Merge pull request #30 from startmunich/27-add-openai-api-key-and-cli…

…ents-in-code

first try of openAI

apps/slackbot/ai_fcts.py

@@ -1,14 +1,11 @@
 import os
 from dotenv import load_dotenv
 from langchain.prompts import PromptTemplate
-from langchain_community.llms import Replicate
+from openai import OpenAI
 from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler
-from langchain.chains import RetrievalQA
 from langchain_community.vectorstores.qdrant import Qdrant
 from qdrant_client import QdrantClient
 from langchain_community.embeddings import InfinityEmbeddings
-from langchain import hub
-from datetime import date
 
 
 
@@ -18,8 +15,6 @@
 # load_dotenv(dotenv_path="apps/slackbot/.env.local")
 
 # trigger new build
-
-# REPLICATE_API_TOKEN  = os.environ["REPLICATE_API_KEY"]
 qdrant_uri = os.environ["QDRANT_URL"]
 qdrant_collection_name = os.environ.get("QDRANT_COLLECTION_NAME")
 infinity_api_url = os.environ.get("INFINITY_URL")
@@ -40,45 +35,11 @@
 retriever = qdrant_db.as_retriever(search_kwargs={"k": 3})
 
 # Create the language model
-llm = Replicate(
-    streaming=True,
-    callbacks=[StreamingStdOutCallbackHandler()],
-    model=llm_model,
-    model_kwargs={"temperature": 0.1, "max_length": 1500, "top_p": 0.9, "top_k": 50, "max_new_tokens": 400,
-        "min_new_tokens": 20, "repetition_penalty": 0.1},
-    verbose = False
-)
-
-
-
-# Create the prompt template
-prompt_template = """ [INST]
-You are StartGPT, an assistant for question-answering tasks.
-The context you get will be from our Notion, Website and Slack. Use this context to answer the question.
-If you utilize context from Notion to answer the question, please provide the source link in your answer.
-If you utilize context from slack or the website, do not provide a link.
-
-<Beginning of context>
-{context} 
-<End of context>
+client = OpenAI(
+    api_key=os.environ.get("OPENAI_KEY"),
+    )
 
-<Beginning of question>
-{question}
-<End of question>
-[INST]
-"""
 
-# Initialize prompt
-prompt_template = PromptTemplate(input_variables=["question", "context"], template=prompt_template)
-
-# prompt = hub.pull("rlm/rag-prompt")
-
-# qa_chain = RetrievalQA.from_chain_type(
-#     llm=llm,
-#     retriever=retriever,
-#     chain_type_kwargs={"prompt": prompt_template},
-#     verbose=False
-# )
 def format_docs(docs):
     context = ""
     for index, doc in enumerate(docs):
@@ -98,10 +59,19 @@ def get_answer(query: str) -> str:
     context = format_docs(docs)
 
     # enter context into the prompt
-
-    prompt = prompt_template.format(context=context, question=query)
-
-    response = llm.invoke(prompt)
+    completion = client.chat.completions.create(
+        model="gpt-4o",
+        messages=[
+            {"role": "system", "content": """You are StartGPT, an assistant for question-answering tasks.
+            The context you get will be from our Notion, Website and Slack. Use this context to answer the question.
+            If you utilize context from Notion to answer the question, please provide the source link in your answer.
+            If you utilize context from slack or the website, do not provide a link."""}, # <-- This is the system message that provides context to the model
+            {"role": "user", "content": f"""<Beginning of context> {context} <End of context> \n
+            <Beginning of question> {query} <End of question>"""}  # <-- This is the user message for which the model will generate a response
+        ]
+    )
+
+    response = completion.choices[0].message.content
     # print(f"Retrieved docs: {docs}\n Retrieved context: {context}\n Generated response: {response}")
 
     return f"{response}"

apps/slackbot/requirements.txt

@@ -4,5 +4,5 @@ python-dotenv~=1.0.1
 langchain == 0.1.12
 langchain_community
 langchainhub
-replicate == 0.24.0
+openai == 1.3.7
 qdrant-client == 1.6.4

infrastructure/docker-compose.yml

@@ -194,7 +194,7 @@ services:
       INFINITY_MODEL: BAAI/bge-large-en-v1.5
       SLACK_APP_TOKEN: $SLACK_APP_TOKEN
       SLACK_BOT_TOKEN: $SLACK_BOT_TOKEN
-      REPLICATE_API_TOKEN: $REPLICATE_API_TOKEN
+      OPENAI_KEY: $OPENAI_KEY
       LLM_MODEL: mistralai/mixtral-8x7b-instruct-v0.1:2b56576fcfbe32fa0526897d8385dd3fb3d36ba6fd0dbe033c72886b81ade93e
 
   webcrawler: