PixelOkay
Private browser workspace
Tools
PIXELOKAY / DEVELOPER TOOLS

Log Analyzer

Find repeated failures, reconstruct incidents, and investigate probable causes with evidence you can trace to the original log lines.

Pattern clusteringException explanationsIncident timelineLocal-first analysis
ANALYSIS WORKSPACE

Log Intelligence Studio

Runs in your tab. No log uploads or API keys. Findings are hypotheses, not confirmed diagnoses.

02 / RESULTS

Investigation overview

Your findings will appear here

Add logs or load the sample, then choose Analyze logs to see error clusters, probable causes and evidence.

03 / OPTIONAL ON-DEVICE AI

Explain log evidence with local AI

Analyze logs first. Rules and exports work without any model.

Model setup is optional. The deterministic analyzer remains available.

Run Analyze logs to prepare evidence. AI uses only bounded, redacted findings and patterns—not your original source files.

AI suggestions are unverified hypotheses. Logs can contain secrets that pattern redaction misses; review before sharing. Model scripts and weights download from external CDNs only after you choose to load a model. Log evidence is processed locally.

LOCAL AI SETUP

Set up AI on this device

First, a model needs to be loaded on your device

Log explanations run inside your browser. Model files may be hundreds of megabytes and can be reused from compatible PixelOkay AI tools on the same origin. Nothing downloads automatically on entry.

  1. 1. Check device
  2. 2. Download / cache
  3. 3. Initialize model
  4. 4. Explain evidence

Device & model resources

Approximate measurements
Device RAM / model memoryChecking device information…

Live system RAM, GPU memory and model working-memory requirements are not reliably exposed in browsers.

Checking device compatibility…

Browser website storageChecking storage…

Site-wide storage estimate, includes other PixelOkay tools. Not RAM.

Model cache not yet inspected.

ONE ACTION · ALL RECOGNIZED AI MODEL DOWNLOADS

Free AI model storage across PixelOkay

Scan this website origin for WebLLM and Transformers.js model caches, plus recognizable model files in other cache buckets. Includes other PixelOkay tools on this origin. Unrelated site data is not cleared automatically.

Run a scan to inspect recognized model downloads.

Unknown caches, browser-native models, inaccessible IndexedDB and OPFS files are not deleted; close other active AI tabs before cleanup.

Having trouble downloading or loading AI?

Refresh Storage & Cache, then retry setup or select a smaller model. Only when needed, inspect recognizable AI model caches; clearing them can require multiple PixelOkay tools to redownload files. Never clear all website data without backing up your work.

Choose a model and check compatibility.
Technical details

WebLLM 0.2.87 · WebGPU, no paid API. Large model weights begin only after explicit consent; runtime assets may load to check cache status.

Remove recognized model caches?

Only identified model-cache entries are targeted. Saved logs, chats, projects, editor work, and unrelated site caches are not targeted; the scanner cannot inspect all browser storage providers.
HOW IT WORKS

Evidence before assumptions.

01

Parse mixed log sources

Text logs, newline-delimited JSON, JSON arrays, ISO timestamps, multiline Java/Python/Node stack traces, and common severity tags. Missing dates remain undated rather than fabricated.

02

Cluster repeat failures

Variable fields such as request IDs, IP addresses, UUIDs, and timings are normalized for deduplication. Click any cluster to inspect the underlying redacted messages.

03

Investigate and export

Signature matches explain common failure families—database, networking, memory, TLS, and more. Every finding points to the original source and line. Export Markdown, CSV or JSON.

What makes this “AI-assisted,” and what are its limits?

The primary analyzer uses transparent pattern matching and heuristic correlation. An optional WebGPU model can explain a bounded evidence summary after explicit model loading. Suggested causes are diagnostic leads, not confirmed root causes. It cannot infer events absent from logs, inspect infrastructure, query live systems, or guarantee that messages are safe to share. Review evidence and keep the original logs for forensic work.

Which formats and limits are supported?

Up to 12 local UTF-8 files and 20 MiB of combined input, 100,000 parsed events, 180,000 lines per source, and 20,000 distinct patterns. Binary files and compressed archives are not decoded. Timezone-free ISO timestamps are interpreted as UTC and flagged; syslog timestamps without a year are left undated. Very large forensic datasets are better handled in a streaming log backend.

Analyze and investigate local log files

Parse supported text and JSON logs, group repeated failures, correlate request or trace identifiers, and inspect probable incident causes with links back to source lines. Optional local AI can explain a bounded, redacted evidence summary.

  1. Paste log lines or add supported UTF-8 text, JSON or JSONL files, then run the deterministic analyzer.
  2. Review repeated-message clusters, incident windows and request/trace correlations. Filter using timestamps found in the logs and inspect evidence before forming a hypothesis.
  3. Export Markdown, JSON or CSV reports, then review the redacted output for secrets before sharing it.

Example: If two checkout errors show ECONNREFUSED with different request IDs, inspect the grouped messages and nearby database warnings, then verify the suspected connection issue against the original lines.

Limits and privacy: Supports up to 12 files and 20 MiB combined, 100,000 events and 180,000 lines per source. Heuristic causes are hypotheses; timestamps are interpreted from log data, and undated events are excluded from time correlations. Pattern redaction can miss secrets. Optional AI runs in-browser on bounded redacted findings; external runtime/model files download only after you choose a model.

Related tools