Open APIs, integration patterns, and the model gateway that lets you choose, swap or self-host every component. No lock-in.
The model gateway sits between your Dropsolid AI use cases and the underlying AI infrastructure. Instead of wiring each use case directly to one vendor's model, it routes through a single configurable layer:
Use case layer: Knowledge Navigator, Trusted Answers, and every other Dropsolid AI capability
Model gateway: Routes requests, handles auth, and abstracts vendor-specific APIs
Pluggable backends: LLMs, embedding models, and vector databases, selected per deployment
Swap a component at the infrastructure layer and every use case built on top keeps working, no re-architecture, and no vendor migration project.
Choice instead of lock-in
| Closed AI platform | Dropsolid open DXP |
| One vendor's model, fixed | Any LLM, embedding model, or vector DB |
| Switching model = rebuilding use cases | Switching model = reconfiguring the gateway |
| Data residency dictated by vendor | Self-hosting options for compliance needs |
| Roadmap set by one provider | Best-of-breed stack, chosen per deployment |
Connect & extend is the architectural foundation under every use case. The model gateway is what lets every deployment pick its own LLM, embedding model, and vector database without rebuilding the use case.
In practice:
Swap models without downtime: Move from one LLM provider to another as pricing, performance, or compliance needs change, without touching the use case logic built on top
Meet data residency requirements: Select a self-hosted or EEA-specific vector database to keep sensitive data where it needs to stay
Standardize on your existing stack: Connect the CRM, DAM, or commerce platform already in place instead of migrating to platform-native equivalents
Future-proof against AI model changes: As new, better, or cheaper models are released, adopt them through the gateway rather than a re-platforming project
