Tobias Glawe

LogCarve is a self-hosted, AI-assisted access-log analyser for website owners, developers, SEO specialists and security reviewers. Browser analytics alone cannot explain everything reaching a web server. Crawlers, automated requests, broken links and suspicious activity often require separate investigation. LogCarve turns Nginx and Apache access logs into searchable requests, traffic charts, crawler breakdowns, error reports and practical investigation evidence. The dashboard, authenticated read-only API and internal AI assistant share one deterministic analytics layer. Users can filter traffic, inspect individual requests, compare time windows and export results. Core analytics remain available without an AI model or Internet connection. Powered by Meta’s Muse Spark 1.3, the assistant investigates questions through bounded, read-only analytics tools. Answers include evidence references, filters and coverage limitations. A local recheck can identify whether saved evidence still matches the current data after imports, log rotation or retention changes. The API also enables external assistants to combine server evidence with separately configured analytics or operational tools. The LogCarve application was developed during the hackathon. Seventy bot-catalog patterns and labels were adapted from LogWerk with their MIT attribution retained; no LogWerk application code or assets were copied. Original LogCarve code is Apache-2.0. The demonstration uses synthetic data.