LLM/docling-serve
1
0
Fork 0
mirror of https://github.com/docling-project/docling-serve.git synced 2026-03-08 06:43:07 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix: race condition in Gradio UI task result retrieval (#454)

Browse Source 
Signed-off-by: tumf <y.takahara@gmail.com>
This commit is contained in:
tumf
2025-12-15 22:29:50 +07:00
committed by GitHub
parent 0ec67a37b7
commit a179338c78
1 changed files with 68 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

80
docling_serve/gradio_ui.py
View File

@@ -256,7 +256,26 @@ def wait_task_finish(auth: str, task_id: str, return_as_file: bool):
verify=ssl_ctx,
timeout=15,
)
task_status = response.json()["task_status"]
# Check response status code first
if response.status_code == 404:
logger.warning(
f"Task {task_id} not found in status poll, it may have completed already"
)
time.sleep(2) # Wait for result to be ready
conversion_sucess = True
task_finished = True
break
response.raise_for_status()
# Safely access task_status
response_data = response.json()
if "task_status" not in response_data:
logger.error(f"Missing task_status in response: {response_data}")
raise RuntimeError("Missing task_status in response")
task_status = response_data["task_status"]
if task_status == "success":
conversion_sucess = True
task_finished = True
@@ -272,18 +291,55 @@ def wait_task_finish(auth: str, task_id: str, return_as_file: bool):
task_finished = True
raise gr.Error(f"Error processing file(s): {e}", print_exception=False)
# Retry logic for result retrieval
if conversion_sucess:
try:
response = httpx.get(
f"{get_api_endpoint()}/v1/result/{task_id}",
headers=headers,
timeout=15,
verify=ssl_ctx,
)
output = response_to_output(response, return_as_file)
return output
except Exception as e:
logger.error(f"Error getting task result: {e}")
max_retries = 3
retry_count = 0
while retry_count < max_retries:
try:
response = httpx.get(
f"{get_api_endpoint()}/v1/result/{task_id}",
headers=headers,
timeout=15,
verify=ssl_ctx,
)
if response.status_code == 404:
retry_count += 1
if retry_count < max_retries:
wait_time = 2**retry_count # Exponential backoff: 2, 4, 8s
logger.warning(
f"Result not ready yet, retrying in {wait_time}s "
f"(attempt {retry_count}/{max_retries})"
)
time.sleep(wait_time)
continue
else:
logger.error(
f"Result not available after {max_retries} retries"
)
raise RuntimeError(
f"Result not available after {max_retries} retries"
)
response.raise_for_status()
output = response_to_output(response, return_as_file)
return output
except Exception as e:
if retry_count >= max_retries - 1:
logger.error(f"Error getting task result: {e}")
raise gr.Error(
f"Error getting task result: {e}", print_exception=False
)
# For non-404 errors on early retries, continue retrying
retry_count += 1
if retry_count < max_retries:
wait_time = 2**retry_count
logger.warning(
f"Error getting result, retrying in {wait_time}s: {e}"
)
time.sleep(wait_time)
raise gr.Error(
f"Error getting task result, conversion finished with status: {task_status}"