How it works

One call in. Deduped, routed, capped, metered — out.

You ask for needs, never a vendor. Conflux hashes the file, serves a cache hit for free, routes a miss to the cheapest healthy provider, and never lets the bill surprise you.

● before — every vendor, by hand
# five keys, five bills, five retries
sightengine.check(file)   $
acrcloud.identify(file)   $
tineye.search(file)       $
vision.moderate(file)     $
openai.describe(file)     $

# same file next week? pay it all again.
● after — one call, deduped & routed
const r = await conflux.check(file, {
  need: ['moderation', 'music', 'ai-gen'],
})
// hash seen before → served from cache, $0
// new file    → cheapest provider per need
// over budget → capped, never a surprise bill

r.aiGenerated // → 0.94
Request lifecycle

What one check() actually does.

  1. 01
    Authenticate
    Resolve the key → org + environment (live/test).
  2. 02
    Hash
    Stream the file's bytes into a content hash — the cache key.
  3. 03
    Cache lookup
    Seen this (org, hash, need) before? Return it. $0, marked saved.
  4. 04
    Route
    Miss → cheapest, healthiest provider that satisfies the need.
  5. 05
    Cap check
    Over the day/org/need ceiling? Degrade or queue — never a surprise charge.
  6. 06
    Call + normalize
    One provider call, one failover hop, a single normalized result.
  7. 07
    Meter
    Write cache + record charged/saved, latency, provider. Respond.

Point one SDK at Conflux. Watch the duplicate bills disappear.

Built by infrastructure engineers who got tired of paying five vendors to check the same file.

Get early accessSee pricing →
We use a single analytics cookie to understand traffic to this site. Nothing loads until you choose. See our Privacy Policy.