Max Mayfield
167d3be2e4
- anomaly.rs: Slack Block Kit webhook + Resend email on 3x cost spike - digest.rs: Weekly HTML digest email via Resend with model usage + savings tables - main.rs: Daily pricing refresh with hardcoded table (OpenAI/Anthropic/Google models) - handler.rs: AES-256-GCM encryption for provider API keys (nonce || ciphertext storage)
dd0c/route — LLM Cost Router & Optimization Dashboard
Drop-in OpenAI-compatible proxy that routes AI requests to the cheapest capable model.
Quick Start
# Start local infra
docker compose up -d
# Run the proxy
OPENAI_API_KEY=sk-your-key cargo run --bin dd0c-proxy
# Test it
curl http://localhost:8080/v1/chat/completions \
-H "Authorization: Bearer your-dd0c-key" \
-H "Content-Type: application/json" \
-d '{"model":"gpt-4o","messages":[{"role":"user","content":"Hello"}]}'
Architecture
- Proxy Engine (Rust/Axum) — <5ms overhead, SSE streaming, async telemetry
- Router Brain — Complexity classification, cost-based routing, fallback chains
- Dashboard API — REST API for config, analytics, team management
- TimescaleDB — Time-series telemetry with continuous aggregates
- PostgreSQL — Config, auth, routing rules
- Redis — API key cache, rate limiting
Project Structure
src/
proxy/ — Proxy server (main entry point)
api/ — Dashboard API server
worker/ — Background jobs (digests, alerts)
router/ — Routing logic & complexity classifier
auth/ — API key validation, JWT, OAuth
config/ — App configuration
data/ — Data layer traits (EventQueue, ObjectStore)
analytics/ — PostHog PLG instrumentation
migrations/ — PostgreSQL + TimescaleDB schemas
tests/ — Unit, integration, E2E tests
infra/ — CDK / deployment configs
Pricing
| Tier | Price | Requests/mo | Retention |
|---|---|---|---|
| Free | $0 | 10K | 7 days |
| Pro | $49/mo | 1M | 90 days |
| Enterprise | Custom | Unlimited | 1 year |