Files
calendar/apps/api/v1
Hariom BalharaandGitHub e3bd90c4f2 fix: Handle calendar-cache with Delegation Credentials
Fixes CAL-5372

# Delegation Credentials with CalendarCache.

Following content is a snapshot of the [internal document](https://calendso.slack.com/docs/T08B8KA2BNF/F08L5JYU3V3)



**Problem-1 :** 

CalendarCache needs SelectedCalendar records to work but SelectedCalendar record is only created when a user connects their calendar and then enables some calendar for conflict checking. Because with Delegation, no manual connection is done by any of the members, we need a way to create SelectedCalendar records automatically.

**Problem-2**

CalendarCache connects to credential(regular credential) which doesn’t exist for Delegation Credential scenario. Also, DelegationCredential is common for all the members(different from Credential which is different for different members) of the organization and we need to identify to which user the CalendarCache belongs.  

**Solution for both problems**
- Create credential records for Delegation Credentials as well - Through Cron(new - we could schedule it every 5mins)
- Now create SelectedCalendar  records for those Credential records -  Through another Cron(new - we could schedule it every 5mins)
- Now CalendarCache records will automatically be created for those SelectedCalendar records -existing cron

## Fixed some Delegation Credentials bugs unrelated to calendar-cache
- If DestinationCalendar wasn't set(which is possible only with Delegation Credentials), then Google Meet wasn't used as a conferencing app - [Added a test]
- If no SelectedCalendar is there but Google Calendar connection exists(possible only with Delegation Credential) then we were not doing conflict checking. It is expected to not do it for Regular Credentials, but for Delegation Credential we must check for conflict in that case too [Added a test]
- Earlier if a user has Regular Credential as well as Delegation Credential for the same external id which is the member email(say member1@acme.com) then availability were retrieved twice because we weren't deduplicating credentials as it wasn't a trivial thing to do. Now that is being done.


**Env Variables:**
Note this PR doesn't introduce any new env variable. The existing env variable has been added to .env.example. But if this env variable isn't already set, it must be set.

`CALCOM_SERVICE_ACCOUNT_ENCRYPTION_KEY={SAME_AS_SET_FOR_V2_API}`

**Deployment Plan:**
1. Add Observability for SelectedCalendar when _error_ field is set
2. Follow https://github.com/calcom/cal.com/blob/calendar-cache-dwd-support/apps/web/app/(use-page-wrapper)/settings/(settings-layout)/organizations/delegation-credential/delegation-credential.md#setting-up-delegation-credential-for-google-calendar-api to enable Delegation Credential for i.cal.com
3. Note that to be able to see the option to enable Delegation Credential for an organization, you need to enable `teamFeature` and `feature` for `delegation-credential`

## Automation Tests
- Introduced tests for calendar-cache.repository.ts
   - Tests all methods of the repository
- Added more tests for handleNewBooking/delegation-credential flow.
   - Added test to verify the bug fix when no DestinationCalendar exists and Google Meet should be used still 
- Added more tests for Google Calendar/CalendarService targeting DelegationCredential
- Added more tests for getCalendarsEvents. 
    - To test the new logic of calling getAvailability still if there are no selectedCalendars in case of Delegation Credential
    - Also introduced tests for `getAvailabitlityWithTimezones` which was an existing function but now has some new changes.
- Added tests for deduplication logic in CalendarManager.ts

## How to Test
Enable Calendar Cache and Delegation Credential feature for acme org through `features` and `teamFeatures` tables.
- Enable Delegation Credential for acme org
- Enable atleast 1 calendar for conflict checking for one of the users(say owner1)
- Ensure GOOGLE_WEBHOOK_TOKEN is set in .env file
- Ensure GOOGLE_WEBHOOK_URL is set to ngrok url of webapp in .env file
- Hit cron endpoint `curl http://localhost:3000/api/calendar-cache/cron\?apiKey\={API_KEY}` that would cache the freebusy result for the selected calendars



Followup 
- https://github.com/calcom/cal.com/pull/20698
- https://github.com/calcom/cal.com/pull/18619/files#r2046795643
2025-04-28 18:11:29 -03:00
..
2025-02-24 14:55:43 +00:00
2025-03-05 14:29:11 +00:00
2024-12-03 18:43:01 +01:00
2025-03-05 14:29:11 +00:00

Commercial Cal.com Public API

Welcome to the Public API ("/apps/api") of the Cal.com.

This is the public REST api for cal.com. It exposes CRUD Endpoints of all our most important resources. And it makes it easy for anyone to integrate with Cal.com at the application programming level.

Stack

  • NextJS
  • TypeScript
  • Prisma

Development

Setup

  1. Clone the main repo (NOT THIS ONE)

    git clone --recurse-submodules -j8 https://github.com/calcom/cal.com.git
    
  2. Go to the project folder

    cd cal.com
    
  3. Copy apps/api/.env.example to apps/api/.env

    cp apps/api/.env.example apps/api/.env
    cp .env.example .env
    
  4. Install packages with yarn

    yarn
    
  5. Start developing

    yarn workspace @calcom/api dev
    
  6. Open http://localhost:3002 with your browser to see the result.

API Authentication (API Keys)

The API requires a valid apiKey query param to be passed: You can generate them at https://app.cal.com/settings/security

For example:

GET https://api.cal.com/v1/users?apiKey={INSERT_YOUR_CAL.COM_API_KEY_HERE}

API Keys optionally may have expiry dates, if they are expired they won't work. If you create an apiKey without a userId relation, it won't work either for now as it relies on it to establish the current authenticated user.

