Status

Is Lexera up?

The check below hits our production API from your own browser, right now. Below it: what we depend on, what has actually broken before, and what we changed each time.

Checking…api · cloud run · asia-south1
first check running…

Contacting the production API from your browser…

This check runs from your browser straight to the production API — the same endpoint the app uses. No proxy, no cache, no synthetic green badge.

Known dependencies

Lexera is built on four external services. When one of them degrades, so do we — listed plainly, with what each failure looks like from your side.

Google Cloud Run

API compute — asia-south1 (Mumbai)

If it fails, the API is down and the app cannot answer at all.

Supabase (PostgreSQL)

Database + verified statute corpus — AWS Singapore

If it fails, statute lookups, workspaces and billing checks degrade or stop.

Indian Kanoon API

Live judgment retrieval

If it fails, fresh case-law retrieval degrades — answers lose live citations and say so.

Google Gemini API

Model inference

If it fails, answer generation stops until it recovers.

Incident history — resolved

Every production incident class we have actually had, with its root cause and what now watches for it. Root causes and canaries, not marketing apologies.

Cloud billing account closureJuly 2026 · resolved

The backend's billing account was closed, taking the API down (503s) until it was reopened. Resolved — a nightly canary now probes the billing tier so this class is caught before users see it.

Indian Kanoon balance exhaustionJuly 2026 · resolved

The prepaid retrieval balance ran out; fresh case-law retrieval silently degraded while cached queries looked fine. Resolved by top-up — a nightly balance canary now runs a live probe search against the IK API.

Upstream model retirementJuly 2026 · resolved

The model provider retired a version in service, causing errors and then a latency regression. Resolved by pinning maintained model aliases and re-tuning — the nightly canary checks model reachability.

Why there is no uptime percentage here

We do not publish a 99.-something badge because we have not been measuring uptime over a window long enough to state one honestly. Inventing one would defeat the point of this page. When we have a trailing figure worth defending, it will appear here — with its measurement window and method. Until then: the live check above, the dependency list, and the incident record are the honest picture.

Seeing something broken that this page calls healthy? Write to founder@lexera.co.in — it goes to the founder, not a queue. And for how the product verifies its own answers, read how we verify.