Adjust the way LLM class is instantiated + fix issue where .env file …

…GEN_AI_API_KEY wasn't being used (#630)

backend/danswer/configs/model_configs.py

@@ -82,10 +82,7 @@
 )
 
 # If the Generative AI model requires an API key for access, otherwise can leave blank
-GEN_AI_API_KEY = (
-    os.environ.get("GEN_AI_API_KEY", os.environ.get("OPENAI_API_KEY"))
-    or "dummy_llm_key"
-)
+GEN_AI_API_KEY = os.environ.get("GEN_AI_API_KEY", os.environ.get("OPENAI_API_KEY"))
 
 # If using GPT4All, HuggingFace Inference API, or OpenAI - specify the model version
 GEN_AI_MODEL_VERSION = os.environ.get(

backend/danswer/direct_qa/open_ai.py

@@ -48,11 +48,12 @@
 
 
 def _ensure_openai_api_key(api_key: str | None) -> str:
-    try:
-        return api_key or get_gen_ai_api_key()
-    except ConfigNotFoundError:
+    final_api_key = api_key or get_gen_ai_api_key()
+    if final_api_key is None:
         raise OpenAIKeyMissing()
 
+    return final_api_key
+
 
 def _build_openai_settings(**kwargs: Any) -> dict[str, Any]:
     """

backend/danswer/direct_qa/qa_utils.py

@@ -22,6 +22,7 @@
 from danswer.direct_qa.qa_prompts import QUOTE_PAT
 from danswer.direct_qa.qa_prompts import UNCERTAINTY_PAT
 from danswer.dynamic_configs import get_dynamic_config_store
+from danswer.dynamic_configs.interface import ConfigNotFoundError
 from danswer.llm.utils import check_number_of_tokens
 from danswer.utils.logger import setup_logger
 from danswer.utils.text_processing import clean_model_quote
@@ -31,11 +32,15 @@
 logger = setup_logger()
 
 
-def get_gen_ai_api_key() -> str:
-    return (
-        cast(str, get_dynamic_config_store().load(GEN_AI_API_KEY_STORAGE_KEY))
-        or GEN_AI_API_KEY
-    )
+def get_gen_ai_api_key() -> str | None:
+    # first check if the key has been provided by the UI
+    try:
+        return cast(str, get_dynamic_config_store().load(GEN_AI_API_KEY_STORAGE_KEY))
+    except ConfigNotFoundError:
+        pass
+
+    # if not provided by the UI, fallback to the env variable
+    return GEN_AI_API_KEY
 
 
 def extract_answer_quotes_freeform(

backend/danswer/llm/azure.py

@@ -5,7 +5,6 @@
 from danswer.configs.model_configs import API_BASE_OPENAI
 from danswer.configs.model_configs import API_VERSION_OPENAI
 from danswer.configs.model_configs import AZURE_DEPLOYMENT_ID
-from danswer.configs.model_configs import GEN_AI_API_KEY
 from danswer.llm.llm import LangChainChatLLM
 from danswer.llm.utils import should_be_verbose
 
@@ -23,11 +22,6 @@ def __init__(
         *args: list[Any],
         **kwargs: dict[str, Any]
     ):
-        # set a dummy API key if not specified so that LangChain doesn't throw an
-        # exception when trying to initialize the LLM which would prevent the API
-        # server from starting up
-        if not api_key:
-            api_key = GEN_AI_API_KEY
         self._llm = AzureChatOpenAI(
             model=model_version,
             openai_api_type="azure",

backend/danswer/llm/build.py

@@ -1,5 +1,3 @@
-from typing import Any
-
 from danswer.configs.app_configs import QA_TIMEOUT
 from danswer.configs.constants import DanswerGenAIModel
 from danswer.configs.constants import ModelHostType
@@ -11,48 +9,40 @@
 from danswer.configs.model_configs import GEN_AI_TEMPERATURE
 from danswer.configs.model_configs import INTERNAL_MODEL_VERSION
 from danswer.direct_qa.qa_utils import get_gen_ai_api_key
-from danswer.dynamic_configs.interface import ConfigNotFoundError
 from danswer.llm.azure import AzureGPT
 from danswer.llm.google_colab_demo import GoogleColabDemo
 from danswer.llm.llm import LLM
 from danswer.llm.openai import OpenAIGPT
 
