This repository contains the back-end implementation of CoMMA, a conversational mashup modification agent.

The paper is published in the CHI 2022 conference as follows: https://doi.org/10.1145/3491102.3517655 (Open Access on ACM DL)

The implementation has been tested on Python 3.5.

1. Install requirements.

   pip3 install -r requirements.txt

2. Locate a valid SSL certificate on the root project directory with the filename of server.crt and server.key.
3. Import the agent to your dialogflow console, and configure the webhook settings. The agent files are located in the data folder.
4. Command sudo python3 mashup.py to initialize the mashup module.
5. Command sudo python3 app.py to initialize the back-end module.

• app.py : This Flask module handles Webhook responses from the Dialogflow front-end. A valid SSL certificate is required.

  The API endpoint description is as follows:

  • POST / handles a Dialogflow webhook request. Check the webhook request document for the detailed information.

• mashup.py : This Flask module handles the registered IoT mashups and their history.

  The API endpoint description is as follows:

  • GET / responds the current history data and modification log in a numbered list.
  • GET /{number} switches the current history & mashup set based on the given number.
    It would return a {"setId": "{number}", "status": "ok"} if okay, or a {"status": "error"} response otherwise. This operation will clear the current modification log.
  • GET /check responds the health of the mashup module.
    It would return a {"status": "ok"} json response if okay, or a {"status": "error"} response otherwise.
  • POST /search responds the search results of RulesHistory.search() method.
    It would return a {"status": "ok", "searchResults": "..."} json response if okay (even if found nothing), or a {"status": "error", "what": "... the reason of error ..."} response otherwise.
    The search results will be dumped by pickle library and be encoded into base64, so be cautious of any security issues!
  • POST /modify accumulates any modification inputs from a user.
    Similar to /search operation, it will search the target mashup using RulesHistory.search() with additional arguments named modifyType and modifyData, which indicate the user modification input.
  • POST /type receives a Google Dialogflow intent and checks the intent matches to the trigger or action of given search context. The detailed specification of json response is as follows:
    • boolean trigger: true if the trigger of given search context matches to the user input intent, false otherwise.
    • boolean action: true if the action of given search context matches to the user input intent, false otherwise.
    • list of NominalAction | NominalTrigger intentData: the nominal objects of user input intent, which are mapped from intents.json.

• logger.py : The logging module.

• test/* : Unit test scripts.

• src/*: They are the core source codes of this implementation. Check src/README.md for more details.

If you found these helpful, please consider citing our paper.

@inproceedings{kim2022comma,
  author = {Kim, Sanghoon and Ko, In-Young},
  title = {A Conversational Approach for Modifying Service Mashups in IoT Environments},
  year = {2022},
  publisher = {Association for Computing Machinery},
  address = {New York, NY, USA},
  url = {https://doi.org/10.1145/3491102.3517655},
  doi = {10.1145/3491102.3517655},
  booktitle = {Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems},
  pages = {1–16},
  series = {CHI '22}
}