feat: weaviate vector database feature added for openai

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()

examples/openai-bot/main.py

@@ -5,10 +5,18 @@
 # Load your OpenAI API key
 OpenAI.api_key = ""
 
+# optional if you want to use vector database 
+# currently textbase support weaviate
+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!
+pleasant chat!. Use the information provided by vector database here : {vector_database_response}.
 """
 
 @bot()
@@ -41,4 +49,4 @@ def on_message(message_history: List[Message], state: dict = None):
     return {
         "status_code": 200,
         "response": response
-    }
+    }

textbase/__init__.py

@@ -1,2 +1,4 @@
 from .bot import bot
-from .message import Message
+from .message import Message
+from .message import Content
+from .vector_database import WeaviateClass

textbase/models.py

@@ -4,8 +4,7 @@
 import time
 import typing
 import traceback
-
-from textbase import Message
+from textbase import Message, WeaviateClass
 
 # Return list of values of content.
 def get_contents(message: Message, data_type: str):
@@ -28,7 +27,10 @@ def extract_content_values(message: Message):
 
 class OpenAI:
     api_key = None
-
+    weaviate_host = None
+    weaviate_auth_key = None
+    weaviate_data_class = None
+    max_weaviate_res_length = None
     @classmethod
     def generate(
         cls,
@@ -40,7 +42,6 @@ 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:
@@ -49,6 +50,14 @@ def generate(
             if contents:
                 filtered_messages.extend(contents)
 
+        weaviate_response = None
+        # 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)
         response = openai.ChatCompletion.create(
             model=model,
             messages=[
@@ -61,7 +70,6 @@ def generate(
             temperature=temperature,
             max_tokens=max_tokens,
         )
-
         return response["choices"][0]["message"]["content"]
 
 class HuggingFace:

textbase/vector_database.py

@@ -0,0 +1,50 @@
+import weaviate
+from textbase import Message,Content
+import json
+from langchain.vectorstores import Weaviate
+from langchain.embeddings.openai import OpenAIEmbeddings
+
+class WeaviateClass:
+    @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,
+    ):
+        weaviate_client = weaviate.Client(
+            url = host,
+            auth_client_secret=auth_key,
+            additional_headers = {
+                model_header_key: api_key,
+            }
+        )
+        # take out user query 
+        user_query = message_query['content'][0]['value']
+        embeddings = OpenAIEmbeddings(
+                openai_api_key = api_key
+            )
+
+        weaviate_vectorstore = Weaviate(
+            weaviate_client,
+            weaviate_data_class,
+            "text",
+            embedding = embeddings
+        )
+        weaviate_response = weaviate_vectorstore.similarity_search(user_query)
+        messages = []
+        for response in weaviate_response:
+            messages.append(response.page_content)
+        response_string = json.dumps(messages)
+
+        # 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
+
+
+