)."
|
| 138 |
|
| 139 |
|
| 140 |
def edit_like_i2i(api_key, model, prompt, input_image_path, aspect_ratio, progress=gr.Progress(track_tqdm=False)):
|
| 141 |
headers = auth_headers(api_key)
|
|
|
|
| 142 |
if not input_image_path:
|
| 143 |
raise gr.Error("Please upload an image.")
|
| 144 |
if not (prompt or "").strip():
|
| 145 |
raise gr.Error("Please enter a prompt.")
|
|
|
|
| 146 |
payload = {
|
| 147 |
"model": model or DEFAULT_IMAGE_MODEL,
|
| 148 |
"prompt": prompt.strip(),
|
| 149 |
-
"image": {
|
|
|
|
|
|
|
|
|
|
| 150 |
"response_format": "b64_json",
|
| 151 |
}
|
|
|
|
| 152 |
if aspect_ratio:
|
| 153 |
payload["aspect_ratio"] = aspect_ratio
|
|
|
|
| 154 |
progress(0.2, desc="Editing image...")
|
| 155 |
resp = requests.post(f"{API_BASE}/images/edits", headers=headers, json=payload, timeout=300)
|
| 156 |
if not resp.ok:
|
| 157 |
raise gr.Error(f"xAI image edit failed:\n{safe_json_error(resp)}")
|
|
|
|
| 158 |
data = resp.json().get("data", [])
|
| 159 |
if not data:
|
| 160 |
raise gr.Error("xAI returned no image.")
|
|
|
|
| 161 |
item = data[0]
|
| 162 |
if item.get("b64_json"):
|
| 163 |
out = download_bytes_to_temp(base64.b64decode(item["b64_json"]), ".png")
|
|
@@ -165,6 +186,7 @@ def edit_like_i2i(api_key, model, prompt, input_image_path, aspect_ratio, progre
|
|
| 165 |
out = download_url_to_temp(item["url"], ".png")
|
| 166 |
else:
|
| 167 |
raise gr.Error("xAI returned no output image payload.")
|
|
|
|
| 168 |
progress(1.0, desc="Done")
|
| 169 |
return out, out, f"Completed: {Path(out).name}"
|
| 170 |
|
|
@@ -172,29 +194,50 @@ def edit_like_i2i(api_key, model, prompt, input_image_path, aspect_ratio, progre
|
|
| 172 |
def poll_video_result(api_key, request_id, timeout_seconds, poll_interval, progress):
|
| 173 |
headers = {"Authorization": f"Bearer {api_key.strip()}"}
|
| 174 |
started = time.time()
|
|
|
|
| 175 |
while True:
|
| 176 |
if time.time() - started > timeout_seconds:
|
| 177 |
raise gr.Error("Timed out waiting for xAI video generation.")
|
|
|
|
| 178 |
resp = requests.get(f"{API_BASE}/videos/{request_id}", headers=headers, timeout=120)
|
| 179 |
if not resp.ok:
|
| 180 |
raise gr.Error(f"xAI video polling failed:\n{safe_json_error(resp)}")
|
|
|
|
| 181 |
data = resp.json()
|
| 182 |
status = data.get("status", "unknown")
|
| 183 |
progress(None, desc=f"Video status: {status}")
|
|
|
|
| 184 |
if status == "done":
|
| 185 |
-
|
|
|
|
| 186 |
if not video_url:
|
| 187 |
raise gr.Error("xAI returned no video URL.")
|
| 188 |
-
return video_url,
|
| 189 |
-
|
| 190 |
-
|
|
|
|
|
|
|
| 191 |
time.sleep(int(poll_interval))
|
| 192 |
|
| 193 |
|
| 194 |
-
def generate_i2v(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 195 |
source_image = last_t2i_first_image if use_last_t2i_image and last_t2i_first_image else uploaded_image_path
|
| 196 |
if not source_image:
|
| 197 |
raise gr.Error("Upload an image or use the first T2I result.")
