From 5bee915f6d5ad77b881e12c891eda1777913d9da Mon Sep 17 00:00:00 2001
From: hashcode-ankit <2698ankitsharma@gmail.com>
Date: Sun, 3 Sep 2023 20:18:54 +0530
Subject: [PATCH 1/6] weaviate vector database feature added for openai only

---
 examples/openai-bot/main.py | 10 ++++++-
 textbase/__init__.py        |  4 ++-
 textbase/models.py          | 59 ++++++++++++++++++++++++++-----------
 textbase/vector_database.py | 52 ++++++++++++++++++++++++++++++++
 4 files changed, 105 insertions(+), 20 deletions(-)
 create mode 100644 textbase/vector_database.py

diff --git a/examples/openai-bot/main.py b/examples/openai-bot/main.py
index ebc25837..bb92f965 100644
--- a/examples/openai-bot/main.py
+++ b/examples/openai-bot/main.py
@@ -3,7 +3,15 @@
 from typing import List
 
 # Load your OpenAI API key
-OpenAI.api_key = ""
+OpenAI.api_key = "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"
+
+# optional if you want to use vector database 
+# currently textbase support weaviate
+OpenAI.vector_db_host = "http://20.193.138.42:8080"
+
+OpenAI.vector_db_data_class = "Documents"
+OpenAI.max_vector_database_objects = 1
+
 
 # Prompt for GPT-3.5 Turbo
 SYSTEM_PROMPT = """You are chatting with an AI. There are no specific prefixes for responses, so you can ask or talk about anything you like.
diff --git a/textbase/__init__.py b/textbase/__init__.py
index a19f84ae..5e923b6a 100644
--- a/textbase/__init__.py
+++ b/textbase/__init__.py
@@ -1,2 +1,4 @@
 from .bot import bot
-from .message import Message
\ No newline at end of file
+from .message import Message
+from .message import Content
+from .vector_database import WeaviateClass
\ No newline at end of file
diff --git a/textbase/models.py b/textbase/models.py
index 814ed533..0124eb41 100644
--- a/textbase/models.py
+++ b/textbase/models.py
@@ -4,8 +4,8 @@
 import time
 import typing
 import traceback
-
-from textbase import Message
+from textbase import Message, WeaviateClass
+# from vector_database import Weaviate
 
 # Return list of values of content.
 def get_contents(message: Message, data_type: str):
@@ -28,7 +28,10 @@ def extract_content_values(message: Message):
 
 class OpenAI:
     api_key = None
-
+    vector_db_host = None
+    vector_db_auth_key = None
+    vector_db_data_class = None
+    max_vector_database_objects = None
     @classmethod
     def generate(
         cls,
@@ -40,7 +43,6 @@ def generate(
     ):
         assert cls.api_key is not None, "OpenAI API key is not set."
         openai.api_key = cls.api_key
-
         filtered_messages = []
 
         for message in message_history:
@@ -48,20 +50,41 @@ def generate(
             contents = get_contents(message, "STRING")
             if contents:
                 filtered_messages.extend(contents)
-
-        response = openai.ChatCompletion.create(
-            model=model,
-            messages=[
-                {
-                    "role": "system",
-                    "content": system_prompt
-                },
-                *map(dict, filtered_messages),
-            ],
-            temperature=temperature,
-            max_tokens=max_tokens,
-        )
-
+        weaviate_response = None
+        # if vector database host provided get response from weaviate
+        if cls.vector_db_host :
+            WeaviateClass.host = cls.vector_db_host
+            WeaviateClass.auth_key = cls.vector_db_auth_key
+            WeaviateClass.api_key = cls.api_key
+            WeaviateClass.vector_db_data_class = cls.vector_db_data_class
+            weaviate_response = WeaviateClass.search_in_weaviate(message_history[-1],"X-OpenAI-Api-Key",cls.max_vector_database_objects)
+        
+            response = openai.ChatCompletion.create(
+                model=model,
+                messages=[
+                    {
+                        "role": "system",
+                        "content": system_prompt
+                    },
+                    *map(dict, filtered_messages),
+                    weaviate_response,
+                ],
+                temperature=temperature,
+                max_tokens=max_tokens,
+            )
+        else :
+            response = openai.ChatCompletion.create(
+                model=model,
+                messages=[
+                    {
+                        "role": "system",
+                        "content": system_prompt
+                    },
+                    *map(dict, filtered_messages),
+                ],
+                temperature=temperature,
+                max_tokens=max_tokens,
+            )
         return response["choices"][0]["message"]["content"]
 
