Add basic chain with memory

src/config.yaml

@@ -8,3 +8,6 @@ nlp:
   token: "${OPENAI_TOKEN}"
   model: "gpt-3.5-turbo"
 
+api:
+  serpapi: "${SERPAPI_TOKEN}"
+

src/language_parser.py

@@ -4,7 +4,8 @@ from config import config
 import nlp
 import logging
 
-completion = nlp.CompletionOpenAI(config["nlp.token"])
+completion = nlp.CompletionOpenAI(token=config["nlp.token"],
+                                  serpapi=config["api.serpapi"])
 
 async def parser(client, message: str):
     """

src/main.py

@@ -21,10 +21,9 @@ async def on_ready():
 
 @client.event
 async def on_message(message):
-    print(message)
     logging.info(f"Message received : {message.content}")
-    if message.author == client.user:
-        return
+    #if message.author == client.user:
+    #    return
     if len(message.content) == 0: # Ignore empty messages
         return
     #if message.content.startswith(config["discord.prefix"]):

src/nlp/Completion.py

@@ -1,9 +1,24 @@
+from langchain.agents import load_tools
+from langchain.agents import ZeroShotAgent, Tool, AgentExecutor
+from langchain.memory import ConversationBufferMemory
+from langchain.agents import initialize_agent
+from langchain.agents import AgentType
+from langchain import LLMChain
+
+
 class CompletionMeta(type):
     """
     The meta class for completion interface
     """
 
+tools = None
+
 class Completion():
+    def __init__(self):
+        self.prefix = None
+        self.suffix = None
+        self.promptVars = None
+
     def complete(self, message: str) -> str:
         """
         Perform a text completion using the language model
@@ -15,3 +30,52 @@ class Completion():
         Return the model name
         """
         pass
+
+    def getChain(self, llm):
+        pass
+
+    def getAgent(self, llm):
+        # Load tools
+        tools = load_tools(["serpapi", "llm-math"], llm=llm)
+
+        # Build a prompt
+        prompt = ZeroShotAgent.create_prompt(
+                tools=tools,
+                prefix=self.getPromptPrefix(),
+                suffix=self.getPromptSuffix(),
+                input_variables=self.getPromptVariables()
+                )
+        memory = ConversationBufferMemory(memory_key="chat_history")
+
+        # Build the LLM Chain
+        llm_chain = LLMChain(llm=llm, prompt=prompt)
+        agent = ZeroShotAgent(llm_chain=llm_chain,
+                              tools=tools,
+                              verbose=True)
+        self.agent_chain = AgentExecutor.from_agent_and_tools(agent=agent,
+                                                         tools=tools,
+                                                         verbose=True,
+                                                         memory=memory)
+
+        return self.agent_chain
+
+    def getPromptVariables(self):
+        if self.promptVars is None:
+            self.promptVars = ["input", "chat_history", "agent_scratchpad"]
+        return self.promptVars
+
+    def getPromptPrefix(self):
+        if self.prefix is None:
+            self.prefix = """Have a conversation with a human, answering the following
+            questions as best you can. You have access to the following tools:"""
+        return self.prefix
+
+    def getPromptSuffix(self):
+        if self.suffix is None:
+            self.suffix = """Begin!
+            {chat_history}
+            Question: {input}
+            {agent_scratchpad}"""
+        return self.suffix
+
+

src/nlp/CompletionOpenAI.py

@@ -1,23 +1,26 @@
-import openai
 from .Completion import Completion
+from langchain.llms import OpenAI
+import os
 
 class CompletionOpenAI(Completion):
     # Constructor
-    def __init__(self, token: str, model: str = "gpt-3.5-turbo"):
+    def __init__(self,
+                 token: str,
+                 serpapi: str,
+                 temperature: float = 0.5):
+        super().__init__()
+
+        os.environ["SERPAPI_API_KEY"] = serpapi
         self.token = token
-        self.model = model
-        openai.api_key = self.token
+
+        self.temperature = temperature
+        self.llm = OpenAI(temperature=self.temperature,
+                     openai_api_key=self.token)
+        self.agent = self.getAgent(self.llm)
+
 
     def complete(self, message: str) -> str:
-        chat_completion = openai.ChatCompletion.create(
-                                        model=self.model,
-                                        messages=[
-                                           {
-                                               "role": "user",
-                                               "content": message
-                                           }
-                                        ])
-        return chat_completion.choices[0].message.content
+        return self.agent_chain.run(message)
 
     def getModelName(self) -> str:
-        return self.model
+        return ""

src/requirements.txt

@@ -2,10 +2,33 @@ aiohttp==3.8.4
 aiosignal==1.3.1
 async-timeout==4.0.2
 attrs==23.1.0
+certifi==2023.5.7
 charset-normalizer==3.1.0
+dataclasses-json==0.5.7
 discord==2.2.3
 discord.py==2.2.3
+envyaml==1.10.211231
 frozenlist==1.3.3
+google-search-results==2.4.2
+greenlet==2.0.2
 idna==3.4
+langchain==0.0.178
+marshmallow==3.19.0
+marshmallow-enum==1.5.1
 multidict==6.0.4
+mypy-extensions==1.0.0
+numexpr==2.8.4
+numpy==1.24.3
+openai==0.27.7
+openapi-schema-pydantic==1.2.4
+packaging==23.1
+pydantic==1.10.8
+PyYAML==6.0
+requests==2.31.0
+SQLAlchemy==2.0.15
+tenacity==8.2.2
+tqdm==4.65.0
+typing-inspect==0.8.0
+typing_extensions==4.6.0
+urllib3==2.0.2
 yarl==1.9.2