How Navigator's Proxy Works¶
UF IT operates a LiteLLM proxy at api.ai.it.ufl.edu that gives AIBHS projects a unified way to call Claude and other models. This is what the AI Models Guidelines refers to as the Navigator Toolkit endpoint.
Two API Surfaces on the Same Proxy¶
- OpenAI-compatible endpoint at
/v1/chat/completions— what most tools speak: Cline, Aider, Cursor, Continue, Zed, etc. - Anthropic pass-through endpoint at the proxy root — what Claude Code speaks natively,
configured via
ANTHROPIC_BASE_URL.
Both are backed by the same LiteLLM router. Requests eventually hit AWS Bedrock, which serves the Claude models. From the client's perspective this is invisible — the proxy translates as needed.
Authentication¶
Every request needs a bearer token in the Authorization header. Navigator keys are self-service
through UF IT's dashboard — see Getting a Navigator API Key. Store the
key in a project-local .env file (see The .env Pattern).
Which Models Are Available to Your Key¶
The model catalog available to an AIBHS key can change as UF IT updates the Navigator configuration. Do not assume a specific model ID — always discover the current one for your key with:
The response is a JSON list of every model your key is allowed to call. Copy the id field of the
model you want — that string is what every tool integration needs. Throughout this guide, that
string is referred to as "your model ID."
Thinking-mode models have a hard constraint. If your model ID contains -thinking or
otherwise indicates extended thinking / reasoning mode, the model requires temperature to be
exactly 1. Any other temperature value is rejected with a BedrockException. Most coding tools
default to temperature 0 or 0.2 for deterministic output — with a thinking-mode model, you
must override that default to 1 or every request fails. If your model ID does NOT indicate
thinking mode, tools work with their default temperature settings and no override is needed. Check
this before configuring any tool.