Currently, SwissPipe only supports one LLM provider. This provider needs to be on the cloud, making self-hosting limited.
Instead of a custom node, it would benefit from having a generic AI agent node similar to n8n.
One option is using rig and wrapping it with whatever fields/response is needed by swisspipe. Using rig as a dependency will give you access to +20 model providers (including self-hosted and all major AI providers) and +10 vector databases.
https://github.com/0xPlaygrounds/rig
Providers: https://github.com/0xPlaygrounds/rig/tree/main/rig-core/src/providers