digiteinfotech/kairon digiteinfotech/kairon

  • Stars
    star
    247
  • Rank 164,117 (Top 4 %)
  • Language
    Python
  • License
    Apache License 2.0
  • Created over 4 years ago
  • Updated about 2 months ago
Twitter logo Share LinkedIn logo Share Facebook logo Share
digiteinfotech/kairon
github

digiteinfotech

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

Conversational AI Platform to build effective Proactive Digital Assistants using Visual LLM Chaining

Python application Codacy Badge Coverage Status

Kairon : You can directly converse with Kairon on Whatsapp and stay informed of our latest releases !

qrcode

Kairon is now envisioned as a conversational digital transformation platform that helps build LLM based digital assistants at scale. It is designed to make the lives of those who work with ai-assistants easy, by giving them a no-coding web interface to adapt , train , test and maintain such assistants . We are now enhancing the backbone of Kairon with a full fledged context management system to build proactive digital assistants .

What is Kairon?

Kairon is currently a set of tools built on the RASA framework with a helpful UI interface . While RASA focuses on technology of chatbots itself. Kairon on the other hand focuses on technology that deal with pre-processing of data that are needed by this framework. These include question augmentation and generation of knowledge graphs that can be used to automatically generate intents, questions and responses. It also deals with the post processing and maintenance of these bots such metrics / follow-up messages etc.

What can it do?

Kairon is open-source. It is a Conversational digital transformation platform: Kairon is a platform that allows companies to create and deploy digital assistants to interact with customers in a conversational manner.

End-to-end lifecycle management: Kairon takes care of the entire digital assistant lifecycle, from creation to deployment and monitoring, freeing up company resources to focus on other tasks. Tethered digital assistants: Kairon’s digital assistants are tethered to the platform, which allows for real-time monitoring of their performance and easy maintenance and updates as needed.

Low-code/no-code interface: Kairon’s interface is designed to be easy for functional users, such as marketing teams or product management, to define how the digital assistant responds to user queries without needing extensive coding skills. Secure script injection: Kairon’s digital assistants can be easily deployed on websites and SAAS products through secure script injection, enabling organizations to offer better customer service and support.

Kairon Telemetry: Kairon’s telemetry feature monitors how users are interacting with the website/product where Kairon was injected and proactively intervenes if they are facing problems, improving the overall user experience. Chat client designer: Kairon’s chat client designer feature allows organizations to create customized chat clients for their digital assistants, which can enhance the user experience and help build brand loyalty.

Analytics module: Kairon’s analytics module provides insights into how users are interacting with the digital assistant, enabling organizations to optimize their performance and provide better service to customers. Robust integration suite: Kairon’s integration suite allows digital assistants to be served in an omni-channel, multi-lingual manner, improving accessibility and expanding the reach of the digital assistant.

Realtime struggle analytics: Kairon’s digital assistants use real-time struggle analytics to proactively intervene when users are facing friction on the product/website where Kairon has been injected, improving user satisfaction and reducing churn. This website can be found at Kairon and is hosted by NimbleWork Inc.

Who uses it ?

Kairon is built for two personas Teams and Individuals who want an easy no-coding interface to create, train, test and deploy digital assistants . One can directly access these features from our hosted website. Teams who want to host the chatbot trainer in-house. They can build it using docker compose. Our teams current focus within NLP is Knowledge Graphs – Do let us know if you are interested.

At this juncture it layers on top of Rasa Open Source

Deployment

Kairon only requires a recent version of Docker and Docker Compose.

Please do the below changes in docker/docker-compose.yml

  1. set env variable server to public IP of the machine where trainer api docker container is running for example: http://localhost:81

  2. Optional, if you want to have google analytics enabled then uncomment trackingid and set google analytics tracking id

  3. set env variable SECRET_KEY to some random key.

    use below command for generating random secret key

    openssl rand -hex 32

  4. run the command.

    cd kairon/docker
docker-compose up -d

  5. Open http://localhost/ in browser.

  6. To Test use username: [email protected] and password: Changeit@123 to try with demo user

Development

Installation

  1. Kairon requires python 3.8 and mongo 4.0+

  2. Then clone this repo

    git clone https://github.com/digiteinfotech/kairon.git
cd kairon/

  3. For creating Virtual environment, please follow the link

  4. For installing dependencies

    Windows

    setup.bat   ```

**No Matching distribution found tensorflow-text** - remove the dependency from requirements.txt file, as **window version is not available** [#44](https://github.com/tensorflow/text/issues/44)

