Janus Edge documentation
Everything an administrator or a user needs: installing and licensing Janus Edge, connecting models, controlling who uses what, and the full API and configuration reference.
For Janus Edge 2026.11.1, the current release. The same documentation ships inside every gateway at /docs.
Install and licenseInstall Janus EdgeLinux server, Kubernetes, Docker and offline archives, then licensing and upgrades.
Get startedQuick start1 page: Quick start
User guideWorking with tokens5 pages: Working with tokens, Choosing a model, Quotas and limits, Streaming responses…
Admin guideChat web pages21 pages: Chat web pages, Setting up Janus Chat, Configuring upstreams, Self-hosted engines…
ReferenceDivergences from the original decisions8 pages: Divergences from the original decisions, Troubleshooting, API reference, Proxy endpoints…
ReferenceCommand line and environmentEvery
janus and janus-ctl option and all 89 environment variables, with defaults, on one page.
ReferenceError catalogEvery error code Janus returns, when it happens and what to do.
Get started
- Quick startFrom first sign-in to a metered request in about five minutes.
User guide
- Working with tokensCreate, rotate, and revoke credentials safely.
- Choosing a modelHow the model list is filtered and priced.
- Quotas and limitsWhat a quota measures, when it resets, and what happens on breach.
- Streaming responsesHow Janus relays Server-Sent Events and still counts tokens.
- Janus ChatChat with every model you can use, in the browser, metered and protected like API traffic.
Admin guide
- Chat web pagesWhich web pages chat models may open: the public internet by default, or your intranet on offline networks.
- Setting up Janus ChatWho can use Chat, which models it offers, its tools, and the instructions every chat starts with.
- Configuring upstreamsAdd a provider, discover its models, and price them.
- Self-hosted enginesServer flags that give Janus exact usage, real speed, the right context size and live load for vLLM, llama.cpp, Ollama and TEI.
- Setting quotasChoose a metric, a window, and a breach behavior.
- Pricing by modalityEvery rate card is per million tokens — here is how each kind of model actually bills, and what to type in.
- Policy blocking rulesSteer well-behaved clients — and understand what these rules cannot do.
- Usage insightsLabel requests by the kind of work they ask for (coding, legal, writing, …) and see the mix by team, model and person.
- Auto-routingLet Janus pick the model for each request: a classifier decides what kind of work it is and how hard, and a router sends it to the model you chose for that work.
- Classifiers: speed and accuracyHow Janus asks decision models several questions at once, which models suit which job, and how to benchmark what your classifier choices cost each request.
- Security GatewayPolicy-driven filtering of prompts and responses: secrets, PII, codenames, request shape, and prompt injection.
- Accounts and sign-inJanus accounts (email + password + two-factor), SSO, first-run setup and break-glass.
- LicensingEditions, seats, what happens when a key expires, and how to install one.
- Operating JanusConfiguration, retention, health checks, and observability.
- Janus FIPS editionThe separate build whose cryptography runs in a FIPS 140-3 validated module, what it changes, and how to prove which edition is running.
- Listener TLSEncrypt the hop from your proxy to Janus: an issued certificate, or a temporary self-signed one.
- Trusted certificatesOne trust store for every outbound TLS connection: the host’s roots, the built-in DoD PKI roots, mounted ConfigMaps and files, and certificates uploaded in the app.
- STIG configuration profileJANUS_STIG_PROFILE=true: one switch that sets the STIG defaults, refuses unsafe settings at startup, and makes administrative changes fail closed when they cannot be audited.
- Security settingsSession time-outs, the sign-in notice, lockout, password rules, audit forwarding to syslog or a SIEM, and smart-card (CAC/PIV) sign-in.
- Bootstrapping the first administratorBreak-glass access for a fresh deployment.
- Test identity provider (Authentik)Run a local Authentik to exercise real OIDC sign-in, group sync, and auto-grants.
Reference
- Divergences from the original decisionsWhere the shipped system departs from the decision record, why, and what it costs you.
- TroubleshootingSymptom-first diagnosis, and what every error code you can receive means.
- API referenceThe three API surfaces, how to authenticate against each, and the machine-readable OpenAPI 3.1 description.
- Proxy endpointsPer-endpoint reference for the OpenAI-compatible surface at PUBLIC_URL/v1.
- Environment variablesEvery configuration variable the gateway reads, with defaults and constraints.
- FAQShort answers to the questions new users and administrators ask first.
- ChangelogWhat changed in each release of this gateway. Generated at build time from the repository CHANGELOG.md, the authoritative copy of this history.
- Third-party noticesAttribution for the open-source software this product includes.