|
|
|
|
| 198 |
headers = auth_headers(api_key)
|
| 199 |
payload = {
|
| 200 |
"model": model or DEFAULT_VIDEO_MODEL,
|
|
@@ -203,43 +246,66 @@ def generate_i2v(api_key, model, prompt, uploaded_image_path, use_last_t2i_image
|
|
| 203 |
"duration": int(duration),
|
| 204 |
"resolution": resolution,
|
| 205 |
}
|
|
|
|
| 206 |
if not payload["prompt"]:
|
| 207 |
raise gr.Error("Please enter an I2V prompt.")
|
| 208 |
if aspect_ratio:
|
| 209 |
payload["aspect_ratio"] = aspect_ratio
|
|
|
|
| 210 |
progress(0.2, desc="Submitting I2V...")
|
| 211 |
resp = requests.post(f"{API_BASE}/videos/generations", headers=headers, json=payload, timeout=300)
|
| 212 |
if not resp.ok:
|
| 213 |
raise gr.Error(f"xAI I2V request failed:\n{safe_json_error(resp)}")
|
|
|
|
| 214 |
request_id = resp.json().get("request_id")
|
| 215 |
if not request_id:
|
| 216 |
raise gr.Error("xAI did not return request_id.")
|
| 217 |
-
|
|
|
|
|
|
|
|
|
|
| 218 |
out = download_url_to_temp(video_url, ".mp4")
|
| 219 |
return out, out, f"I2V complete. Request ID: {request_id}. Duration: {actual_duration}s"
|
| 220 |
|
| 221 |
|
| 222 |
-
def generate_v2v(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 223 |
if not uploaded_video_path:
|
| 224 |
raise gr.Error("Please upload an MP4 source video.")
|
| 225 |
if Path(uploaded_video_path).suffix.lower() != ".mp4":
|
| 226 |
raise gr.Error("xAI V2V expects .mp4 input.")
|
|
|
|
| 227 |
headers = auth_headers(api_key)
|
| 228 |
public_video_url = local_file_to_public_url(uploaded_video_path, request)
|
|
|
|
| 229 |
payload = {
|
| 230 |
"model": model or DEFAULT_VIDEO_MODEL,
|
| 231 |
"prompt": (prompt or "").strip(),
|
| 232 |
"video_url": public_video_url,
|
| 233 |
}
|
|
|
|
| 234 |
if not payload["prompt"]:
|
| 235 |
raise gr.Error("Please enter a V2V prompt.")
|
|
|
|
| 236 |
progress(0.2, desc="Submitting V2V...")
|
| 237 |
resp = requests.post(f"{API_BASE}/videos/generations", headers=headers, json=payload, timeout=300)
|
| 238 |
if not resp.ok:
|
| 239 |
raise gr.Error(f"xAI V2V request failed:\n{safe_json_error(resp)}")
|
|
|
|
| 240 |
request_id = resp.json().get("request_id")
|
| 241 |
if not request_id:
|
| 242 |
raise gr.Error("xAI did not return request_id.")
