Skip to content
Cette page a été générée et traduite avec l'aide de l'IA. Si vous remarquez des inexactitudes, n'hésitez pas à contribuer. Modifier sur GitHub

Automix Router ​

The Automix router optimizes for cost by starting every query avec un cheap model and only escalating vers un premium model lorsque le initial response's confidence is below a threshold.

Fonctionnement ​

  1. Initial query -- send the query vers le cheap model
  2. Confidence check -- evaluate la reponse confidence score
  3. Escalate if needed -- if confidence is below threshold, re-query with premium model
  4. Retour -- retour the first confident response

Confidence Scoring ​

Confidence is assessed based on:

  • Self-reported confidence in la reponse
  • Presence of hedging language ("I'm not sure", "pourrait etre")
  • Token-level entropy of la reponse
  • Tool call success rate

Configuration ​

toml
[router]
strategy = "automix"

[router.automix]
enabled = true
confidence_threshold = 0.7
cheap_model = "anthropic/claude-haiku"
premium_model = "anthropic/claude-opus-4-6"
max_escalations = 1

Cost Savings ​

In typical usage, Automix route 60-80% of interroge vers le cheap model, achieving significant cost savings tandis que maintaining quality for complex queries.

Voir aussi Pages ​

Released under the Apache-2.0 License.