Five moves to a first decision
Get Started / Path
1
Install
Install the packaged runtime from GitHub Releases so role-model-router is on your PATH.
2
Connect models
Start runtime and connect the local or remote endpoints that will form your real candidate set.
3
Full benchmark
Run a full benchmark so quality, latency, and cost are measured and inform routing decisions.
4
Choose strategy
Pick Balanced, Quality, Latency, or Cost only after the benchmark has evidence.
5
Observe
View routed requests, cache hit rates and more under the Observe page.
First-time setup architecture
Establish the real candidate set first, then benchmark, then choose strategy from evidence.
01 · Install & launch
Start the packaged runtime
Launch the router and open the operator UI on the local machine.
02 · Connect endpoints
Wire local and remote paths
Connect the exact providers and backends you plan to route across.
03 · Activate & assign
Publish the real inventory
Activate models and bind each endpoint only to the roles it should serve.
04 · Full benchmark
Grade the candidate set
Write measured quality into observed profiles for the endpoints you configured.
05 · Choose strategy
Save after the evidence
Set balanced, quality, latency, or cost once benchmark results exist.
06 · Validate request
Inspect a real route
Confirm winner, fallbacks, and reasons in Router and Observe.
macOS & Linux
install.sh
# Packaged release · recommendedcurl -fsSL https://raw.githubusercontent.com/try-works/role-model/main/scripts/install.sh | sh
Installs under ~/.local/share and puts role-model-router on your PATH. Open a new terminal if the command is missing.
Windows
install.ps1
# PowerShell · packaged releaseirm https://raw.githubusercontent.com/try-works/role-model/main/scripts/install.ps1 | iex
Installs under %LOCALAPPDATA%\Programs\RoleModelRouter. Prefer the packaged release over a source build unless you are contributing.
Pi extension improves routing with request metadata
The package discovers an already-running role-model — it never installs, starts, or owns the process.
01
Runtime first
Install and launch the role-model router before adding the Pi package.
02
Set up pi-role-model
Discovers the runtime, registers the role-model provider, and exposes setup / doctor / alias commands.
03
Router stays authority
Pi sends enriched requests to the downstream OpenAI endpoint; role-model still picks the concrete endpoint.
Install Pi extension
Install the package, then setup and choose an alias.
installpi install npm:@try-works/pi-role-model
/role-model setup
/role-model doctor
/role-model alias choose
Default endpoint is http://127.0.0.1:3456. Remote runtime access stays blocked unless you explicitly allow it.
Inspect the decision
A RouterDecision you can read — winner, fallbacks, exclusions, and reasons.
RouterDecision{ "chosen": "kimi-k3", "strategy": "balanced", "fallbacks": ["gpt-5.6-sol", "opus-5"], "reason": "highest measured quality for role-coder" }
Selection signals
domain · Engineering
role · coder
task · implement
scope · repo
winner · Kimi K3