|
| 243 |
-
|
|
|
|
|
|
|
|
|
|
| 244 |
out = download_url_to_temp(video_url, ".mp4")
|
| 245 |
-
return out, out, f"V2V complete. Request ID: {request_id}. Duration: {actual_duration}s"
|
|
|
|
| 15 |
DEFAULT_VIDEO_MODEL = "grok-imagine-video"
|
| 16 |
|
| 17 |
IMAGE_ASPECT_RATIOS = [
|
| 18 |
+
"auto", "1:1", "16:9", "9:16", "4:3", "3:4", "3:2", "2:3",
|
| 19 |
+
"2:1", "1:2", "19.5:9", "9:19.5", "20:9", "9:20"
|
| 20 |
]
|
| 21 |
IMAGE_RESOLUTIONS = ["1k", "2k"]
|
| 22 |
VIDEO_ASPECT_RATIOS = ["16:9", "9:16", "1:1", "4:3", "3:4", "3:2", "2:3"]
|
|
|
|
| 65 |
space_host = os.getenv("SPACE_HOST")
|
| 66 |
if space_host:
|
| 67 |
return f"https://{space_host}".rstrip("/")
|
| 68 |
+
|
| 69 |
if request is not None:
|
| 70 |
headers = getattr(request, "headers", None) or {}
|
| 71 |
host = headers.get("x-forwarded-host") or headers.get("host")
|
| 72 |
proto = headers.get("x-forwarded-proto") or "https"
|
| 73 |
if host:
|
| 74 |
return f"{proto}://{host}".rstrip("/")
|
| 75 |
+
|
| 76 |
return None
|
| 77 |
|
| 78 |
|
|
|
|
| 87 |
headers = auth_headers(api_key)
|
| 88 |
image_models = [DEFAULT_IMAGE_MODEL]
|
| 89 |
video_models = [DEFAULT_VIDEO_MODEL]
|
| 90 |
+
|
| 91 |
try:
|
| 92 |
r = requests.get(f"{API_BASE}/image-generation-models", headers=headers, timeout=60)
|
| 93 |
if r.ok:
|
| 94 |
image_models = [m["id"] for m in r.json().get("models", []) if m.get("id")] or image_models
|
| 95 |
except Exception:
|
| 96 |
pass
|
| 97 |
+
|
| 98 |
try:
|
| 99 |
r = requests.get(f"{API_BASE}/video-generation-models", headers=headers, timeout=60)
|
| 100 |
if r.ok:
|
| 101 |
video_models = [m["id"] for m in r.json().get("models", []) if m.get("id")] or video_models
|
| 102 |
except Exception:
|
| 103 |
pass
|
| 104 |
+
|
| 105 |
return (
|
| 106 |
gr.update(choices=image_models, value=image_models[0]),
|
| 107 |
gr.update(choices=video_models, value=video_models[0]),
|
| 108 |
+
f"Loaded {len(image_models)} image model(s) and {len(video_models)} video model(s).",
|
| 109 |
)
|
| 110 |
|
| 111 |
|
|
|
|
| 117 |
"n": int(n),
|
| 118 |
"response_format": "b64_json",
|
| 119 |
}
|
| 120 |
+
|
| 121 |
if not payload["prompt"]:
|
| 122 |
raise gr.Error("Please enter a prompt.")
|
| 123 |
if aspect_ratio:
|
| 124 |
payload["aspect_ratio"] = aspect_ratio
|
| 125 |
if resolution:
|
| 126 |
payload["resolution"] = resolution
|
| 127 |
+
|
| 128 |
progress(0.2, desc="Generating images...")
|
| 129 |
resp = requests.post(f"{API_BASE}/images/generations", headers=headers, json=payload, timeout=300)
|
| 130 |
if not resp.ok:
|
| 131 |
raise gr.Error(f"xAI image generation failed:\n{safe_json_error(resp)}")
|
| 132 |
+
|
| 133 |
gallery, paths = [], []
|
| 134 |
for i, item in enumerate(resp.json().get("data", []), start=1):
|
| 135 |
if item.get("b64_json"):
|
|
|
|
| 138 |
out = download_url_to_temp(item["url"], ".png")
|
| 139 |
else:
|
| 140 |
continue
|
| 141 |
+
|
| 142 |
paths.append(out)
|
| 143 |
gallery.append((out, f"Image {i}"))
|
| 144 |
+
|
| 145 |
if not paths:
|
| 146 |
raise gr.Error("xAI returned no images.")
|
| 147 |
+
|
| 148 |
progress(1.0, desc="Done")
|
| 149 |
return gallery, paths[0], paths, f"Generated {len(paths)} image(s)."
|
| 150 |
|
| 151 |
|
| 152 |
def edit_like_i2i(api_key, model, prompt, input_image_path, aspect_ratio, progress=gr.Progress(track_tqdm=False)):
|
| 153 |
headers = auth_headers(api_key)
|
| 154 |
+
|
| 155 |
if not input_image_path:
|
| 156 |
raise gr.Error("Please upload an image.")
|
| 157 |
if not (prompt or "").strip():
|
| 158 |
raise gr.Error("Please enter a prompt.")
