“Vibe-coding” using systems like Claude Code or OpenAI Codex can be very effective, especially when combining clearly defined interfaces like PRC’s gRPC based environment with environments for that a lot of references exist, e.g. a GUI built on top of Python/Tkinter.
The challenge is clearly defining how data should be structured so that it can be processed by the server, in this case the PRC.Server.
We have therefore expanded the comments on the prc.proto file (PRC.Integrations/Protos at main · jbraumann/PRC.Integrations) and created a separate LLM integration guide (PRC.Integrations/PRC_LLM_Integration_Guide.md at main · jbraumann/PRC.Integrations) which was itself generated by an LLM with the goal of collecting all relevant information in order to successfully integrate PRC into different environments. Being LLM-generated and updated, it may contain inaccuracies.
In our work we have been successful in providing the prc.protos file alongside the the LLM Integration Guide as context, plus a prompt defining the app/plugin that you want to build.
We are happy to hear about any suggestions on how to improve the context files!