 class HuggingFace:
diff --git a/textbase/vector_database.py b/textbase/vector_database.py
new file mode 100644
index 00000000..9fab96ff
--- /dev/null
+++ b/textbase/vector_database.py
@@ -0,0 +1,52 @@
+import weaviate
+from textbase import Message,Content
+import json
+from langchain.vectorstores import Weaviate
+from langchain.embeddings.openai import OpenAIEmbeddings
+
+class WeaviateClass:
+    api_key = None
+    host = None
+    auth_key = None
+    vector_db_data_class = None
+    @classmethod
+    def search_in_weaviate(
+        cls,
+        message_query:Message,
+        model_header_key: str,
+        max_vector_database_objects: int,
+    ):
+        assert cls.api_key is not None, "OpenAI API key is not set."
+        assert cls.host is not None, "Waaviate Host is not set."
+        # auth key is optional so can be skipped for None verification
+        weaviate_client = weaviate.Client(
+            url = cls.host,
+            auth_client_secret=cls.auth_key,
+            additional_headers = {
+                model_header_key: cls.api_key,
+            }
+        )
+        user_query = message_query['content'][0]['value']
+        embeddings = OpenAIEmbeddings(
+                openai_api_key = cls.api_key
+            )
+
+        weaviate_vectorstore = Weaviate(
+            weaviate_client,
+            "Documents",
+            "text",
+            embedding = embeddings
+        )
+        weaviate_response = weaviate_vectorstore.similarity_search(user_query)
+        messages = []
+        for response in weaviate_response:
+            messages.append(response.page_content)
+        
+        return   {
+            "role": message_query['role'],
+            # token limit will exceed so taking only 1000 char 
+            "content": json.dumps(messages,indent=4)[:1000]
+        }
+    
+
+        
\ No newline at end of file

From 799cfaf811e8b4bbd1f1a2019497990c38a9d987 Mon Sep 17 00:00:00 2001
From: hashcode-ankit <2698ankitsharma@gmail.com>
Date: Sun, 3 Sep 2023 20:26:17 +0530
Subject: [PATCH 2/6] reformatting

---
 examples/openai-bot/main.py | 9 +++++----
 textbase/models.py          | 3 +--
 textbase/vector_database.py | 8 +++++---
 3 files changed, 11 insertions(+), 9 deletions(-)

diff --git a/examples/openai-bot/main.py b/examples/openai-bot/main.py
index bb92f965..638750c6 100644
--- a/examples/openai-bot/main.py
+++ b/examples/openai-bot/main.py
@@ -3,14 +3,15 @@
 from typing import List
 
 # Load your OpenAI API key
-OpenAI.api_key = "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"
+OpenAI.api_key = ""
 
 # optional if you want to use vector database 
 # currently textbase support weaviate
-OpenAI.vector_db_host = "http://20.193.138.42:8080"
-
+OpenAI.vector_db_host = "http://20.345.234.42:8080"
+# vector database class where you stored the objects
 OpenAI.vector_db_data_class = "Documents"
-OpenAI.max_vector_database_objects = 1
+# lenght of response you want to take from weaviate
+OpenAI.max_weaviate_res_length = 1000
 