**Linux**

    chmod 777 ./setup.sh sh ./setup.sh

  5. For starting augmentation services run

    python -m uvicorn augmentation.paraphrase.server:app --host 0.0.0.0

  6. For starting trainer-api services run

    python -m uvicorn kairon.api.app.main:app --host 0.0.0.0 --port 8080

System Configuration

Email verification setup

The email.yaml file can be used to configure the process for account confirmation through a verification link sent to the user's mail id. It consists of the following parameters:

  • enable -

    set value to True for enabling email verification, and False to disable.

    You can also use the environment variable EMAIL_ENABLE to change the values.

  • url -

    this url, along with a unique token is sent to the user's mail id for account verification as well as for password reset tasks.

    You can also use the environment variable APP_URL to change the values.

  • email -

    the mail id of the account which sends the confirmation mail.

    You can also use the environment variable EMAIL_SENDER_EMAIL to change the values.

  • password -

    the password of the account which sends the confirmation mail.

    You can also use the environment variable EMAIL_SENDER_PASSWORD to change the values.

  • port -

    the port that is used to send the mail [For ex. "587"].

    You can also use the environment variable EMAIL_SENDER_PORT to change the values.

  • service -

    the mail service that is used to send the confirmation mail [For ex. "gmail"].

    You can also use the environment variable EMAIL_SENDER_SERVICE to change the values.

  • tls -

    set value to True for enabling transport layer security, and False to disable.

    You can also use the environment variable EMAIL_SENDER_TLS to change the values.

  • userid -

    the user ID for the mail service if you're using a custom service for sending mails.

    You can also use the environment variable EMAIL_SENDER_USERID to change the values.

  • confirmation_subject -

    the subject of the mail to be sent for confirmation.

    You can also use the environment variable EMAIL_TEMPLATES_CONFIRMATION_SUBJECT to change the subject.

  • confirmation_body -

    the body of the mail to be sent for confirmation.

    You can also use the environment variable EMAIL_TEMPLATES_CONFIRMATION_BODY to change the body of the mail.

  • confirmed_subject -

    the subject of the mail to be sent after confirmation.

    You can also use the environment variable EMAIL_TEMPLATES_CONFIRMED_SUBJECT to change the subject.

  • confirmed_body -

    the body of the mail to be sent after confirmation.

    You can also use the environment variable EMAIL_TEMPLATES_CONFIRMED_BODY to change the body of the mail.

  • password_reset_subject -

    the subject of the mail to be sent for password reset.

    You can also use the environment variable EMAIL_TEMPLATES_PASSWORD_RESET_SUBJECT to change the subject.

  • password_reset_body -

    the body of the mail to be sent for password reset.

    You can also use the environment variable EMAIL_TEMPLATES_PASSWORD_RESET_BODY to change the body of the mail.

  • password_changed_subject -

    the subject of the mail to be sent after changing the password.

    You can also use the environment variable EMAIL_TEMPLATES_PASSWORD_CHANGED_SUBJECT to change the subject.

  • password_changed_body -

    the body of the mail to be sent after changing the password.

    You can also use the environment variable EMAIL_TEMPLATES_PASSWORD_CHANGED_BODY to change the body of the mail.

Documentation

Documentation for all APIs for Kairon are still being fleshed out. A intermediary version of the documentation is available here. Documentation

Contribute

We ❤️ contributions of all size and sorts. If you find a typo, if you want to improve a section of the documentation or if you want to help with a bug or a feature, here are the steps:

  1. Fork the repo and create a new branch, say rasa-dx-issue1

  2. Fix/improve the codebase

  3. write test cases and documentation for code'

  4. run test cases.

python -m pytest
  1. reformat code using black
python -m black bot_trainer

  1. Commit the changes, with proper comments about the fix.

  2. Make a pull request. It can simply be one of your commit messages.

  3. Submit your pull request and wait for all checks passed.

  4. Request reviews from one of the developers from our core team.

  5. Get a đź‘Ť and PR gets merged.

Built With

Authors

The repository is being maintained and supported by NimbleWork Inc.

See also the list of contributors who participated in this project.

License

Licensed under the Apache License, Version 2.0. Copy of the license

A list of the Licenses of the dependencies of the project can be found at the Link