 
-def get_llm_from_model(model: str, **kwargs: Any) -> LLM:
-    if model == DanswerGenAIModel.OPENAI_CHAT.value:
-        if API_TYPE_OPENAI == "azure":
-            return AzureGPT(**kwargs)
-        return OpenAIGPT(**kwargs)
-    if (
-        model == DanswerGenAIModel.REQUEST.value
-        and kwargs.get("model_host_type") == ModelHostType.COLAB_DEMO
-    ):
-        return GoogleColabDemo(**kwargs)
-
-    raise ValueError(f"Unknown LLM model: {model}")
-
-
 def get_default_llm(
-    api_key: str | None = None, timeout: int = QA_TIMEOUT, **kwargs: Any
+    api_key: str | None = None,
+    timeout: int = QA_TIMEOUT,
 ) -> LLM:
     """NOTE: api_key/timeout must be a special args since we may want to check
     if an API key is valid for the default model setup OR we may want to use the
     default model with a different timeout specified."""
     if api_key is None:
-        try:
-            api_key = get_gen_ai_api_key()
-        except ConfigNotFoundError:
-            # if no API key is found, assume this model doesn't need one
-            pass
+        api_key = get_gen_ai_api_key()
+
+    model_args = {
+        # provide a dummy key since LangChain will throw an exception if not
+        # given, which would prevent server startup
+        "api_key": api_key or "dummy_api_key",
+        "timeout": timeout,
+        "model_version": GEN_AI_MODEL_VERSION,
+        "endpoint": GEN_AI_ENDPOINT,
+        "max_output_tokens": GEN_AI_MAX_OUTPUT_TOKENS,
+        "temperature": GEN_AI_TEMPERATURE,
+    }
+    if INTERNAL_MODEL_VERSION == DanswerGenAIModel.OPENAI_CHAT.value:
+        if API_TYPE_OPENAI == "azure":
+            return AzureGPT(**model_args)  # type: ignore
+        return OpenAIGPT(**model_args)  # type: ignore
+    if (
+        INTERNAL_MODEL_VERSION == DanswerGenAIModel.REQUEST.value
+        and GEN_AI_HOST_TYPE == ModelHostType.COLAB_DEMO
+    ):
+        return GoogleColabDemo(**model_args)  # type: ignore
 
-    return get_llm_from_model(
-        model=INTERNAL_MODEL_VERSION,
-        api_key=api_key,
-        timeout=timeout,
-        model_version=GEN_AI_MODEL_VERSION,
-        endpoint=GEN_AI_ENDPOINT,
-        model_host_type=GEN_AI_HOST_TYPE,
-        max_output_tokens=GEN_AI_MAX_OUTPUT_TOKENS,
-        temperature=GEN_AI_TEMPERATURE,
-        **kwargs,
-    )
+    raise ValueError(f"Unknown LLM model: {INTERNAL_MODEL_VERSION}")

backend/danswer/llm/openai.py

@@ -3,7 +3,6 @@
 
 from langchain.chat_models.openai import ChatOpenAI
 
-from danswer.configs.model_configs import GEN_AI_API_KEY
 from danswer.configs.model_configs import GEN_AI_TEMPERATURE
 from danswer.llm.llm import LangChainChatLLM
 from danswer.llm.utils import should_be_verbose
@@ -20,12 +19,6 @@ def __init__(
         *args: list[Any],
         **kwargs: dict[str, Any]
     ):
-        # set a dummy API key if not specified so that LangChain doesn't throw an
-        # exception when trying to initialize the LLM which would prevent the API
-        # server from starting up
-        if not api_key:
-            api_key = GEN_AI_API_KEY
-
         self._llm = ChatOpenAI(
             model=model_version,
             openai_api_key=api_key,

backend/danswer/server/manage.py

@@ -502,14 +502,9 @@ def validate_existing_genai_api_key(
         # First time checking the key, nothing unusual
         pass
 
-    try:
-        genai_api_key = get_gen_ai_api_key()
-    except ConfigNotFoundError:
+    genai_api_key = get_gen_ai_api_key()
+    if genai_api_key is None:
         raise HTTPException(status_code=404, detail="Key not found")
-    except ValueError as e:
-        raise HTTPException(status_code=400, detail=str(e))
-
-    get_dynamic_config_store().store(check_key_time, curr_time.timestamp())
 
     try:
         is_valid = check_model_api_key_is_valid(genai_api_key)
@@ -520,6 +515,9 @@ def validate_existing_genai_api_key(
     if not is_valid:
         raise HTTPException(status_code=400, detail="Invalid API key provided")
 
+    # mark check as successful
+    get_dynamic_config_store().store(check_key_time, curr_time.timestamp())
+
 
 @router.get("/admin/genai-api-key", response_model=ApiKey)
 def get_gen_ai_api_key_from_dynamic_config_store(