 
 # Prompt for GPT-3.5 Turbo
diff --git a/textbase/models.py b/textbase/models.py
index 0124eb41..ee0a3410 100644
--- a/textbase/models.py
+++ b/textbase/models.py
@@ -5,7 +5,6 @@
 import typing
 import traceback
 from textbase import Message, WeaviateClass
-# from vector_database import Weaviate
 
 # Return list of values of content.
 def get_contents(message: Message, data_type: str):
@@ -56,7 +55,7 @@ def generate(
             WeaviateClass.host = cls.vector_db_host
             WeaviateClass.auth_key = cls.vector_db_auth_key
             WeaviateClass.api_key = cls.api_key
-            WeaviateClass.vector_db_data_class = cls.vector_db_data_class
+            WeaviateClass.max_weaviate_res_length = cls.max_weaviate_res_length
             weaviate_response = WeaviateClass.search_in_weaviate(message_history[-1],"X-OpenAI-Api-Key",cls.max_vector_database_objects)
         
             response = openai.ChatCompletion.create(
diff --git a/textbase/vector_database.py b/textbase/vector_database.py
index 9fab96ff..f2d087dc 100644
--- a/textbase/vector_database.py
+++ b/textbase/vector_database.py
@@ -18,6 +18,7 @@ def search_in_weaviate(
     ):
         assert cls.api_key is not None, "OpenAI API key is not set."
         assert cls.host is not None, "Waaviate Host is not set."
+        assert cls.max_weaviate_res_length is not None,"max_weaviate_res_length is not set "
         # auth key is optional so can be skipped for None verification
         weaviate_client = weaviate.Client(
             url = cls.host,
@@ -26,6 +27,7 @@ def search_in_weaviate(
                 model_header_key: cls.api_key,
             }
         )
+        # take out user query 
         user_query = message_query['content'][0]['value']
         embeddings = OpenAIEmbeddings(
                 openai_api_key = cls.api_key
@@ -33,7 +35,7 @@ def search_in_weaviate(
 
         weaviate_vectorstore = Weaviate(
             weaviate_client,
-            "Documents",
+            cls.vector_db_data_class,
             "text",
             embedding = embeddings
         )
@@ -44,8 +46,8 @@ def search_in_weaviate(
         
         return   {
             "role": message_query['role'],
-            # token limit will exceed so taking only 1000 char 
-            "content": json.dumps(messages,indent=4)[:1000]
+            # token limit will exceed so taking limit
+            "content": json.dumps(messages,indent=4)[:cls.max_weaviate_res_length]
         }
     
 

From 25f3479eb6c94269d704ca139f4e0ad6d52be377 Mon Sep 17 00:00:00 2001
From: hashcode-ankit <2698ankitsharma@gmail.com>
Date: Sun, 3 Sep 2023 20:44:09 +0530
Subject: [PATCH 3/6] maxmax_weaviate_res_length introduces

---
 examples/openai-bot/main.py | 12 +++++-------
 textbase/models.py          |  5 +++--
 textbase/vector_database.py |  1 -
 3 files changed, 8 insertions(+), 10 deletions(-)

diff --git a/examples/openai-bot/main.py b/examples/openai-bot/main.py
index 638750c6..e24ca1cd 100644
--- a/examples/openai-bot/main.py
+++ b/examples/openai-bot/main.py
@@ -3,21 +3,19 @@
 from typing import List
 
 # Load your OpenAI API key
-OpenAI.api_key = ""
+OpenAI.api_key = "sk-6IUNBlbkFJkzBaDKIYHubR6q2I23"
 
 # optional if you want to use vector database 
 # currently textbase support weaviate
-OpenAI.vector_db_host = "http://20.345.234.42:8080"
-# vector database class where you stored the objects
+OpenAI.vector_db_host = "http://193.123.12.42:8080"
 OpenAI.vector_db_data_class = "Documents"
-# lenght of response you want to take from weaviate
-OpenAI.max_weaviate_res_length = 1000
+OpenAI.max_weaviate_res_length = 1
 
