Remove unused env variables (#2299)

backend/danswer/configs/model_configs.py

@@ -51,33 +51,9 @@ CROSS_ENCODER_RANGE_MIN = 0
 # Generative AI Model Configs
 #####
 
-# If changing GEN_AI_MODEL_PROVIDER or GEN_AI_MODEL_VERSION from the default,
+# NOTE: settings like `GEN_AI_MODEL_PROVIDER`, `GEN_AI_MODEL_VERSION`, etc. which
-# be sure to use one that is LiteLLM compatible:
+# used to be here are now solely configured via the UI and stored in Postgres.
-# https://litellm.vercel.app/docs/providers/azure#completion---using-env-variables
-# The provider is the prefix before / in the model argument
 
-# Additionally Danswer supports GPT4All and custom request library based models
-# Set GEN_AI_MODEL_PROVIDER to "custom" to use the custom requests approach
-# Set GEN_AI_MODEL_PROVIDER to "gpt4all" to use gpt4all models running locally
-GEN_AI_MODEL_PROVIDER = os.environ.get("GEN_AI_MODEL_PROVIDER") or "openai"
-# If using Azure, it's the engine name, for example: Danswer
-GEN_AI_MODEL_VERSION = os.environ.get("GEN_AI_MODEL_VERSION")
-
-# For secondary flows like extracting filters or deciding if a chunk is useful, we don't need
-# as powerful of a model as say GPT-4 so we can use an alternative that is faster and cheaper
-FAST_GEN_AI_MODEL_VERSION = os.environ.get("FAST_GEN_AI_MODEL_VERSION")
-
-# 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 None
-)
-
-# API Base, such as (for Azure): https://danswer.openai.azure.com/
-GEN_AI_API_ENDPOINT = os.environ.get("GEN_AI_API_ENDPOINT") or None
-# API Version, such as (for Azure): 2023-09-15-preview
-GEN_AI_API_VERSION = os.environ.get("GEN_AI_API_VERSION") or None
-# LiteLLM custom_llm_provider
-GEN_AI_LLM_PROVIDER_TYPE = os.environ.get("GEN_AI_LLM_PROVIDER_TYPE") or None
 # Override the auto-detection of LLM max context length
 GEN_AI_MAX_TOKENS = int(os.environ.get("GEN_AI_MAX_TOKENS") or 0) or None
 

backend/danswer/llm/chat_llm.py

@@ -25,9 +25,6 @@ from langchain_core.messages.tool import ToolMessage
 from danswer.configs.app_configs import LOG_ALL_MODEL_INTERACTIONS
 from danswer.configs.app_configs import LOG_DANSWER_MODEL_INTERACTIONS
 from danswer.configs.model_configs import DISABLE_LITELLM_STREAMING
-from danswer.configs.model_configs import GEN_AI_API_ENDPOINT
-from danswer.configs.model_configs import GEN_AI_API_VERSION
-from danswer.configs.model_configs import GEN_AI_LLM_PROVIDER_TYPE
 from danswer.configs.model_configs import GEN_AI_TEMPERATURE
 from danswer.llm.interfaces import LLM
 from danswer.llm.interfaces import LLMConfig
@@ -192,10 +189,10 @@ class DefaultMultiLLM(LLM):
         timeout: int,
         model_provider: str,
         model_name: str,
+        api_base: str | None = None,
+        api_version: str | None = None,
         max_output_tokens: int | None = None,
-        api_base: str | None = GEN_AI_API_ENDPOINT,
+        custom_llm_provider: str | None = None,
-        api_version: str | None = GEN_AI_API_VERSION,
-        custom_llm_provider: str | None = GEN_AI_LLM_PROVIDER_TYPE,
         temperature: float = GEN_AI_TEMPERATURE,
         custom_config: dict[str, str] | None = None,
         extra_headers: dict[str, str] | None = None,

backend/danswer/llm/custom_llm.py

@@ -7,7 +7,6 @@ from langchain_core.messages import AIMessage
 from langchain_core.messages import BaseMessage
 from requests import Timeout
 
