ai root cause
AI log analysis that
finds the root cause
Ask what broke. Get an answer, not a trace.
Logier's AI reads the entries around an error — the ones before it that set the stage, and the ones after it that show the fallout — then writes a plain-English explanation of what happened, when, and why. It works on any parsed log, including custom and proprietary formats.
translation, not just detection
Root cause, not just
the error line
What you'd read
ConnectionPoolExhaustedError: timeout after 30000ms at Pool.acquire (pool.js:142) at WebhookWorker.process (worker.js:88) at RetryQueue.drain (queue.js:53) ... 12 identical entries in the next 90s
What Logier's AI says
The payment webhook started failing at 14:02:07 because the retry queue filled up after the database connection pool was exhausted three minutes earlier. The dozen errors that follow are the same underlying failure repeating, not a dozen separate problems.
what it actually looks at
What the model reads
before it answers
Timing
What happened immediately before the failure, not just the failure line itself.
Clustering
Whether this is one of thousands of identical, already-triaged errors, or something genuinely new.
Surrounding context
The request, session, or job ID that ties related lines together across the file, even when they're far apart.
History
Whether the same signature showed up earlier in the session, and whether it was already dismissed.
bring your own key
Bring your own API key — or
run a local model offline
Logier doesn't route through a hosted model of its own. Connect the provider you already trust — including a fully local one — and only the specific excerpt you're asking about is ever sent.
OpenAI
gpt-4o-mini by default
Anthropic
Claude models
Gemini 1.5 Flash
Local
Ollama, LM Studio, or any OpenAI-compatible endpoint
ask it anything
Questions developers
ask their logs
“Why did this spike start at 14:02?”
A deploy landed at 14:01:52. The first timeout appears nine seconds later, and every failing request shares the same downstream service.
“Is this the same error as yesterday?”
No — yesterday's cluster was a client-side validation error. This one is a new signature: a connection reset that hasn't appeared in this file before.
“What changed right before this?”
The retry queue crossed its configured size limit about three minutes earlier, which lines up with when the connection pool started rejecting new requests.
worth knowing
What it won't do: guess
Logier's AI only reasons over what's actually in the log you've given it. If the real root cause lives outside that window — an upstream provider, a config change that was never logged — it says so instead of inventing an explanation that sounds confident but isn't grounded in anything. Treat it as a fast, honest first read, not a final verdict.
where this lives
Copilot reads the file you already have open
There's no separate console to paste logs into. The assistant sits inside the file view and already has the parsed entries in context.
Explorer
Cloud
- 1
Open the assistant from any file view. It answers against the entries already parsed in this tab.
- 2
Select an error and ask about it directly — only the excerpt around that entry is sent to your provider.
- 3
Ask about a spike you can see rather than describing it: the surrounding window is what gets read.
the data boundary
What gets sent, and what
never leaves your machine
Every other part of Logier runs locally whether or not you configure an AI provider. Only the excerpt around the error you ask about is sent, and only to the provider you connected yourself.
| What | Where it goes |
|---|---|
| Format detection and parsing | Always local — never sent |
| Clustering, timeline, anomaly detection | Always local — never sent |
| Search and filtering | Always local — never sent |
| The full log file | Never sent, under any configuration |
| The excerpt around the error you ask about | Sent to the provider you configured — or to your local model |
faq
Common questions
keep reading
Before and after the explanation
AI log analyzer
How desktop AI log analysis compares to cloud observability platforms, and what Logier does with no AI configured at all.
Automatic format detection
The AI reasons over parsed entries, so it works the same on a custom in-house format as on Apache access logs.
Anomaly detection
Anomalies surface what's worth asking about; this page explains why it happened.
Local-first log privacy
The full data-handling picture, including running Logier on an air-gapped machine.
Logier vs. lnav
lnav gives you SQL over your logs; Logier gives you an explanation. Both are useful.
Offline log viewer
Running the whole thing — including AI analysis — against a local model, with nothing leaving your machine.
Let the AI read it first.
Point it at an error and get a plain-English explanation before you've opened the stack trace.