Skip to content
AI usage dashboard

Qwen Code 0.25.0 with default automatic memory

On 2026-10-05, ran the official npm client in rootless Podman against local llama.cpp and Qwen2.5-0.5B GGUF. JSONL contains selected native assistant fields, not the complete transcript: anonymized IDs, same-day shifted timestamps, unchanged token values, and no bodies/project paths/other non-usage fields. Version, image/model digests and independent counts are in container acceptance.

  • One main-loop call: prompt=10,226, output=2, cache read=0, total=10,228.
  • CLI stats.bySource separately reports one automatic-memory extraction, total=5,854, absent from the session’s per-call records. CLI totals of two calls/16,082 do not become this sample’s count/total; do not subtract to invent a background event.
  • Retain schema_version=0.25.0 and raw local model name. canonical/provider remain unknown. Missing cache write/reasoning relationships remain unknown; thoughts=0 does not justify values.
  • Rescan adds/updates nothing; still one call and 10,228 tokens.