Issues

• Turn has no latency field, so "how long the AI took" is unmeasurable — and the sources disagree on whether it is knowable

  enhancement
  m8a-io/modelog#24 · Filter by author smolinari opened on Oct 29, 2026

• Measure whether prompt-cache keepalives actually save money (low priority)

  question
  m8a-io/modelog#23 · Filter by author smolinari opened on Oct 29, 2026

• Session outcome records: measure what the work was worth, without letting the model author the record

  enhancement
  m8a-io/modelog#22 · Filter by author smolinari opened on Oct 29, 2026
  2

• Detect and describe the subscription→usage-credits billing state (hasExtraUsageEnabled, fallback_credit, cost-state)

  enhancement
  m8a-io/modelog#12 · Filter by author smolinari opened on Oct 12, 2026
  2

• Cost silently stops being reported after a billing-state change

  enhancement
  m8a-io/modelog#10 · Filter by author smolinari opened on Oct 11, 2026
  3

• compare_models reports means with no dispersion or sample size

  enhancement
  m8a-io/modelog#6 · Filter by author smolinari opened on Oct 4, 2026

• No tool returns token counts, so a per-turn cost difference cannot be decomposed

  enhancement
  m8a-io/modelog#5 · Filter by author smolinari opened on Oct 4, 2026

• Price each turn at the rate in force when it ran, instead of one global snapshot

  enhancement
  m8a-io/modelog#3 · Filter by author smolinari opened on Oct 4, 2026

• Add a setting to override date/time formatting, independent of the host locale

  enhancement
  m8a-io/modelog#2 · Filter by author smolinari opened on Oct 4, 2026

• Add a toggle for real-time dashboard updates (and give Rescan a purpose)

  m8a-io/modelog#1 · Filter by author smolinari opened on Sep 28, 2026