A browser URL and an API URL are different.
A tool’s web interface opens in your browser. An API base URL belongs in a client configuration. Saving one as a shortcut does not configure the other. Record which type of address a tool expects before copying it between applications.
Ollama documents http://localhost:11434/api for its native local API and http://localhost:11434/v1 for its OpenAI-compatible local API. These refer to the machine running your browser or client—not a server owned by RMA.
Match the running server.
For llama.cpp, inspect the address and port printed by your running llama-server and use its server documentation for the supported route. A port number in someone else’s setup is not proof that your own server is listening there.
Check the client’s required protocol and exact base-URL format. Adding /v1 twice, dropping it when it is required, or sending a native API payload to a compatibility route can make a healthy server look broken. Do not treat an OpenAI-compatible endpoint as support for every OpenAI feature.
Save the connection facts.
Keep the runtime version, model identifier, API family, base URL, and client version next to your working configuration. Store a browser URL as a tool shortcut if you want one-click access to its interface. Keep credentials in the client’s credential mechanism, outside setup text.
# Local connection
Runtime and version:
Model identifier:
Client and version:
API family:
Base URL:
Browser URL, if available:
Verification request:
Observed response:
Last checked:
Local and hosted are separate choices.
A loopback address such as localhost points to the current device. A hosted endpoint sends requests to another operator. Moving between them can change authentication, data handling, limits, and costs. Make that choice explicitly.
RMA Connect supports local connections and optional remote provider routing. The workspace itself stores projects, setups, and tool links; it does not turn a saved URL into a managed inference service.
Sources
- Ollama: native and compatible base URLs ↗
- llama.cpp: server routes and configuration ↗
- RMA Connect and current routing boundaries ↗
Product behavior and software documentation were checked on the review date. Upstream versions can change.