This is a customized fork of the open-source repository attendee by Noah Duncan, optimized to work with our specific architecture.

The Meeting Recorder is a Django application designed to run as a single Docker image, supported by Postgres and Redis.

Core Architecture:

1. Main API: Handles authentication, account management, and bot orchestration.
2. Bots: The API spins up recorder pods (workers) called "Bots". These bots:
   • Join a meeting (Zoom, Teams, etc.).
   • Record the audio/video.
   • Upload the raw data to Object Storage (OpenStack Swift).
   • Leave the meeting and destroy themselves.
   • Trigger a callback to the transcript-gateway.

This fork deviates from the original attendee repository in several key ways:

• Custom Storage: We utilize OpenStack Swift (via Infomaniak) instead of AWS S3.
  • Location: /bots/storage/infomaniak_storage.py
• Callbacks: Post-recording, the bot triggers a callback to our transcript-gateway.
  • Location: /transcript_services/v1/api_service.py
• Kubernetes Native: We have removed non-essential bloat and added more logs to support running inside a K8s cluster.

Best for quick logic changes and database testing.

1. Build the image: docker compose -f dev.docker-compose.yml build
2. Generate Environment Variables: docker compose -f dev.docker-compose.yml run --rm recorder-api python init_env.py > .env (Edit the generated .env file to add your AWS/Storage credentials)
3. Start Services: docker compose -f dev.docker-compose.yml up
4. Run Migrations: docker compose -f dev.docker-compose.yml exec recorder-api python manage.py migrate
5. Create Account:
   • Go to localhost:8001 to sign up.
   • Find the confirmation link in your terminal logs (e.g., http://localhost:8001/accounts/confirm-email/<key>/).
   • Login to obtain your API Key.

Best for testing integration with the wider cluster architecture.

Prerequisites:

• skaffold (brew install skaffold)
• Docker Desktop (Kubernetes enabled)
• Clone of the transcript-gitops repository (for manifests).

Configuration:

1. Update skaffold.yaml to point to your local path for the manifests:

   manifests:
       rawYaml:
       - /Users/USERNAME/PATH_TO/transcript-k8s/deployment-meeting-recorder-api-dev.yaml

2. Secrets Management:
   • Copy examples: cp ./secret-examples/secret-docker.yaml secret-docker.yaml
   • Apply: kubectl apply -f secret-docker.yaml (and repeat for secret-transcript.yaml)
3. Apply Configs & Deps:
   • kubectl apply -f configmap-transcript.yaml
   • kubectl apply -f deployment-redis.yaml
   • kubectl apply -f deployment-postgres.yaml
   • kubectl apply -f role-meeting-recorder-bot.yaml
4. Run: skaffold dev

• Create Migrations: docker compose -f dev.docker-compose.yml exec recorder-api python manage.py makemigrations bots
• Apply Migrations: docker compose -f dev.docker-compose.yml exec recorder-api python manage.py migrate

• Build the Docker image (Takes about 5 minutes)

docker build --platform=linux/amd64 -t azytaku/transcript-meeting-recorder:1.0.33 -f Dockerfile.dev .

• Push the image to Docker Hub

docker push azytaku/transcript-meeting-recorder:1.0.33

Join a meeting with a POST request to /bots:

curl -X POST http://localhost:8000/api/v1/bots \
-H 'Authorization: Token <YOUR_API_KEY>' \
-H 'Content-Type: application/json' \
-d '{"meeting_url": "https://us05web.zoom.us/j/84315220467?pwd=9M1SQg2Pu2l0cB078uz6AHeWelSK19.1", "bot_name": "My Bot"}'

Response: {"id":"bot_3hfP0PXEsNinIZmh","meeting_url":"https://us05web.zoom.us/j/4849920355?pwd=aTBpNz760UTEBwUT2mQFtdXbl3SS3i.1","state":"joining","transcription_state":"not_started"}

The API will respond with an object that represents your bot's state in the meeting.

Make a GET request to /bots/<id> to poll the bot:

curl -X GET http://localhost:8000/api/v1/bots/bot_3hfP0PXEsNinIZmh \
-H 'Authorization: Token <YOUR_API_KEY>' \
-H 'Content-Type: application/json'

Response: {"id":"bot_3hfP0PXEsNinIZmh","meeting_url":"https://us05web.zoom.us/j/88669088234?pwd=AheaMumvS4qxh6UuDtSOYTpnQ1ZbAS.1","state":"ended","transcription_state":"complete"}

When the endpoint returns a state of ended, it means the meeting has ended. When the transcription_state is complete it means the meeting recording has been transcribed.

Once the meeting has ended and the transcript is ready make a GET request to /bots/<id>/transcript to retrieve the meeting transcripts:

curl -X GET http://localhost:8000/api/v1/bots/bot_3hfP0PXEsNinIZmh/transcript \
-H 'Authorization: Token mpc67dedUlzEDXfNGZKyC30t6cA11TYh' \
-H 'Content-Type: application/json'

Response:

[{
"speaker_name":"Alan Turing",
"speaker_uuid":"16778240","speaker_user_uuid":"AAB6E21A-6B36-EA95-58EC-5AF42CD48AF8",
"timestamp_ms":1079,"duration_ms":7710,
"transcription":"You can totally record this, buddy. You can totally record this. Go for it, man."
},...]

You can also query this endpoint while the meeting is happening to retrieve partial transcripts.