🍇 Berri AI: 1-click deploy for your LLM Apps ⚡️

Berri AI is a python package that helps you quickly and easily deploy your LLM App from Colab/VSCode/etc. directly to production. Just install the package, import the function, and run deploy. At the end of the deploy (~10-15mins), you will get:

1. 🎉 A web app to interact with your agent 👉 example (hint: ask the agent if you can accept international payments from india)
2. 😱 An endpoint you can query 👉 https://agent-repo-35aa2cf3-a0a1-4cf8-834f-302e5b7fe07e-45247-8aqi.zeet-team-ishaan-jaff.zeet.app/langchain_agent?query="who is obama?"

Demo Video / Walkthrough

🤓 Usage

There are 5 major ways you can use Berri

1. Pipelines: Best way to get started. Pipelines let you spin up an LLM App in 2-lines of code.
2. GPT-Index: If you're writing an LLM app with the primary method of interaction being gpt-index's .query() function.
3. Langchain: If you're writing an LLM app with the primary method of interaction being Langchain's 'initialize_agent' or 'AgentExecutor()' functions
4. Wrapper functions: For more complex use-cases. If you're taking a user query and doing multiple things (LLM calls, api calls, etc.) with it, you can put them in a wrapper function and pass the wrapper function to Berri.
5. Search Strategies: Improve search results for LLM agents.

👷 Pipelines

Today we support 2 pipelines:

• docQAPipeline: This lets you paste a url link to your documentation and get a shareable web app to use it, in 15mins. We'll handle the chunking, vectorizing, agent initialization, deployment for you.

1. Install the package:

   pip install berri-ai
   

2. Import the deploy method:

   from berri_ai import docQAPipeline
   

3. Initiate the deployment by providing your email address:

   docQAPipeline(user_email, open_ai_key, input_url)
   # example docQAPipeline(user_email="[email protected]", open_ai_key="sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx", input_url="https://stripe.com/docs/india-accept-international-payments#TransactionPurposeCode")
   

   go here to get your openai api key

• ComplexInformationQA: This let's you pass an index from GPT-Index, and create a Langchain agent that answers complex questions (e.g.: "my order didn't arrive, even though I'd paid for express shipping") + provides references (i.e. documents that the agent looked at)

1. Install the package:

   pip install berri-ai
   

2. Import the ComplexInformationQA method:

   from berri_ai.ComplexInformationQA import ComplexInformationQA
   

3. There are 2 ways to initialize a Complex Information Agent:

Before we go into the approaches, let's see how to initialize a CIQ Agent:

To initialize the CIQ Agent class, you will need to provide the following parameters:

• model_key: This is the API key used to access the OpenAI API.
• index: This is an optional parameter used to provide an index for the knowledgebase.
• prompt: This is an optional parameter used to provide a prompt for the agent.
• functions: This is a list of functions to be used by the agent (optional if you're passing an index).
• descriptions: This is a list of descriptions for each function in the list (optional if you're passing an index).

Sample Usage:

• Default: Initiate the Complex Information Agent by providing your stored vector-index:

  index = GPTSimpleVectorIndex.load_from_disk("./doc_qa.json")
  CIQAgent = ComplexInformationQA(<openai_api_key>, index)
  response = CIQAgent.run("my order didn't arrive, even though I'd paid for express shipping")
  

  Note: you will need to initialize your environment with your openai key (os.environ["OPENAI_API_KEY"] = <openai_api_key>) go here to get your openai api key

• Custom Tools: Pass your own tools to an agent

  functions = [search, noneInput]
  descriptions = ["This is the knowledge base to query. Only use keywords while querying this database. Do not use full sentences.", "This function takes a none input and returns a none output"]
  CIQAgent = ComplexInformationQA(<openai_api_key>, None, None, functions, descriptions)
  response = CIQAgent.run(user_input)
  

  Note: you will need to initialize your environment with your openai key (os.environ["OPENAI_API_KEY"] = <openai_api_key>) go here to get your openai api key

GPT-Index

To use Berri AI w/ GPT-Index, follow these steps:

1. Install the package:

   pip install berri-ai
   

2. Import the deploy method:

   from berri_ai import deploy_gpt_index
   

3. Initiate the deployment by providing your email address:

   deploy_gpt_index(user_email=<your email>)
   # example deploy_gpt_index(user_email="[email protected]")
   

   Note: Today, Berri will only look for the '.query()' function. Let us know if there are other use-cases you would like us to support.

Langchain

To use Berri AI w/ Langchain, follow these steps:

1. Install the package:

   pip install berri-ai
   

2. Import the deploy method:

   from berri_ai import deploy
   

3. Initiate the deployment by providing your email address:

   deploy(user_email=<your email>)
   # example deploy(user_email="[email protected]")
   

   Note: Today, berri will look for the initialize_agent() and AgentExecutor() functions in your code. If you're using another way of initializing your agent, let us know and we'll update the package to account for that.

Once deployment is complete, you will receive an email notification. The entire process usually takes 10-15 minutes.

Wrapper Functions

To use Berri AI w/ Wrapper Functions, follow these steps:

1. Install the package:

   pip install berri-ai
   

2. Import the deploy method:

   from berri_ai import deploy_func
   

3. Initiate the deployment by providing your email address:

   deploy_func(user_email=<your email>, executing_function=<stringified name of your executing function>, test_str=<test_user_input_query>)
   # example deploy_func(user_email="[email protected]", executing_function="print_answer", test_str="what is ManimML?")
   

📚 Examples

1. Berri LangChain Youtube Agent Example
2. Berri LangChain Search Agent Example
3. Berri GPT Index + Langchain Document QA Example

🚨 Support

If you have any questions or need help using Berri AI, join the Discord or Text/WhatsApp us @ +17708783106 📱.