🧙 Log Alchemist — Agentic Incident Response for HPC

Upload one or many log files, or paste logs directly. Uploading a screenshot will OCR automatically and fill the Logs box.

MCP tools exposed: extract_incident, assess_incident_scope, build_incident_bundle, parse_hpc_failure_artifacts, build_collection_manifest, open_incident_session, append_incident_artifact, assess_evidence_gaps, build_agentic_response_plan, suggest_mcp_companions, generate_playwright_mission, resolve_incident_session, find_similar_incidents, suggest_fix.

What makes this different from a generic chat upload: it preserves privacy, mines recurring log templates, distinguishes access issues from platform/orchestrator incidents and true runtime failures, parses artifacts from systems like TorchElastic, Slurm, and containerized shell platforms, builds local incident memory over time, and can orchestrate a stateful MCP investigation across filesystem, shell, Playwright, docs, and telemetry companions.

Advanced agentic views stay compact in the UI until the upload is grounded enough to trust them.