 
 # Prompt for GPT-3.5 Turbo
 SYSTEM_PROMPT = """You are chatting with an AI. There are no specific prefixes for responses, so you can ask or talk about anything you like.
 The AI will respond in a natural, conversational manner. Feel free to start the conversation with any question or topic, and let's have a
-pleasant chat!
+pleasant chat! You can use the data given at 
 """
 
 @bot()
@@ -50,4 +48,4 @@ def on_message(message_history: List[Message], state: dict = None):
     return {
         "status_code": 200,
         "response": response
-    }
\ No newline at end of file
+    }
diff --git a/textbase/models.py b/textbase/models.py
index ee0a3410..54dda705 100644
--- a/textbase/models.py
+++ b/textbase/models.py
@@ -30,7 +30,7 @@ class OpenAI:
     vector_db_host = None
     vector_db_auth_key = None
     vector_db_data_class = None
-    max_vector_database_objects = None
+    max_weaviate_res_length = None
     @classmethod
     def generate(
         cls,
@@ -56,7 +56,8 @@ def generate(
             WeaviateClass.auth_key = cls.vector_db_auth_key
             WeaviateClass.api_key = cls.api_key
             WeaviateClass.max_weaviate_res_length = cls.max_weaviate_res_length
-            weaviate_response = WeaviateClass.search_in_weaviate(message_history[-1],"X-OpenAI-Api-Key",cls.max_vector_database_objects)
+            WeaviateClass.vector_db_data_class = cls.vector_db_data_class
+            weaviate_response = WeaviateClass.search_in_weaviate(message_history[-1],"X-OpenAI-Api-Key")
         
             response = openai.ChatCompletion.create(
                 model=model,
diff --git a/textbase/vector_database.py b/textbase/vector_database.py
index f2d087dc..f35f8a36 100644
--- a/textbase/vector_database.py
+++ b/textbase/vector_database.py
@@ -14,7 +14,6 @@ def search_in_weaviate(
         cls,
         message_query:Message,
         model_header_key: str,
-        max_vector_database_objects: int,
     ):
         assert cls.api_key is not None, "OpenAI API key is not set."
         assert cls.host is not None, "Waaviate Host is not set."

From 25350a18bf51a49b550228e325f48f11ff1cf225 Mon Sep 17 00:00:00 2001
From: hashcode-ankit <2698ankitsharma@gmail.com>
Date: Sun, 3 Sep 2023 21:27:31 +0530
Subject: [PATCH 4/6] more refactoring

---
 examples/openai-bot/main.py | 11 ++++----
 textbase/models.py          | 56 +++++++++++++------------------------
 textbase/vector_database.py | 33 ++++++++++------------
 3 files changed, 41 insertions(+), 59 deletions(-)

diff --git a/examples/openai-bot/main.py b/examples/openai-bot/main.py
index e24ca1cd..fd62c7a1 100644
--- a/examples/openai-bot/main.py
+++ b/examples/openai-bot/main.py
@@ -3,19 +3,20 @@
 from typing import List
 
 # Load your OpenAI API key
-OpenAI.api_key = "sk-6IUNBlbkFJkzBaDKIYHubR6q2I23"
+OpenAI.api_key = ""
 
 # optional if you want to use vector database 
 # currently textbase support weaviate
-OpenAI.vector_db_host = "http://193.123.12.42:8080"
-OpenAI.vector_db_data_class = "Documents"
-OpenAI.max_weaviate_res_length = 1
+OpenAI.weaviate_host = "http://12.345.138.42:8080"
+OpenAI.weaviate_data_class = "Documents"
+# if using free OpenAI key set limit to avoid limit reach error
+OpenAI.max_weaviate_res_length = 1000
 