|
| 159 |
+
|
| 160 |
payload = {
|
| 161 |
"model": model or DEFAULT_IMAGE_MODEL,
|
| 162 |
"prompt": prompt.strip(),
|
| 163 |
+
"image": {
|
| 164 |
+
"url": file_to_data_uri(input_image_path),
|
| 165 |
+
"type": "image_url",
|
| 166 |
+
},
|
| 167 |
"response_format": "b64_json",
|
| 168 |
}
|
| 169 |
+
|
| 170 |
if aspect_ratio:
|
| 171 |
payload["aspect_ratio"] = aspect_ratio
|
| 172 |
+
|
| 173 |
progress(0.2, desc="Editing image...")
|
| 174 |
resp = requests.post(f"{API_BASE}/images/edits", headers=headers, json=payload, timeout=300)
|
| 175 |
if not resp.ok:
|
| 176 |
raise gr.Error(f"xAI image edit failed:\n{safe_json_error(resp)}")
|
| 177 |
+
|
| 178 |
data = resp.json().get("data", [])
|
| 179 |
if not data:
|
| 180 |
raise gr.Error("xAI returned no image.")
|
| 181 |
+
|
| 182 |
item = data[0]
|
| 183 |
if item.get("b64_json"):
|
| 184 |
out = download_bytes_to_temp(base64.b64decode(item["b64_json"]), ".png")
|
|
|
|
| 186 |
out = download_url_to_temp(item["url"], ".png")
|
| 187 |
else:
|
| 188 |
raise gr.Error("xAI returned no output image payload.")
|
| 189 |
+
|
| 190 |
progress(1.0, desc="Done")
|
| 191 |
return out, out, f"Completed: {Path(out).name}"
|
| 192 |
|
|
|
|
| 194 |
def poll_video_result(api_key, request_id, timeout_seconds, poll_interval, progress):
|
| 195 |
headers = {"Authorization": f"Bearer {api_key.strip()}"}
|
| 196 |
started = time.time()
|
| 197 |
+
|
| 198 |
while True:
|
| 199 |
if time.time() - started > timeout_seconds:
|
| 200 |
raise gr.Error("Timed out waiting for xAI video generation.")
|
| 201 |
+
|
| 202 |
resp = requests.get(f"{API_BASE}/videos/{request_id}", headers=headers, timeout=120)
|
| 203 |
if not resp.ok:
|
| 204 |
raise gr.Error(f"xAI video polling failed:\n{safe_json_error(resp)}")
|
| 205 |
+
|
| 206 |
data = resp.json()
|
| 207 |
status = data.get("status", "unknown")
|
| 208 |
progress(None, desc=f"Video status: {status}")
|
| 209 |
+
|
| 210 |
if status == "done":
|
| 211 |
+
video = data.get("video") or {}
|
| 212 |
+
video_url = video.get("url")
|
| 213 |
if not video_url:
|
| 214 |
raise gr.Error("xAI returned no video URL.")
|
| 215 |
+
return video_url, video.get("duration"), data
|
| 216 |
+
|
| 217 |
+
if status in {"failed", "error", "cancelled", "expired"}:
|
| 218 |
+
raise gr.Error(f"xAI video job ended with status: {status}\n{json.dumps(data, indent=2)}")
|
| 219 |
+
|
| 220 |
time.sleep(int(poll_interval))
|
| 221 |
|
| 222 |
|
| 223 |
+
def generate_i2v(
|
| 224 |
+
api_key,
|
| 225 |
+
model,
|
| 226 |
+
prompt,
|
| 227 |
+
uploaded_image_path,
|
| 228 |
+
use_last_t2i_image,
|
| 229 |
+
last_t2i_first_image,
|
| 230 |
+
duration,
|
| 231 |
+
aspect_ratio,
|
| 232 |
+
resolution,
|
| 233 |
+
timeout_seconds,
|
| 234 |
+
poll_interval,
|
| 235 |
+
progress=gr.Progress(track_tqdm=False),
|
| 236 |
+
):
|
| 237 |
source_image = last_t2i_first_image if use_last_t2i_image and last_t2i_first_image else uploaded_image_path
|
| 238 |
if not source_image:
|
| 239 |
raise gr.Error("Upload an image or use the first T2I result.")
