Skip to content

v6.18.0

Choose a tag to compare

View all tags
@github-actions github-actions released this 27 Oct 03:26
· 64 commits to master since this release
v6.18.0
70ac706
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Auto-released from merge of PR #1749 (trigger: push).

  • Cache trace: why a ChatGPT (codex) turn missed the prompt cache, not only how much it read. Every request on that path now records, next to its cached tokens, what it shared with the previous request on its prompt_cache_key: the gap in milliseconds, whether its input extends the previous one item for item and where it diverged if not, and whether the seat or the model, system prompt and tool names changed. Hashes and counts only. It rides on the /analytics row as cache, on the --log-file record as cache_*, and as one stdout line per request with the new --cache-trace flag or DARIO_CACHE_TRACE=1 (and under -v). Docs: docs/analytics.md "The cache trace".
  • Docs: the multi-instance page no longer says there is no shared-state mode. Its closing section predated --pool-shared-state (6.12.7). It now says what the lock and the shared state cover, what is still per instance (ChatGPT seats, the ledger, the overage halt), why routing does not read Redis on the request path, and where to ask for the rest (#1733).

Built + tested + npm-published inline by cc-drift-auto-release.yml on this run — tokenless via OIDC trusted publishing, provenance-attested. See the workflow log for the publish output.

Assets 5
Loading