CONDUITWe cut LLM costs by 95.9% without touching your code.
Stop overpaying for every request. Conduit routes automatically between free-tier providers, caches repeat questions, and never goes down all without touching your application code.
✓ $0 API cost · ✓ 5 provider tiers · ✓ Self-hosted
How It Works
Every request is classified, routed to the cheapest capable provider, cached for reuse, and automatically retried elsewhere if a provider fails, all in milliseconds, invisibly.
What It Does
Learned Routing
A fine-tuned classifier scores every request and routes it to the cheapest model that can actually handle it. It only escalates to something expensive when it truly has to.
Semantic Caching
Near-duplicate questions are answered instantly from a pgvector cache, at effectively zero cost. The false-hit rate is measured, not guessed.
Automatic Failover
Circuit breakers detect a failing provider in real time and reroute through the chain automatically. Requests always land on a self-hosted local model that never goes offline.
Full Observability
Every request is traced end to end with OpenTelemetry and Prometheus. Routing decisions, latency, and cost are never a black box.
Zero-Cost Operation
Built entirely on free-tier providers and self-hosted inference. No credit card, no trial period, no hidden spend. ₹0, verified.
Secure API Keys
Bearer tokens are hashed before they ever touch the database. Every request is authenticated against the hash, never the raw key.
Redis Rate Limiting
An atomic Lua-scripted token bucket in Redis enforces per-key limits with no race conditions, even under concurrent bursts.
Self-Hosted or Cloud
Route to hosted providers like Groq and Gemini, or fail over to a self-hosted Ollama model that runs entirely on your own hardware.