Skip to content

v1.15.3

Choose a tag to compare

View all tags
@github-actions github-actions released this 22 Jul 19:58
v1.15.3
b35af71
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

v1.15.3 — Cache Metrics Honesty

A small fix release answering the "I do not see cached values" report: the stats line showed cache: 0 write / 0 read / 0 total on every turn for some providers.

Fixes

  • DeepSeek cache metrics are now parsed. odek understood Anthropic (cache_creation/read_input_tokens) and OpenAI (prompt_tokens_details.cached_tokens) — but not DeepSeek's native prompt_cache_hit_tokens / prompt_cache_miss_tokens, which DeepSeek endpoints and gateways often return instead of the OpenAI-style details. DeepSeek hits now count as cache read, misses as cache write (DeepSeek caches missed prefixes automatically). Verified against a live DeepSeek account: native hit/miss numbers agree exactly with the OpenAI-style cached_tokens, with no double counting.
  • No more fake zeros. When a provider returns no cache metrics at all, zeros falsely implied "caching ran and missed". A new CacheReported flag (LLM client → loop → stats) distinguishes "no data" from "0 tokens", and the Telegram stats line now shows cache: n/a (provider reports no cache metrics) instead.

What you'll see after upgrading

  • Real numbers (cache: 8143 write / 15488 read / …) if your provider is DeepSeek or returns the OpenAI/Anthropic fields.
  • cache: n/a if your provider doesn't report cache usage — a definitive answer, not a confusing zero.

Assets

Binaries are available for linux/amd64, linux/arm64, darwin/amd64, and darwin/arm64. Verify with checksums.txt.

Full Changelog: v1.15.2...v1.15.3

Assets 7
Loading