Disable Gpt4all due to mac not supporting it currently (#233)

backend/danswer/configs/model_configs.py

@@ -35,9 +35,12 @@ BATCH_SIZE_ENCODE_CHUNKS = 8
 # Valid list:
 # - openai-completion
 # - openai-chat-completion
-# - gpt4all-completion
+# - gpt4all-completion -> Due to M1 Macs not having compatible gpt4all version, please install dependency yourself
-# - gpt4all-chat-completion
+# - gpt4all-chat-completion-> Due to M1 Macs not having compatible gpt4all version, please install dependency yourself
-INTERNAL_MODEL_VERSION = os.environ.get("INTERNAL_MODEL", "openai-chat-completion")
+# To use gpt4all, run: pip install --upgrade gpt4all==1.0.5
+INTERNAL_MODEL_VERSION = os.environ.get(
+    "INTERNAL_MODEL_VERSION", "openai-chat-completion"
+)
 # For GPT4ALL, use "ggml-model-gpt4all-falcon-q4_0.bin" for the below for a tested model
 GEN_AI_MODEL_VERSION = os.environ.get("GEN_AI_MODEL_VERSION", "gpt-3.5-turbo")
 GEN_AI_MAX_OUTPUT_TOKENS = 512

backend/danswer/direct_qa/__init__.py

@@ -3,14 +3,16 @@ from typing import Any
 from danswer.configs.app_configs import QA_TIMEOUT
 from danswer.configs.model_configs import INTERNAL_MODEL_VERSION
 from danswer.direct_qa.exceptions import UnknownModelError
-from danswer.direct_qa.gpt_4_all import GPT4AllChatCompletionQA
-from danswer.direct_qa.gpt_4_all import GPT4AllCompletionQA
 from danswer.direct_qa.interfaces import QAModel
 from danswer.direct_qa.open_ai import OpenAIChatCompletionQA
 from danswer.direct_qa.open_ai import OpenAICompletionQA
 from openai.error import AuthenticationError
 from openai.error import Timeout
 
+# Imports commented out temporarily due to incompatibility of gpt4all with M1 Mac hardware currently
+# from danswer.direct_qa.gpt_4_all import GPT4AllChatCompletionQA
+# from danswer.direct_qa.gpt_4_all import GPT4AllCompletionQA
+
 
 def check_model_api_key_is_valid(model_api_key: str) -> bool:
     if not model_api_key:
@@ -41,9 +43,10 @@ def get_default_backend_qa_model(
         return OpenAICompletionQA(timeout=timeout, api_key=api_key, **kwargs)
     elif internal_model == "openai-chat-completion":
         return OpenAIChatCompletionQA(timeout=timeout, api_key=api_key, **kwargs)
-    elif internal_model == "gpt4all-completion":
+    # Note GPT4All is not supported for M1 Mac machines currently, removing until support is added
-        return GPT4AllCompletionQA(**kwargs)
+    # elif internal_model == "gpt4all-completion":
-    elif internal_model == "gpt4all-chat-completion":
+    #    return GPT4AllCompletionQA(**kwargs)
-        return GPT4AllChatCompletionQA(**kwargs)
+    # elif internal_model == "gpt4all-chat-completion":
+    #    return GPT4AllChatCompletionQA(**kwargs)
     else:
         raise UnknownModelError(internal_model)

backend/requirements/default.txt

@@ -10,7 +10,9 @@ filelock==3.12.0
 google-api-python-client==2.86.0
 google-auth-httplib2==0.1.0
 google-auth-oauthlib==1.0.0
-gpt4all==1.0.5
+# GPT4All library does not support M1 Mac architecture
+# will reintroduce this when library version catches up
+# gpt4all==1.0.5
 httpcore==0.16.3
 httpx==0.23.3
 httpx-oauth==0.11.2

deployment/docker_compose/docker-compose.dev.yml

@@ -45,6 +45,8 @@ services:
       - vector_db
     restart: always
     environment:
+      - INTERNAL_MODEL_VERSION=${INTERNAL_MODEL_VERSION:-openai-chat-completion}
+      - GEN_AI_MODEL_VERSION=${GEN_AI_MODEL_VERSION:-gpt-3.5-turbo}
       - POSTGRES_HOST=relational_db
       - QDRANT_HOST=vector_db
       - TYPESENSE_HOST=search_engine