diagrams/resilience-3layers.mmd

Scope: entire provider (e.g., , , ).

Purpose: stop sending traffic to a provider that is repeatedly failing at the upstream/service level.

Implementation:

• ,

States:

— normal traffic allowed
• — provider temporarily blocked; combo routing skips it
• — reset timeout elapsed; probe request allowed

Defaults ():

Trip codes: only provider-level statuses . Do NOT trip for account-level errors (most 401/403/429 — those belong to cooldown or lockout).

Lazy recovery: when expires, , , refresh state to . No background timer needed.

---

Scope: single provider connection/account/key.

Purpose: skip one bad key while other connections for the same provider keep serving.

Implementation:

• in same file

Fields per connection:

— timestamp until cooldown expires
• , ,
• — exponential backoff counter

Default cooldowns:

• /reset headers/parseable reset text

Anti-thundering-herd guard: prevents concurrent failures from over-extending cooldown or double-incrementing .

Terminal states (NOT cooldowns):

Lazy recovery: when is past, connection becomes eligible again. On successful use, clears all error fields.

---

Scope: provider + connection + model triple.

Purpose: avoid disabling a whole connection when only one model is unavailable or quota-limited.

Examples:

Implementation: — , , .

)

REST API:

— list active lockouts
• — manual re-enable. Body: . Auth: management.

---

14 routing strategies (priority, weighted, round-robin, context-relay, fill-first, p2c, random, least-used, cost-optimized, reset-aware, strict-random, auto, lkgp, context-optimized) — see AUTO-COMBO.md.
• Reset-aware routing (v3.8.0) — prioritizes connections by quota reset time.
• Background mode degradation — Responses API degraded to sync with warning.
• Dynamic tool limit detection — backs off providers when tool count limits hit.

---

/.
• instead of /.

---

STEALTH_GUIDE.md.

---

Architecture Guide — System architecture and internals
• User Guide — Providers, combos, CLI integration
• Auto-Combo Engine — 6-factor scoring, mode packs