klbr/khoj
1
0
Fork 0
mirror of https://github.com/khoaliber/khoj.git synced 2026-03-06 05:39:12 +00:00
Code Issues Packages Projects Releases Wiki Activity

Stop OpenAI chat from emitting reference notes directly in chat body

Browse Source 
The Chat models sometime output reference notes directly in the chat
body in unformatted form, specifically as Notes:\n['. Prevent that.
Reference notes are shown in clean, formatted form anyway
This commit is contained in:
Debanjum Singh Solanky
2023-11-10 22:27:24 -08:00
parent 8585976f37
commit cba371678d
2 changed files with 5 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
src/khoj/processor/conversation/openai/gpt.py
View File

@@ -134,4 +134,5 @@ def converse(
temperature=temperature, temperature=temperature,
openai_api_key=api_key, openai_api_key=api_key,
completion_func=completion_func, completion_func=completion_func,
model_kwargs={"stop": ["Notes:\n["]},
) )

7
src/khoj/processor/conversation/openai/utils.py
View File

@@ -69,15 +69,15 @@ def completion_with_backoff(**kwargs):
reraise=True, reraise=True,
) )
def chat_completion_with_backoff( def chat_completion_with_backoff(
messages, compiled_references, model_name, temperature, openai_api_key=None, completion_func=None messages, compiled_references, model_name, temperature, openai_api_key=None, completion_func=None, model_kwargs=None
): ):
g = ThreadedGenerator(compiled_references, completion_func=completion_func) g = ThreadedGenerator(compiled_references, completion_func=completion_func)
t = Thread(target=llm_thread, args=(g, messages, model_name, temperature, openai_api_key)) t = Thread(target=llm_thread, args=(g, messages, model_name, temperature, openai_api_key, model_kwargs))
t.start() t.start()
return g return g
def llm_thread(g, messages, model_name, temperature, openai_api_key=None): def llm_thread(g, messages, model_name, temperature, openai_api_key=None, model_kwargs=None):
callback_handler = StreamingChatCallbackHandler(g) callback_handler = StreamingChatCallbackHandler(g)
chat = ChatOpenAI( chat = ChatOpenAI(
streaming=True, streaming=True,
@@ -86,6 +86,7 @@ def llm_thread(g, messages, model_name, temperature, openai_api_key=None):
model_name=model_name, # type: ignore model_name=model_name, # type: ignore
temperature=temperature, temperature=temperature,
openai_api_key=openai_api_key or os.getenv("OPENAI_API_KEY"), openai_api_key=openai_api_key or os.getenv("OPENAI_API_KEY"),
model_kwargs=model_kwargs,
request_timeout=20, request_timeout=20,
max_retries=1, max_retries=1,
client=None, client=None,