Skip to content

Latest commit

 

History

History
95 lines (69 loc) · 2.73 KB

README.md

File metadata and controls

95 lines (69 loc) · 2.73 KB
Raw

✨ Versa: Your Personal AI Companion ✨

Relaxed conversations, Real results!

Versa's API

Overview

The Versa API provides programmatic access to the Versa chatbot, allowing developers to integrate its functionalities into their applications.

Getting Started:

Using the API

Doc Bot

To use the Doc Bot functionality, set the chatbot_type to DOCBOT and provide the file path to the document:

from chatbot.engine import DocBotEngine

doc_bot = DocBotEngine(
        file_path=<file_path>,
        model_name_or_path=<model_name_or_path>,
        quantize=<whether to quantize model>,
        llm_temp=<llm temperature>,
        access_token=<(optional) access_token>,
)
question = "What is the typical size of a blue whale?"
response_message = doc_bot.get_response(query=question)
print(
    f"{CHAT_SEPARATOR}Question: {response_message.query}{CHAT_SEPARATOR}Response: {response_message.response}"
)

Chat Bot

To use the general chatbot functionality, choose from the available chatbot_type options:

from chatbot.engine import ChatBotEngine

# Available chatbot types: {'Therapist', 'Comedian', 'Default', 'Child', 'Expert'}
chat_bot = ChatBotEngine(
        chatbot_type=<type of chatbot>,
        model_name_or_path=<model_name_or_path>,
        quantize=<whether to quantize model>,
        llm_temp=<llm temperature>,
        access_token=<(optional) access_token>,
)
question = "What is 2/2?"
response_message = chat_bot.get_response(query=question)
print(
    f"{CHAT_SEPARATOR}Question: {response_message.query}{CHAT_SEPARATOR}Response: {response_message.response}"
)

Chatbot Types:

The exact names of chatbot types are:

  1. Therapist
  2. Expert
  3. Comedian
  4. Child
  5. Default

These can be imported from chatbot.constants

Arguments Available

ChatBotEngine:

  1. chatbot_type: type of chatbot to be used.

DocBotEngine:

  1. file_path: File path for processing.

Common Arguments:

  1. model_name_or_path: model name or path.
  2. quantize: whether to quantize model or not.
  3. llm_temp: Parameter influencing the balance between predictability and creativity in generated text (less than 1 for more deterministic or greater than 1 for more creative)
  4. access_token: (Optional). Access token for gated models in huggingface

Considerations

  • When using Doc Bot, passing file_path is mandatory.
  • If you are passing a gated model, please provide access_token.

Response Format

The get_response method returns a ResponseMessage object with the following attributes:

  • query: The original query.
  • response: The generated response.

License

This project is licensed under the GNU Affero General Public License (AGPL) (see it here - LICENSE ).