005
OSS / EXPERIMENTAL V0.3 · AI & models
// One route through the model mess
Adaptive LLM Gateway
An OpenAI- and Anthropic-compatible gateway for local models, subscription CLIs and APIs — with routing, audit, PII redaction and injection defenses built in.
LLM routingSecurityOpen source
GITHUB_PULSE / SYNCINGPUBLIC REPOSITORYAUTO-SYNC / PUBLIC DATA ONLYLAST PUSH…syncing with GitHub
LANGUAGE…primary repository language
RELEASENO FORMAL RELEASElatest public GitHub release
LICENSENOT DECLAREDrepository metadata
01 // THE_PROBLEMWHY THIS
EXISTS.
Developers now pay for several AI subscriptions, run local models and still need separate integrations for every agent. Traditional gateways assume pay-per-token API keys and rarely make security or audit part of the default path.
02 // THE_APPROACHTHE USEFUL
LOOP.
The gateway discovers local and subscription-backed providers, exposes compatible endpoints and passes every request through routing, privacy and audit layers before it reaches a model.
03 // WHAT_SHIPPED
WHAT'S
INSIDE.
01ONE COMPATIBLE EDGE
OpenAI and Anthropic clients can reach local models, APIs and supported subscription bridges through one endpoint.
02SECURITY IN THE PIPE
Prompt-injection checks, PII redaction and output-stream defense run as first-class gateway stages.
03ROUTING WITH RECEIPTS
Cost-aware routing, quota tracking, replay and audit records make model choice explainable and debuggable.