The core logic is contained in [agent/route.ts](/apps/ai/src/app/api/agent/route.ts). Here, a [LangChain Agent Executor](https://docs.langchain.com/docs/components/agents/agent-executor) is tasked with following your instructions. Given your last-known timezone, working hours, and busy times, it attempts to CRUD your bookings.
_The AI agent can only choose from a set of tools, without ever seeing your API key._
Emails are cleaned and routed in [receive/route.ts](/apps/ai/src/app/api/receive/route.ts) using [MailParser](https://nodemailer.com/extras/mailparser/).
Incoming emails are routed by email address. Addresses are verified by [DKIM record](https://support.google.com/a/answer/174124?hl=en), making them hard to spoof.
Before running the app, please see [env.mjs](./src/env.mjs) for all required environment variables. Run `cp .env.example .env` in this folder to get started. You'll need:
The scheduling agent in [agent/route.ts](/apps/ai/src/app/api/agent/route.ts) calls an LLM (in this case, GPT-4) in a loop to accomplish a multi-step task. We use an [OpenAI Functions agent](https://js.langchain.com/docs/modules/agents/agent_types/openai_functions_agent), which is fine-tuned to output text suited for passing to tools.
Tools (eg. [`createBooking`](/apps/ai/src/tools/createBooking.ts)) are simply JavaScript methods wrapped by Zod schemas, telling the agent what format to output.
To expose the AI app, you can use either [Tunnelmole](https://github.com/robbie-cahill/tunnelmole-client), an open source tunnelling tool; or [nGrok](https://ngrok.com/), a popular closed source tunnelling tool.
For Tunnelmole, run `tmole 3005` (or the AI app's port number) in a new terminal. Please replace `3005` with the port number if it is different. In the output, you'll see two URLs, one http and a https (we recommend using the https url for privacy and security). To install Tunnelmole, use `curl -O https://install.tunnelmole.com/8dPBw/install && sudo bash install`. (On Windows, download [tmole.exe](https://tunnelmole.com/downloads/tmole.exe))
For nGrok, run `ngrok http 3005` (or the AI app's port number) in a new terminal. You may need to install nGrok first.
To forward incoming emails to the serverless function at `/agent`, we use [SendGrid's Inbound Parse](https://docs.sendgrid.com/for-developers/parsing-email/setting-up-the-inbound-parse-webhook).
1. Ensure you have a [SendGrid account](https://signup.sendgrid.com/)
2. Ensure you have an authenticated domain. Go to Settings > Sender Authentication > Authenticate. For DNS host, select `I'm not sure`. Click Next and add your domain, eg. `example.com`. Choose Manual Setup. You'll be given three CNAME records to add to your DNS settings, eg. in [Vercel Domains](https://vercel.com/dashboard/domains). After adding those records, click Verify. To troubleshoot, see the [full instructions](https://docs.sendgrid.com/ui/account-and-settings/how-to-set-up-domain-authentication).
3. Authorize your domain for email with MX records: one with name `[your domain].com` and value `mx.sendgrid.net.`, and another with name `bounces.[your domain].com` and value `feedback-smtp.us-east-1.amazonses.com`. Set the priority to `10` if prompted.
5. In the Destination URL field, use the Tunnelmole or ngrok URL from above along with the path, `/api/receive`, and one param, `parseKey`, which lives in [this app's .env](/apps/ai/.env.example) under `PARSE_KEY`. The full URL should look like `https://abc.tunnelmole.net/api/receive?parseKey=ABC-123` or `https://abc.ngrok.io/api/receive?parseKey=ABC-123`.
