estel/dify
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

fix: first agent latency (#2334)

Browse Source
This commit is contained in:
Yeuoly
2024-02-01 15:30:50 +08:00
committed by GitHub
parent 09acf215f0
commit 3b357f51a6
1 changed files with 6 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
api/core/features/assistant_fc_runner.py
View File

@@ -97,7 +97,6 @@ class AssistantFunctionCallApplicationRunner(BaseAssistantApplicationRunner):
tool_input='', tool_input='',
messages_ids=message_file_ids messages_ids=message_file_ids
) )
self.queue_manager.publish_agent_thought(agent_thought, PublishFrom.APPLICATION_MANAGER)
# recale llm max tokens # recale llm max tokens
self.recale_llm_max_tokens(self.model_config, prompt_messages) self.recale_llm_max_tokens(self.model_config, prompt_messages)
@@ -124,7 +123,11 @@ class AssistantFunctionCallApplicationRunner(BaseAssistantApplicationRunner):
current_llm_usage = None current_llm_usage = None
if self.stream_tool_call: if self.stream_tool_call:
is_first_chunk = True
for chunk in chunks: for chunk in chunks:
if is_first_chunk:
self.queue_manager.publish_agent_thought(agent_thought, PublishFrom.APPLICATION_MANAGER)
is_first_chunk = False
# check if there is any tool call # check if there is any tool call
if self.check_tool_calls(chunk): if self.check_tool_calls(chunk):
function_call_state = True function_call_state = True
@@ -183,6 +186,8 @@ class AssistantFunctionCallApplicationRunner(BaseAssistantApplicationRunner):
if not result.message.content: if not result.message.content:
result.message.content = '' result.message.content = ''
self.queue_manager.publish_agent_thought(agent_thought, PublishFrom.APPLICATION_MANAGER)
yield LLMResultChunk( yield LLMResultChunk(
model=model_instance.model, model=model_instance.model,
prompt_messages=result.prompt_messages, prompt_messages=result.prompt_messages,