 
 # Prompt for GPT-3.5 Turbo
 SYSTEM_PROMPT = """You are chatting with an AI. There are no specific prefixes for responses, so you can ask or talk about anything you like.
 The AI will respond in a natural, conversational manner. Feel free to start the conversation with any question or topic, and let's have a
-pleasant chat! You can use the data given at 
+pleasant chat!. Use the information provided by vector database here : {vector_database_response}.
 """
 
 @bot()
diff --git a/textbase/models.py b/textbase/models.py
index 54dda705..272746a6 100644
--- a/textbase/models.py
+++ b/textbase/models.py
@@ -27,9 +27,9 @@ def extract_content_values(message: Message):
 
 class OpenAI:
     api_key = None
-    vector_db_host = None
-    vector_db_auth_key = None
-    vector_db_data_class = None
+    weaviate_host = None
+    weaviate_auth_key = None
+    weaviate_data_class = None
     max_weaviate_res_length = None
     @classmethod
     def generate(
@@ -49,42 +49,26 @@ def generate(
             contents = get_contents(message, "STRING")
             if contents:
                 filtered_messages.extend(contents)
+
         weaviate_response = None
         # if vector database host provided get response from weaviate
-        if cls.vector_db_host :
-            WeaviateClass.host = cls.vector_db_host
-            WeaviateClass.auth_key = cls.vector_db_auth_key
-            WeaviateClass.api_key = cls.api_key
-            WeaviateClass.max_weaviate_res_length = cls.max_weaviate_res_length
-            WeaviateClass.vector_db_data_class = cls.vector_db_data_class
-            weaviate_response = WeaviateClass.search_in_weaviate(message_history[-1],"X-OpenAI-Api-Key")
+        if cls.weaviate_host :
+            weaviate_response = WeaviateClass.search_in_weaviate(cls.api_key,cls.weaviate_host,cls.weaviate_auth_key,cls.weaviate_data_class,message_history[-1],cls.max_weaviate_res_length,"X-OpenAI-Api-Key")
         
-            response = openai.ChatCompletion.create(
-                model=model,
-                messages=[
-                    {
-                        "role": "system",
-                        "content": system_prompt
-                    },
-                    *map(dict, filtered_messages),
-                    weaviate_response,
-                ],
-                temperature=temperature,
-                max_tokens=max_tokens,
-            )
-        else :
-            response = openai.ChatCompletion.create(
-                model=model,
-                messages=[
-                    {
-                        "role": "system",
-                        "content": system_prompt
-                    },
-                    *map(dict, filtered_messages),
-                ],
-                temperature=temperature,
-                max_tokens=max_tokens,
-            )
+        # append the vector databases result in system prompt for better answers 
+        system_prompt= system_prompt.format(vector_database_response = weaviate_response)
+        response = openai.ChatCompletion.create(
+            model=model,
+            messages=[
+                {
+                    "role": "system",
+                    "content": system_prompt
+                },
+                *map(dict, filtered_messages),
+            ],
+            temperature=temperature,
+            max_tokens=max_tokens,
+        )
         return response["choices"][0]["message"]["content"]
 
 class HuggingFace:
diff --git a/textbase/vector_database.py b/textbase/vector_database.py
index f35f8a36..2b6c32c8 100644
--- a/textbase/vector_database.py
+++ b/textbase/vector_database.py
@@ -5,36 +5,33 @@
 from langchain.embeddings.openai import OpenAIEmbeddings
 
 class WeaviateClass:
-    api_key = None
-    host = None
-    auth_key = None
-    vector_db_data_class = None
     @classmethod
     def search_in_weaviate(
         cls,
+        api_key: str,
+        host: str,
+        auth_key:str,
+        weaviate_data_class,
         message_query:Message,
+        max_weaviate_res_length: int,
         model_header_key: str,
     ):
-        assert cls.api_key is not None, "OpenAI API key is not set."
-        assert cls.host is not None, "Waaviate Host is not set."
-        assert cls.max_weaviate_res_length is not None,"max_weaviate_res_length is not set "
-        # auth key is optional so can be skipped for None verification
         weaviate_client = weaviate.Client(
-            url = cls.host,
-            auth_client_secret=cls.auth_key,
+            url = host,
+            auth_client_secret=auth_key,
             additional_headers = {
-                model_header_key: cls.api_key,
+                model_header_key: api_key,
             }
         )
         # take out user query 
         user_query = message_query['content'][0]['value']
         embeddings = OpenAIEmbeddings(
-                openai_api_key = cls.api_key
+                openai_api_key = api_key
             )
 