|
| 240 |
+
|
| 241 |
headers = auth_headers(api_key)
|
| 242 |
payload = {
|
| 243 |
"model": model or DEFAULT_VIDEO_MODEL,
|
|
|
|
| 246 |
"duration": int(duration),
|
| 247 |
"resolution": resolution,
|
| 248 |
}
|
| 249 |
+
|
| 250 |
if not payload["prompt"]:
|
| 251 |
raise gr.Error("Please enter an I2V prompt.")
|
| 252 |
if aspect_ratio:
|
| 253 |
payload["aspect_ratio"] = aspect_ratio
|
| 254 |
+
|
| 255 |
progress(0.2, desc="Submitting I2V...")
|
| 256 |
resp = requests.post(f"{API_BASE}/videos/generations", headers=headers, json=payload, timeout=300)
|
| 257 |
if not resp.ok:
|
| 258 |
raise gr.Error(f"xAI I2V request failed:\n{safe_json_error(resp)}")
|
| 259 |
+
|
| 260 |
request_id = resp.json().get("request_id")
|
| 261 |
if not request_id:
|
| 262 |
raise gr.Error("xAI did not return request_id.")
|
| 263 |
+
|
| 264 |
+
video_url, actual_duration, _ = poll_video_result(
|
| 265 |
+
api_key, request_id, int(timeout_seconds), int(poll_interval), progress
|
| 266 |
+
)
|
| 267 |
out = download_url_to_temp(video_url, ".mp4")
|
| 268 |
return out, out, f"I2V complete. Request ID: {request_id}. Duration: {actual_duration}s"
|
| 269 |
|
| 270 |
|
| 271 |
+
def generate_v2v(
|
| 272 |
+
api_key,
|
| 273 |
+
model,
|
| 274 |
+
prompt,
|
| 275 |
+
uploaded_video_path,
|
| 276 |
+
timeout_seconds,
|
| 277 |
+
poll_interval,
|
| 278 |
+
request: gr.Request,
|
| 279 |
+
progress=gr.Progress(track_tqdm=False),
|
| 280 |
+
):
|
| 281 |
if not uploaded_video_path:
|
| 282 |
raise gr.Error("Please upload an MP4 source video.")
|
| 283 |
if Path(uploaded_video_path).suffix.lower() != ".mp4":
|
| 284 |
raise gr.Error("xAI V2V expects .mp4 input.")
|
| 285 |
+
|
| 286 |
headers = auth_headers(api_key)
|
| 287 |
public_video_url = local_file_to_public_url(uploaded_video_path, request)
|
| 288 |
+
|
| 289 |
payload = {
|
| 290 |
"model": model or DEFAULT_VIDEO_MODEL,
|
| 291 |
"prompt": (prompt or "").strip(),
|
| 292 |
"video_url": public_video_url,
|
| 293 |
}
|
| 294 |
+
|
| 295 |
if not payload["prompt"]:
|
| 296 |
raise gr.Error("Please enter a V2V prompt.")
|
| 297 |
+
|
| 298 |
progress(0.2, desc="Submitting V2V...")
|
| 299 |
resp = requests.post(f"{API_BASE}/videos/generations", headers=headers, json=payload, timeout=300)
|
| 300 |
if not resp.ok:
|
| 301 |
raise gr.Error(f"xAI V2V request failed:\n{safe_json_error(resp)}")
|
| 302 |
+
|
| 303 |
request_id = resp.json().get("request_id")
|
| 304 |
if not request_id:
|
| 305 |
raise gr.Error("xAI did not return request_id.")
|
| 306 |
+
|
| 307 |
+
video_url, actual_duration, _ = poll_video_result(
|
| 308 |
+
api_key, request_id, int(timeout_seconds), int(poll_interval), progress
|
| 309 |
+
)
|
| 310 |
out = download_url_to_temp(video_url, ".mp4")
|
| 311 |
+
return out, out, f"V2V complete. Request ID: {request_id}. Duration: {actual_duration}s"
|