ConvoAI is an open-source conversational chatbot framework designed to simplify and streamline the development of powerful conversational AI chatbots. Built on the robust sentence transformer model 'paraphrase-distilroberta-base-v1,' ConvoAI excels in understanding semantic similarity and efficiently mapping user inputs to intents with minimal data.
-
Developer-Friendly: ConvoAI prioritizes ease of use for developers, simplifying the chatbot creation process.
-
Versatile Custom Actions: The framework allows developers to integrate custom actions, enabling external operations such as API calls and database queries to enhance the bot's capabilities.
-
Rule-Based Development: Chatbot development is rule-based, with defined intent prompts mapped to responses or custom actions. Training is skipped, leveraging the transformer model's proficiency in capturing semantic similarity.
-
Flexible Intent Definitions: Intent prompts require a minimum of one example, but developers can include additional examples for intents with diverse phrasings. Responses can be either text or custom actions.
-
Entities and Slots Management: ConvoAI efficiently manages entities and slots, facilitating the extraction and storage of crucial information from user input.
- Data.yaml: Defines intents, prompts, and responses.
- Entities.yaml: Maps entity examples to entity types.
- Slots.yaml: Configures memory slots for storing important information.
- Forms.yaml: Defines structured forms for information acquisition.
- Rules.yaml: Designs conversation flows and paths.
- Config.yaml: Manages the integrated action execution engine.
- Define intents, prompts, and responses in Data.yaml.
- Map entity examples to entity types in Entities.yaml.
- Configure memory slots in Slots.yaml.
- Define structured forms in Forms.yaml.
- Design conversation flows in Rules.yaml.
- Execute actions seamlessly with the integrated action execution engine.
pip install convoai
Refer to the documentation and example files for detailed usage instructions and best practices.
Contributions are welcome! Check out the Contribution Guidelines for details.
This project is licensed under the MIT License.
