LLM calls with budgets, caching, and fallbacks.
FlyAI gives you a single fx:// interface for the LLM providers you already use, with multi-provider fallback, semantic caching, a weekly free quota, and metered wallet billing at cost plus a transparent markup. Six official functions: ai-auto, ai-fast, ai-code, ai-reason, ai-vision, ai-embed.
What you get
Inference with the boring production primitives already done: caching, fallbacks, quota, billing, prompt-injection guard, BYOK.
Multi-provider fallback
A ladder of provider rungs lights up one after another when a higher rung fails. The router picks the best rung per request from real outcome signals.
Semantic cache
An exact-match cache plus a semantic cache. Cache hits claim a fraction of the quota, so effective throughput is much higher than the plan number. Threshold auto-tunes to a target false-positive rate.
Weekly free tier
A generous weekly allowance per plan. Daily cap = one-third of the weekly budget. Cache hits consume fewer units than fresh requests.
BYOK via Vault
Bring your own OpenAI / Groq / DeepSeek / Together / OpenRouter / Fireworks keys. Stored in Vault AI connections. BYOK bypasses platform quota but enforces the preset allowlist, per-connection RPM, and daily/monthly spend caps.
TokenFabric (delta sync)
Optional: re-send only the files that changed since the last call. Unchanged files are summarized by hash. Saves most tokens on codebase-aware prompts. Injection-guarded before cache write.
Substrate receipts
Every FlyAI call becomes a typed, replayable, receipt-bearing execution record with a per-request budget and circuit-broken verification.
Production-launch readiness
What's in general availability today, what's running in shadow mode, and what's queued for staging.
Stop hand-rolling LLM routing
Six fx:// functions, multi-provider fallback, semantic cache, free weekly quota, metered billing — bundled into every plan.