-from danswer.configs.model_configs import GEN_AI_API_ENDPOINT
 from danswer.configs.model_configs import GEN_AI_NUM_RESERVED_OUTPUT_TOKENS
 from danswer.llm.interfaces import LLM
 from danswer.llm.interfaces import ToolChoiceOptions
@@ -37,7 +36,7 @@ class CustomModelServer(LLM):
         # Not used here but you probably want a model server that isn't completely open
         api_key: str | None,
         timeout: int,
-        endpoint: str | None = GEN_AI_API_ENDPOINT,
+        endpoint: str,
         max_output_tokens: int = GEN_AI_NUM_RESERVED_OUTPUT_TOKENS,
     ):
         if not endpoint:

backend/danswer/llm/llm_initialization.py

@@ -1,80 +0,0 @@
-from sqlalchemy.orm import Session
-
-from danswer.configs.app_configs import DISABLE_GENERATIVE_AI
-from danswer.configs.model_configs import FAST_GEN_AI_MODEL_VERSION
-from danswer.configs.model_configs import GEN_AI_API_ENDPOINT
-from danswer.configs.model_configs import GEN_AI_API_KEY
-from danswer.configs.model_configs import GEN_AI_API_VERSION
-from danswer.configs.model_configs import GEN_AI_MODEL_PROVIDER
-from danswer.configs.model_configs import GEN_AI_MODEL_VERSION
-from danswer.db.llm import fetch_existing_llm_providers
-from danswer.db.llm import update_default_provider
-from danswer.db.llm import upsert_llm_provider
-from danswer.llm.llm_provider_options import AZURE_PROVIDER_NAME
-from danswer.llm.llm_provider_options import BEDROCK_PROVIDER_NAME
-from danswer.llm.llm_provider_options import fetch_available_well_known_llms
-from danswer.server.manage.llm.models import LLMProviderUpsertRequest
-from danswer.utils.logger import setup_logger
-
-
-logger = setup_logger()
-
-
-def load_llm_providers(db_session: Session) -> None:
-    existing_providers = fetch_existing_llm_providers(db_session)
-    if existing_providers:
-        return
-
-    if not GEN_AI_API_KEY or DISABLE_GENERATIVE_AI:
-        return
-
-    well_known_provider_name_to_provider = {
-        provider.name: provider
-        for provider in fetch_available_well_known_llms()
-        if provider.name != BEDROCK_PROVIDER_NAME
-    }
-
-    if GEN_AI_MODEL_PROVIDER not in well_known_provider_name_to_provider:
-        logger.error(f"Cannot auto-transition LLM provider: {GEN_AI_MODEL_PROVIDER}")
-        return None
-
-    # Azure provider requires custom model names,
-    # OpenAI / anthropic can just use the defaults
-    model_names = (
-        [
-            name
-            for name in [
-                GEN_AI_MODEL_VERSION,
-                FAST_GEN_AI_MODEL_VERSION,
-            ]
-            if name
-        ]
-        if GEN_AI_MODEL_PROVIDER == AZURE_PROVIDER_NAME
-        else None
-    )
-
-    well_known_provider = well_known_provider_name_to_provider[GEN_AI_MODEL_PROVIDER]
-    llm_provider_request = LLMProviderUpsertRequest(
-        name=well_known_provider.display_name,
-        provider=GEN_AI_MODEL_PROVIDER,
-        api_key=GEN_AI_API_KEY,
-        api_base=GEN_AI_API_ENDPOINT,
-        api_version=GEN_AI_API_VERSION,
-        custom_config={},
-        default_model_name=(
-            GEN_AI_MODEL_VERSION
-            or well_known_provider.default_model
-            or well_known_provider.llm_names[0]
-        ),
-        fast_default_model_name=(
-            FAST_GEN_AI_MODEL_VERSION or well_known_provider.default_fast_model
-        ),
-        model_names=model_names,
-        is_public=True,
-        display_model_names=[],
-    )
-    llm_provider = upsert_llm_provider(db_session, llm_provider_request)
-    update_default_provider(db_session, llm_provider.id)
-    logger.notice(
-        f"Migrated LLM provider from env variables for provider '{GEN_AI_MODEL_PROVIDER}'"
-    )

backend/danswer/llm/utils.py

@@ -32,7 +32,6 @@ from litellm.exceptions import UnprocessableEntityError  # type: ignore
 from danswer.configs.constants import MessageType
 from danswer.configs.model_configs import GEN_AI_MAX_TOKENS
 from danswer.configs.model_configs import GEN_AI_MODEL_FALLBACK_MAX_TOKENS
-from danswer.configs.model_configs import GEN_AI_MODEL_PROVIDER
 from danswer.configs.model_configs import GEN_AI_NUM_RESERVED_OUTPUT_TOKENS
 from danswer.db.models import ChatMessage
 from danswer.file_store.models import ChatFileType
@@ -331,7 +330,7 @@ def test_llm(llm: LLM) -> str | None:
 def get_llm_max_tokens(
     model_map: dict,
     model_name: str,
-    model_provider: str = GEN_AI_MODEL_PROVIDER,
+    model_provider: str,
 ) -> int:
     """Best effort attempt to get the max tokens for the LLM"""
     if GEN_AI_MAX_TOKENS:
@@ -371,7 +370,7 @@ def get_llm_max_tokens(
 def get_llm_max_output_tokens(
     model_map: dict,
     model_name: str,
-    model_provider: str = GEN_AI_MODEL_PROVIDER,
+    model_provider: str,
 ) -> int:
     """Best effort attempt to get the max output tokens for the LLM"""
     try:

