sij/khoj
1
0
Fork 0
mirror of https://github.com/khoj-ai/khoj.git synced 2024-11-23 15:38:55 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 5

Default to gpt-4o-mini instead of gpt-3.5-turbo in tests, func args

Browse source 
GPT-4o-mini is cheaper, smarter and can hold more context than
GPT-3.5-turbo. In production, we also default to gpt-4o-mini, so makes
sense to upgrade defaults and tests to work with it
This commit is contained in:
Debanjum Singh Solanky 2024-08-22 19:04:49 -07:00
parent 8a4c20d59a
commit 9986c183ea
5 changed files with 7 additions and 6 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
src/khoj/processor/conversation/openai/gpt.py
View file

@ -112,7 +112,7 @@ def converse(
user_query, user_query,
online_results: Optional[Dict[str, Dict]] = None, online_results: Optional[Dict[str, Dict]] = None,
conversation_log={}, conversation_log={},
model: str = "gpt-3.5-turbo", model: str = "gpt-4o-mini",
api_key: Optional[str] = None, api_key: Optional[str] = None,
api_base_url: Optional[str] = None, api_base_url: Optional[str] = None,
temperature: float = 0.2, temperature: float = 0.2,

3
src/khoj/processor/conversation/utils.py
View file

@ -23,6 +23,7 @@ model_to_prompt_size = {
"gpt-3.5-turbo-0125": 12000, "gpt-3.5-turbo-0125": 12000,
"gpt-4-0125-preview": 20000, "gpt-4-0125-preview": 20000,
"gpt-4-turbo-preview": 20000, "gpt-4-turbo-preview": 20000,
"gpt-4o-mini": 20000,
"TheBloke/Mistral-7B-Instruct-v0.2-GGUF": 3500, "TheBloke/Mistral-7B-Instruct-v0.2-GGUF": 3500,
"NousResearch/Hermes-2-Pro-Mistral-7B-GGUF": 3500, "NousResearch/Hermes-2-Pro-Mistral-7B-GGUF": 3500,
"bartowski/Meta-Llama-3.1-8B-Instruct-GGUF": 20000, "bartowski/Meta-Llama-3.1-8B-Instruct-GGUF": 20000,
@ -136,7 +137,7 @@ def generate_chatml_messages_with_context(
user_message, user_message,
system_message=None, system_message=None,
conversation_log={}, conversation_log={},
model_name="gpt-3.5-turbo", model_name="gpt-4o-mini",
loaded_model: Optional[Llama] = None, loaded_model: Optional[Llama] = None,
max_prompt_size=None, max_prompt_size=None,
tokenizer_name=None, tokenizer_name=None,

2
src/khoj/utils/rawconfig.py
View file

@ -89,7 +89,7 @@ class SearchConfig(ConfigBase):
class OpenAIProcessorConfig(ConfigBase): class OpenAIProcessorConfig(ConfigBase):
api_key: str api_key: str
chat_model: Optional[str] = "gpt-3.5-turbo" chat_model: Optional[str] = "gpt-4o-mini"
class OfflineChatProcessorConfig(ConfigBase): class OfflineChatProcessorConfig(ConfigBase):

4
tests/conftest.py
View file

@ -192,7 +192,7 @@ def offline_agent():
@pytest.mark.django_db @pytest.mark.django_db
@pytest.fixture @pytest.fixture
def openai_agent(): def openai_agent():
chat_model = ChatModelOptionsFactory(chat_model="gpt-3.5-turbo", model_type="openai") chat_model = ChatModelOptionsFactory(chat_model="gpt-4o-mini", model_type="openai")
return Agent.objects.create( return Agent.objects.create(
name="Accountant", name="Accountant",
chat_model=chat_model, chat_model=chat_model,
@ -301,7 +301,7 @@ def chat_client_builder(search_config, user, index_content=True, require_auth=Fa
# Initialize Processor from Config # Initialize Processor from Config
if os.getenv("OPENAI_API_KEY"): if os.getenv("OPENAI_API_KEY"):
chat_model = ChatModelOptionsFactory(chat_model="gpt-3.5-turbo", model_type="openai") chat_model = ChatModelOptionsFactory(chat_model="gpt-4o-mini", model_type="openai")
chat_model.openai_config = OpenAIProcessorConversationConfigFactory() chat_model.openai_config = OpenAIProcessorConversationConfigFactory()
UserConversationProcessorConfigFactory(user=user, setting=chat_model) UserConversationProcessorConfigFactory(user=user, setting=chat_model)

2
tests/test_conversation_utils.py
View file

@ -6,7 +6,7 @@ from khoj.processor.conversation import utils
class TestTruncateMessage: class TestTruncateMessage:
max_prompt_size = 10 max_prompt_size = 10
model_name = "gpt-3.5-turbo" model_name = "gpt-4o-mini"
encoder = tiktoken.encoding_for_model(model_name) encoder = tiktoken.encoding_for_model(model_name)
def test_truncate_message_all_small(self): def test_truncate_message_all_small(self):