log-analysis

Log Analysis

Inspects Python application logs from any environment — local filesystem, Docker containers, or a remote server via SSH. Produces a raw tail, filtered error list, pattern search, or period summary. Handles Python tracebacks, Django request logs, uvicorn/gunicorn access logs, structlog JSON, and Celery task logs.

Available actions

#	Action	Role	Input
01	tail	Display the last N lines from a log source	env, source, optional lines
02	parse-errors	Extract and group Python exceptions and HTTP errors	env, source, optional since
03	search	Search for a pattern, status code, or time range	pattern, optional env, source
04	summarize	Period digest: error rate, top exceptions, 5xx count	optional env, optional since

Default flow

Non-sequential — dispatch based on user intent:

• "tail / show / last lines / recent logs / check logs" → tail
• "errors / exceptions / tracebacks / error list" → parse-errors
• "find / search / grep / [specific pattern or status code]" → search
• "summary / digest / report / stats" → summarize
• Env not specified → ask: local, docker, or prod?
• No clear intent → tail on app + web-server, then offer parse-errors

Transversal rules

• Never print or log SSH credentials, passwords, or key paths.
• Always show timestamps in output.
• Truncate individual log lines to 300 chars to avoid flooding context.
• When errors are found, always suggest a likely fix or next investigation step.
• Default env is docker when Docker is detected in the project; otherwise local.
• For prod: if no SSH host is configured, ask user for user@host before proceeding.
• Multiline Python tracebacks must be kept intact — never truncate a traceback at a line boundary.

References

• references/environments.md — env detection, standard log paths, SSH config pattern
• references/log-formats.md — Python/Django/uvicorn log formats and parsing patterns