backend/danswer/main.py

@@ -62,7 +62,6 @@ from danswer.document_index.interfaces import DocumentIndex
 from danswer.dynamic_configs.factory import get_dynamic_config_store
 from danswer.dynamic_configs.interface import ConfigNotFoundError
 from danswer.indexing.models import IndexingSetting
-from danswer.llm.llm_initialization import load_llm_providers
 from danswer.natural_language_processing.search_nlp_models import EmbeddingModel
 from danswer.natural_language_processing.search_nlp_models import warm_up_bi_encoder
 from danswer.natural_language_processing.search_nlp_models import warm_up_cross_encoder
@@ -182,9 +181,6 @@ def setup_postgres(db_session: Session) -> None:
     logger.notice("Verifying default standard answer category exists.")
     create_initial_default_standard_answer_category(db_session)
 
-    logger.notice("Loading LLM providers from env variables")
-    load_llm_providers(db_session)
-
     logger.notice("Loading default Prompts and Personas")
     delete_old_default_personas(db_session)
     load_chat_yamls()

deployment/docker_compose/docker-compose.dev.yml

@@ -35,13 +35,6 @@ services:
       - OPENID_CONFIG_URL=${OPENID_CONFIG_URL:-}
       - TRACK_EXTERNAL_IDP_EXPIRY=${TRACK_EXTERNAL_IDP_EXPIRY:-}
       # Gen AI Settings
-      - GEN_AI_MODEL_PROVIDER=${GEN_AI_MODEL_PROVIDER:-}
-      - GEN_AI_MODEL_VERSION=${GEN_AI_MODEL_VERSION:-}
-      - FAST_GEN_AI_MODEL_VERSION=${FAST_GEN_AI_MODEL_VERSION:-}
-      - GEN_AI_API_KEY=${GEN_AI_API_KEY:-}
-      - GEN_AI_API_ENDPOINT=${GEN_AI_API_ENDPOINT:-}
-      - GEN_AI_API_VERSION=${GEN_AI_API_VERSION:-}
-      - GEN_AI_LLM_PROVIDER_TYPE=${GEN_AI_LLM_PROVIDER_TYPE:-}
       - GEN_AI_MAX_TOKENS=${GEN_AI_MAX_TOKENS:-}
       - QA_TIMEOUT=${QA_TIMEOUT:-}
       - MAX_CHUNKS_FED_TO_CHAT=${MAX_CHUNKS_FED_TO_CHAT:-}
@@ -120,13 +113,6 @@ services:
     environment:
       - ENCRYPTION_KEY_SECRET=${ENCRYPTION_KEY_SECRET:-}
       # Gen AI Settings (Needed by DanswerBot)
-      - GEN_AI_MODEL_PROVIDER=${GEN_AI_MODEL_PROVIDER:-}
-      - GEN_AI_MODEL_VERSION=${GEN_AI_MODEL_VERSION:-}
-      - FAST_GEN_AI_MODEL_VERSION=${FAST_GEN_AI_MODEL_VERSION:-}
-      - GEN_AI_API_KEY=${GEN_AI_API_KEY:-}
-      - GEN_AI_API_ENDPOINT=${GEN_AI_API_ENDPOINT:-}
-      - GEN_AI_API_VERSION=${GEN_AI_API_VERSION:-}
-      - GEN_AI_LLM_PROVIDER_TYPE=${GEN_AI_LLM_PROVIDER_TYPE:-}
       - GEN_AI_MAX_TOKENS=${GEN_AI_MAX_TOKENS:-}
       - QA_TIMEOUT=${QA_TIMEOUT:-}
       - MAX_CHUNKS_FED_TO_CHAT=${MAX_CHUNKS_FED_TO_CHAT:-}

