FFluxen
How it works

Point your existing client at Fluxen

No SDK changes. Fluxen is OpenAI-compatible — swap the base URL and API key, and every request starts flowing through the gateway.

  1. 1

    Connect an application

    Create an application in the dashboard, issue an API key, and point your client's base_url at your Fluxen deployment. Traffic starts appearing within seconds.

  2. 2

    Understand

    Fluxen accounts for every request — cost, tokens, model, provider, latency, cache status — no sampling. Application Detail shows spend, model mix, and trends immediately.

  3. 3

    Identify

    Once an application clears a minimum traffic/spend floor, four detectors run against its real history and surface opportunities with evidence, confidence, and an estimated dollar impact.

  4. 4

    Simulate

    Before anything changes, replay the recommended scenario against real historical requests and see the actual delta — not an estimate extrapolated from a formula.

  5. 5

    Control

    Apply the change with one confirmed action. Fluxen writes a versioned policy and enforces it on the gateway's hot path immediately.

  6. 6

    Measure

    7 and 14 days later, Fluxen checks what actually happened against real traffic and gives an honest verdict — successful, partial, no effect, or regressed. A regression is one click from reverted.

Deployment

docker compose up

The entire product — gateway, control API, background jobs, dashboard, Postgres, Redis — runs from one repository with one command. No message broker, no Kubernetes, no managed service required.