apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-23 00:17:25 +08:00
Code Issues Actions 19 Packages Projects Releases Wiki Activity

[Optimization]update request in tool parser&reasoning parser (#6858)

Browse Source 
* update request in tool parser&reasoning parser
This commit is contained in:
luukunn
2026-03-17 11:51:12 +08:00
committed by GitHub
parent 4ed483d20b
commit fe8d58a094
9 changed files with 39 additions and 22 deletions
Download Patch File Download Diff File Expand all files Collapse all files
fastdeploy/entrypoints/openai/response_processors.py
+8 -1
View File
@@ -72,7 +72,7 @@ class ChatResponseProcessor:
else:
self._multipart_buffer.append({"decode_type": decode_type, "request_output": request_output})
async def process_response_chat(self, request_outputs, stream, include_stop_str_in_output):
async def process_response_chat(self, request_outputs, stream, include_stop_str_in_output, request):
"""
Process a list of responses into a generator that yields each processed response as it's generated.
Args:
@@ -101,6 +101,7 @@ class ChatResponseProcessor:
include_stop_str_in_output=include_stop_str_in_output,
audio_tokens=all_audio_tokens,
tts=tts,
request=request,
)
else:
response = self.data_processor.process_response_dict(
@@ -109,6 +110,7 @@ class ChatResponseProcessor:
include_stop_str_in_output=include_stop_str_in_output,
audio_tokens=all_audio_tokens,
tts=tts,
request=request,
)
yield response
elif decode_type == 2: # audio
@@ -125,6 +127,7 @@ class ChatResponseProcessor:
response_dict=request_output,
stream=stream,
include_stop_str_in_output=include_stop_str_in_output,
request=request,
)
elif stream:
decode_type = request_output["outputs"].get("decode_type", 0)
@@ -156,12 +159,14 @@ class ChatResponseProcessor:
response_dict=request_output,
stream=stream,
include_stop_str_in_output=include_stop_str_in_output,
request=request,
)
else:
self.data_processor.process_response_dict(
response_dict=request_output,
stream=stream,
include_stop_str_in_output=include_stop_str_in_output,
request=request,
)
text = {"type": "text", "text": request_output["outputs"]["text"]}
request_output["outputs"]["multipart"] = [text]
@@ -183,12 +188,14 @@ class ChatResponseProcessor:
response_dict=part["request_output"],
stream=False,
include_stop_str_in_output=include_stop_str_in_output,
request=request,
)
else:
self.data_processor.process_response_dict(
response_dict=request_output,
stream=stream,
include_stop_str_in_output=include_stop_str_in_output,
request=request,
)
text = {"type": "text", "text": part["request_output"]["outputs"]["text"]}
multipart.append(text)