deployment/docker_compose/docker-compose.gpu-dev.yml

@@ -32,13 +32,6 @@ services:
       - EMAIL_FROM=${EMAIL_FROM:-}
       - TRACK_EXTERNAL_IDP_EXPIRY=${TRACK_EXTERNAL_IDP_EXPIRY:-}
       # Gen AI Settings
-      - GEN_AI_MODEL_PROVIDER=${GEN_AI_MODEL_PROVIDER:-}
-      - GEN_AI_MODEL_VERSION=${GEN_AI_MODEL_VERSION:-}
-      - FAST_GEN_AI_MODEL_VERSION=${FAST_GEN_AI_MODEL_VERSION:-}
-      - GEN_AI_API_KEY=${GEN_AI_API_KEY:-}
-      - GEN_AI_API_ENDPOINT=${GEN_AI_API_ENDPOINT:-}
-      - GEN_AI_API_VERSION=${GEN_AI_API_VERSION:-}
-      - GEN_AI_LLM_PROVIDER_TYPE=${GEN_AI_LLM_PROVIDER_TYPE:-}
       - GEN_AI_MAX_TOKENS=${GEN_AI_MAX_TOKENS:-}
       - QA_TIMEOUT=${QA_TIMEOUT:-}
       - MAX_CHUNKS_FED_TO_CHAT=${MAX_CHUNKS_FED_TO_CHAT:-}
@@ -112,13 +105,6 @@ services:
     environment:
       - ENCRYPTION_KEY_SECRET=${ENCRYPTION_KEY_SECRET:-}
       # Gen AI Settings (Needed by DanswerBot)
-      - GEN_AI_MODEL_PROVIDER=${GEN_AI_MODEL_PROVIDER:-}
-      - GEN_AI_MODEL_VERSION=${GEN_AI_MODEL_VERSION:-}
-      - FAST_GEN_AI_MODEL_VERSION=${FAST_GEN_AI_MODEL_VERSION:-}
-      - GEN_AI_API_KEY=${GEN_AI_API_KEY:-}
-      - GEN_AI_API_ENDPOINT=${GEN_AI_API_ENDPOINT:-}
-      - GEN_AI_API_VERSION=${GEN_AI_API_VERSION:-}
-      - GEN_AI_LLM_PROVIDER_TYPE=${GEN_AI_LLM_PROVIDER_TYPE:-}
       - GEN_AI_MAX_TOKENS=${GEN_AI_MAX_TOKENS:-}
       - QA_TIMEOUT=${QA_TIMEOUT:-}
       - MAX_CHUNKS_FED_TO_CHAT=${MAX_CHUNKS_FED_TO_CHAT:-}

deployment/docker_compose/env.prod.template

@@ -7,16 +7,7 @@
 WEB_DOMAIN=http://localhost:3000
 
 
-# Generative AI settings, uncomment as needed, will work with defaults
+# NOTE: Generative AI configurations are done via the UI now
-GEN_AI_MODEL_PROVIDER=openai
-GEN_AI_MODEL_VERSION=gpt-4
-# Provide this as a global default/backup, this can also be set via the UI
-#GEN_AI_API_KEY=
-# Set to use Azure OpenAI or other services, such as https://danswer.openai.azure.com/
-#GEN_AI_API_ENDPOINT=
-# Set up to use a specific API version, such as 2023-09-15-preview (example taken from Azure)
-#GEN_AI_API_VERSION=
-
 
 # If you want to setup a slack bot to answer questions automatically in Slack 
 # channels it is added to, you must specify the two below.

deployment/helm/values.yaml

@@ -392,13 +392,6 @@ configMap:
   # SMTP_PASS: ""  # 'your-gmail-password'
   EMAIL_FROM: ""  # 'your-email@company.com' SMTP_USER missing used instead
   # Gen AI Settings
-  GEN_AI_MODEL_PROVIDER: ""
-  GEN_AI_MODEL_VERSION: ""
-  FAST_GEN_AI_MODEL_VERSION: ""
-  # GEN_AI_API_KEY: ""
-  GEN_AI_API_ENDPOINT: ""
-  GEN_AI_API_VERSION: ""
-  GEN_AI_LLM_PROVIDER_TYPE: ""
   GEN_AI_MAX_TOKENS: ""
   QA_TIMEOUT: "60"
   MAX_CHUNKS_FED_TO_CHAT: ""

deployment/kubernetes/env-configmap.yaml

@@ -14,13 +14,6 @@ data:
   SMTP_PASS: ""  # 'your-gmail-password'
   EMAIL_FROM: ""  # 'your-email@company.com' SMTP_USER missing used instead
   # Gen AI Settings
-  GEN_AI_MODEL_PROVIDER: ""
-  GEN_AI_MODEL_VERSION: ""
-  FAST_GEN_AI_MODEL_VERSION: ""
-  GEN_AI_API_KEY: ""
-  GEN_AI_API_ENDPOINT: ""
-  GEN_AI_API_VERSION: ""
-  GEN_AI_LLM_PROVIDER_TYPE: ""
   GEN_AI_MAX_TOKENS: ""
   QA_TIMEOUT: "60"
   MAX_CHUNKS_FED_TO_CHAT: ""