         weaviate_vectorstore = Weaviate(
             weaviate_client,
-            cls.vector_db_data_class,
+            weaviate_data_class,
             "text",
             embedding = embeddings
         )
@@ -42,12 +39,12 @@ def search_in_weaviate(
         messages = []
         for response in weaviate_response:
             messages.append(response.page_content)
+        response_string = json.dumps(messages)
         
-        return   {
-            "role": message_query['role'],
-            # token limit will exceed so taking limit
-            "content": json.dumps(messages,indent=4)[:cls.max_weaviate_res_length]
-        }
+        # if token limit exceed error come user can configure max_weaviate_res_length to be considered for output
+        if max_weaviate_res_length  and len(response_string)>max_weaviate_res_length :
+            response_string = response_string[:max_weaviate_res_length]
+        return  response_string
     
 
         
\ No newline at end of file

From b2f23ee76026f0d52aac7ecd8c6a22ac1bebaf8b Mon Sep 17 00:00:00 2001
From: hashcode-ankit <2698ankitsharma@gmail.com>
Date: Sun, 3 Sep 2023 21:35:07 +0530
Subject: [PATCH 5/6] todo added

---
 textbase/models.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/textbase/models.py b/textbase/models.py
index 272746a6..e3184736 100644
--- a/textbase/models.py
+++ b/textbase/models.py
@@ -51,9 +51,10 @@ def generate(
                 filtered_messages.extend(contents)
 
         weaviate_response = None
-        # if vector database host provided get response from weaviate
+        # if weaviate_host provided get response from weaviate
         if cls.weaviate_host :
             weaviate_response = WeaviateClass.search_in_weaviate(cls.api_key,cls.weaviate_host,cls.weaviate_auth_key,cls.weaviate_data_class,message_history[-1],cls.max_weaviate_res_length,"X-OpenAI-Api-Key")
+        # Todo: support for other vector database 
         
         # append the vector databases result in system prompt for better answers 
         system_prompt= system_prompt.format(vector_database_response = weaviate_response)

From 96882694dcb34355b4129b1310c808489ad33600 Mon Sep 17 00:00:00 2001
From: hashcode-ankit <2698ankitsharma@gmail.com>
Date: Sun, 3 Sep 2023 22:52:50 +0530
Subject: [PATCH 6/6] documentation update

---
 docs/docs/examples/openai-bot.md | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/docs/docs/examples/openai-bot.md b/docs/docs/examples/openai-bot.md
index a7df7b7c..b981eff3 100644
--- a/docs/docs/examples/openai-bot.md
+++ b/docs/docs/examples/openai-bot.md
@@ -14,10 +14,16 @@ from typing import List
 # Load your OpenAI API key
 OpenAI.api_key = ""
 
+# For using weaviate vector database
+OpenAI.weaviate_host = "http://23.345.138.42:8080"
+OpenAI.weaviate_data_class = "Documents"
+# if using free OpenAI key set limit to avoid limit reach error
+OpenAI.max_weaviate_res_length = 1000
 # Prompt for GPT-3.5 Turbo
 SYSTEM_PROMPT = """You are chatting with an AI. There are no specific prefixes for responses, so you can ask or talk about anything you like.
 The AI will respond in a natural, conversational manner. Feel free to start the conversation with any question or topic, and let's have a
 pleasant chat!
+Use the information provided by vector database here : {vector_database_response}.
 """
 
 @bot()