In the future we might add support for header Bearer Auth if we need to or if our customers require it.

Middlewares

We don't use the new NextJS 12 Beta Middlewares, mainly because they run on the edge, and are not able to call prisma from api endpoints. We use instead a very nifty library called next-api-middleware that let's us use a similar approach building our own middlewares and applying them as we see fit.

  • withMiddleware() requires some default middlewares (verifyApiKey, etc...)

Next.config.js

Redirects

Since this is an API only project, we don't want to have to type /api/ in all the routes, and so redirect all traffic to api, so a call to api.cal.com/v1 will resolve to api.cal.com/api/v1

Likewise, v1 is added as param query called version to final /api call so we don't duplicate endpoints in the future for versioning if needed.

Transpiling locally shared monorepo modules

We're calling several packages from monorepo, this need to be transpiled before building since are not available as regular npm packages. That's what withTM does.

  "@calcom/app-store",
  "@calcom/prisma",
  "@calcom/lib",
  "@calcom/features",

API Endpoint Validation

We validate that only the supported methods are accepted at each endpoint, so in

  • /endpoint: you can only [GET] (all) and [POST] (create new)
  • /endpoint/id: you can read create and edit [GET, PATCH, DELETE]

Zod Validations

The API uses zod library like our main web repo. It validates that either GET query parameters or POST body content's are valid and up to our spec. It gives errors when parsing result's with schemas and failing validation.

We use it in several ways, but mainly, we first import the auto-generated schema from @calcom/prisma for each model, which lives in lib/validations/

We have some shared validations which several resources require, like baseApiParams which parses apiKey in all requests, or querIdAsString or TransformParseInt which deal with the id's coming from req.query.

  • [*]BaseBodyParams that omits any values from the model that are too sensitive or we don't want to pick when creating a new resource like id, userId, etc.. (those are gotten from context or elsewhere)

  • [*]Public that also omits any values that we don't want to expose when returning the model as a response, which we parse against before returning all resources.

  • [*]BodyParams which merges both [*]BaseBodyParams.merge([*]RequiredParams);

Next Validations

Next-Validations Docs Next-Validations Repo We also use this useful helper library that let us wrap our endpoints in a validate HOC that checks the req against our validation schema built out with zod for either query and / or body's requests.

Testing with Jest + node-mocks-http

We aim to provide a fully tested API for our peace of mind, this is accomplished by using jest + node-mocks-http

Endpoints matrix

resource get [id] get all create edit delete
attendees
availabilities
booking-references
event-references
destination-calendars
custom-inputs
event-types
memberships
payments
schedules
selected-calendars
teams
users 👤[1]

Models from database that are not exposed

mostly because they're deemed too sensitive can be revisited if needed. Also they are expected to be used via cal's webapp.

  • Api Keys
  • Credentials
  • Webhooks
  • ResetPasswordRequest
  • VerificationToken
  • ReminderMail

Documentation (OpenAPI)

You will see that each endpoint has a comment at the top with the annotation @swagger with the documentation of the endpoint, please update it if you change the code! This is what auto-generates the OpenAPI spec by collecting the YAML in each endpoint and parsing it in /docs alongside the json-schema (auto-generated from prisma package, not added to code but manually for now, need to fix later)

@calcom/apps/swagger

The documentation of the API lives inside the code, and it's auto-generated, the only endpoints that return without a valid apiKey are the homepage, with a JSON message redirecting you to the docs. and the /docs endpoint, which returns the OpenAPI 3.0 JSON Spec. Which SwaggerUi then consumes and generates the docs on.

Deployment

scripts/vercel-deploy.sh The API is deployed to vercel.com, it uses a similar deployment script to website or webapp, and requires transpilation of several shared packages that are part of our turborepo ["app-store", "prisma", "lib", "ee"] in order to build and deploy properly.

Envirorment variables

Required

DATABASE_URL=DATABASE_URL="postgresql://postgres:@localhost:5450/calendso"

Optional

APIKEY_PREFIX=cal# This can be changed per envirorment so caltest for staging for example.

If you're self-hosting under our commercial license, you can use any prefix you want for api keys. either leave the default cal_ (not providing any envirorment variable) or modify it

Ensure that while testing swagger, API project should be run in production mode We make sure of this by not using next in dev, but next build && next start, if you want hot module reloading and such when developing, please use yarn run next directly on apps/api.

See https://github.com/vercel/next.js/blob/canary/packages/next/server/dev/hot-reloader.ts#L79. Here in dev mode OPTIONS method is hardcoded to return only GET and OPTIONS as allowed method. Running in Production mode would cause this file to be not used. This is hot-reloading logic only. To remove this limitation, we need to ensure that on local endpoints are requested by swagger at /api/v1 and not /v1

How to deploy

We recommend deploying API in vercel.

There's some settings that you'll need to setup.

Under Vercel > Your API Project > Settings

In General > Build & Development Settings BUILD COMMAND: yarn turbo run build --scope=@calcom/api --include-dependencies --no-deps OUTPUT DIRECTORY: apps/api/.next

In Git > Ignored Build Step

Add this command: ./scripts/vercel-deploy.sh

See scripts/vercel-deploy.sh for more info on how the deployment is done.

IMPORTANT: If you're forking the API repo you will need to update the URLs in both the main repo .gitmodules and this repo ./scripts/vercel-deploy.sh

Environment variables

Lastly API requires an env var for DATABASE_URL and CALCOM_LICENSE_KEY