# Tier × lane orchestration metadata (model IDs → config/providers.yaml)

lanes:
  A: economy
  B: balanced
  C: max

# LiteLLM per-lane RPM (proxy-side). Zed agent bursts tool calls — keep high for personal IFT.
# OpenRouter :free upstream may still 429; router fallbacks + cooldown_time apply separately.
models:
  a-simple:
    fallbacks: [b-simple, groq-llama-8b]
    rpm: 300
  b-simple:
    fallbacks: [a-simple, groq-llama-8b]
    rpm: 300
  c-simple:
    fallbacks: [b-simple]
    rpm: 300

  a-medium-ops:
    fallbacks: [a-medium-code, b-medium-ops]
    rpm: 240
  b-medium-ops:
    fallbacks: [a-medium-code, c-medium-ops]
    rpm: 240
  c-medium-ops:
    fallbacks: [b-medium-ops]
    rpm: 240

  a-medium-code:
    fallbacks: [a-medium-ops, groq-qwen-coder]
    rpm: 240
  b-medium-code:
    fallbacks: [b-medium-ops, c-medium-code]
    rpm: 240
  c-medium-code:
    fallbacks: [b-medium-code, c-medium-ops]
    rpm: 180

  a-complex:
    fallbacks: [b-complex, a-reasoning]
    rpm: 180
  b-complex:
    fallbacks: [a-complex, c-complex]
    rpm: 120
  c-complex:
    fallbacks: [b-complex]
    rpm: 120

  a-reasoning:
    fallbacks: [b-reasoning, gemini-flash]
    rpm: 120
  b-reasoning:
    fallbacks: [a-reasoning, c-reasoning]
    rpm: 120
  c-reasoning:
    fallbacks: [b-reasoning, gemini-flash]
    rpm: 120

  a-vision-ocr:
    fallbacks: [a-vision]
    rpm: 120
    no_escalation: true
  a-vision:
    fallbacks: [a-vision-ocr]
    rpm: 120
  b-vision:
    fallbacks: [a-vision, c-vision]
    rpm: 90
  c-vision:
    fallbacks: [b-vision]
    rpm: 90

audit:
  price_drift_threshold_pct: 10
  cheaper_alternative_pct: 15