Route by policy, not model names
Balanced policy chose quality within the request cap, then wrote the decision into the trace.
Default for every key unless an exception matches.
Selected for mixed production traffic.
Exceptions override the default only when conditions match.
Preview shows the route before teams save policy changes.
Tune the policy once, then keep application code stable.
Admins set the team default, add exceptions only when needed, and preview the route before saving.
Default for every key unless an exception matches.
Selected for mixed production traffic.
Exceptions override the default only when conditions match.
A support request shows how policy replaces model names.
The app sends one stable request. LineMod evaluates policy, chooses a model, prepares fallback, and records the decision.
{
"model": "auto:balanced",
"metadata": {
"workload": "support",
"budget_cap": "$0.020"
}
}Team default applies because no exception matched.
Sonnet fits the quality target and budget cap.
Backup chain is prepared but not used.
Application code stays stable while routing policy changes.
Teams can tune model choice in the console without redeploying product code.
client = OpenAI({
baseURL: "https://api.linemod.ai/v1",
apiKey: process.env.LINEMOD_API_KEY
})
client.chat.completions.create({
model: "auto:balanced",
messages,
metadata: { policy: "team_default" }
})Routing changes live in LineMod policy, while developers keep the same SDK workflow.
Put routing policy between apps and models
Start with one API route, then tune cost, quality, and fallback from the gateway.