## AWS

- DNS?
- RDNS (main pg)
- RDNS (keycloak pg)
- Elasticache (redis)
- Qdrant (ec2 with external volume)
How easy would it be to scale this?
- Managed s3 (for server)
- Keycloak (ec2)
- Embedding server
- s3 (for chat)
- s3 (for search)
- s3 (for models)
- s3 (for server)
- secrets manager?
- Tika (ec2)

flowchart TD
    DNS --> C
    DNS --> D
    DNS --> server
    server --> B[s3-files]
    C[s3-chat] --> server
    D[s3-search] --> server
    server --> embedding-server
    server --> G[Tika]
    server --> qdrant-ec2
    qdrant-ec2 --> H[volume mount]
    server --> elastic-cache
    server --> keycloak
    keycloak --> I[RDS-keycloak]
    server --> RDS-main

note: s3 should all be the same, it looks easier on the graph if s3 is denoted for its usecase

### Shrinking rust server image
- [ ] remove imagemagick (feature flag it out)
- [ ] microservice cross encoder (hopefully not a huge perf hit)