From 0b78480977052d88e8945ce50a714339f9f93c43 Mon Sep 17 00:00:00 2001
From: Alex <a@tushynski.me>
Date: Thu, 25 May 2023 15:14:47 +0100
Subject: [PATCH 1/2] init

---
 application/app.py           | 17 +++++++++++++++++
 application/requirements.txt |  6 ++++--
 2 files changed, 21 insertions(+), 2 deletions(-)

diff --git a/application/app.py b/application/app.py
index 7c12721f..a88a1bec 100644
--- a/application/app.py
+++ b/application/app.py
@@ -27,6 +27,7 @@ from langchain.prompts.chat import (
 )
 from pymongo import MongoClient
 from werkzeug.utils import secure_filename
+from langchain.llms import GPT4All
 
 from core.settings import settings
 from error import bad_request
@@ -195,6 +196,9 @@ def api_answer():
             llm = HuggingFaceHub(repo_id="bigscience/bloom", huggingfacehub_api_token=api_key)
         elif settings.LLM_NAME == "cohere":
             llm = Cohere(model="command-xlarge-nightly", cohere_api_key=api_key)
+        elif settings.LLM_NAME == "gpt4all":
+            llm = GPT4All(model="/Users/alextu/Library/Application Support/nomic.ai/GPT4All/",
+                          backend='gpt4all-j-v1.3-groovy')
         else:
             raise ValueError("unknown LLM model")
 
@@ -210,6 +214,19 @@ def api_answer():
             # result = chain({"question": question, "chat_history": chat_history})
             # generate async with async generate method
             result = run_async_chain(chain, question, chat_history)
+        elif settings.LLM_NAME == "gpt4all":
+            question_generator = LLMChain(llm=llm, prompt=CONDENSE_QUESTION_PROMPT)
+            doc_chain = load_qa_chain(llm, chain_type="map_reduce", combine_prompt=p_chat_combine)
+            chain = ConversationalRetrievalChain(
+                retriever=docsearch.as_retriever(k=2),
+                question_generator=question_generator,
+                combine_docs_chain=doc_chain,
+            )
+            chat_history = []
+            # result = chain({"question": question, "chat_history": chat_history})
+            # generate async with async generate method
+            result = run_async_chain(chain, question, chat_history)
+
         else:
             qa_chain = load_qa_chain(llm=llm, chain_type="map_reduce",
                                      combine_prompt=chat_combine_template, question_prompt=q_prompt)
diff --git a/application/requirements.txt b/application/requirements.txt
index 7f737e32..af976297 100644
--- a/application/requirements.txt
+++ b/application/requirements.txt
@@ -26,7 +26,8 @@ ecdsa==0.18.0
 entrypoints==0.4
 faiss-cpu==1.7.3
 filelock==3.9.0
-Flask==2.3.2
+Flask==2.2.3
+Flask-Cors==3.0.10
 frozenlist==1.3.3
 geojson==2.5.0
 greenlet==2.0.2
@@ -39,7 +40,8 @@ Jinja2==3.1.2
 jmespath==1.0.1
 joblib==1.2.0
 kombu==5.2.4
-langchain==0.0.126
+langchain==0.0.179
+loguru==0.6.0
 lxml==4.9.2
 MarkupSafe==2.1.2
 marshmallow==3.19.0

From aaa1249a41215fa5a5400832bad9143df4175ed4 Mon Sep 17 00:00:00 2001
From: Alex <a@tushynski.me>
Date: Thu, 25 May 2023 19:33:37 +0100
Subject: [PATCH 2/2] model fix + env var

---
 application/app.py           | 3 +--
 application/core/settings.py | 1 +
 application/requirements.txt | 1 +
 3 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/application/app.py b/application/app.py
index a88a1bec..14916d06 100644
--- a/application/app.py
+++ b/application/app.py
@@ -197,8 +197,7 @@ def api_answer():
         elif settings.LLM_NAME == "cohere":
             llm = Cohere(model="command-xlarge-nightly", cohere_api_key=api_key)
         elif settings.LLM_NAME == "gpt4all":
-            llm = GPT4All(model="/Users/alextu/Library/Application Support/nomic.ai/GPT4All/",
-                          backend='gpt4all-j-v1.3-groovy')
+            llm = GPT4All(model=settings.MODEL_PATH)
         else:
             raise ValueError("unknown LLM model")
 
diff --git a/application/core/settings.py b/application/core/settings.py
index fa654ed5..3c0672da 100644
--- a/application/core/settings.py
+++ b/application/core/settings.py
@@ -9,6 +9,7 @@ class Settings(BaseSettings):
     CELERY_BROKER_URL: str = "redis://localhost:6379/0"
     CELERY_RESULT_BACKEND: str = "redis://localhost:6379/1"
     MONGO_URI: str = "mongodb://localhost:27017/docsgpt"
+    MODEL_PATH: str = "./models/gpt4all-model.bin"
 
     API_URL: str = "http://localhost:5001"  # backend url for celery worker
 
diff --git a/application/requirements.txt b/application/requirements.txt
index af976297..4d9c1b0d 100644
--- a/application/requirements.txt
+++ b/application/requirements.txt
@@ -31,6 +31,7 @@ Flask-Cors==3.0.10
 frozenlist==1.3.3
 geojson==2.5.0
 greenlet==2.0.2
+gpt4all==0.1.7
 hub==3.0.1
 huggingface-hub==0.12.1
 humbug==0.2.8