ChatGPT Integration

[tothambrus11]

Users should be able to configure the bot with an OpenAI api key so that chatgpt can read the post's description, read the person's introduction, and write a personal message that convinces the landlord / roommates of choosing them.

The configuration should contain these optional config fields: OPENAI_API_KEY, USE_AI_GENERATED_MESSAGE.
The context that is provided to chatgpt in the first system message should be read from the same message file that we currently use for sending the hard-coded message.

Acceptance criteria:

• There should be helpful error reporting, e.g. in case the api key is missing while USE_AI_GENERATED_MESSAGE is set to true.
• The project should use the https://www.npmjs.com/package/chatgpt library and the official way of accessing chatgpt.
• The program should automatically write messages for us with no human interaction.
• We should limit the size of the context that we give to the AI, it can be an optional config field MAX_CONTEXT_WORD_COUNT.

Notes:

• We should research prompt-engineering methods to create the best possible prompt for the AI, so that each message would be as convincing / friendly as possible.
• We should also research what are the best factors of an advertisement reply, so that we can give these as tips to the AI. These factors should also be added to the README.md, so that users of the program don't forget to give enough context for the AI.
• We might need to set some limits for each of our message if it turns out to be too long (MAX_GENERATED_WORD_COUNT). In this case, we might need to ask chatgpt in a follow-up question to shorten the message.
• We might need to break down this issue into multiple smaller ones.