v7.2.148
Added 4
- Support cache write tokens in claude responses
- Add claude fable 5.1 and gemini 3.8 flash models
- Use home model capabilities for thinking
- Make subagent session affinity configurable via session-affinity-subagents
Changed 2
- Upgrade default Claude Code baseline and fingerprint to 2.1.258
- Use request thinking helper for kimi
Fixed 11
- Support native thinking signatures without prefixes in claude translator
- Retry model fetching per endpoint and prioritize daily base url
- Preserve unsigned gemini thinking blocks with trailing carriers
- Harden concurrent session handling, listener lifecycle, and token accounting
- Strip tool config, labels, and session id in token counting
- Bypass quota cooldowns and credit hints when cooling is disabled
- Validate access token expiration and retain valid credentials on refresh failure
- Map user_credits_insufficient to 402 and user_period_limit_exceeded to 429
- Enforce minimum cooldown floor and track attempted credentials on 429
- Preserve home model capability metadata
- Clear unsupported reasoning levels
Removed 1
- Remove minimal level from dynamic_allowed definitions in models
From CLIProxyAPI
Linux release assets
CLIProxyAPI_<version>_linux_<arch>.tar.gzis the default Linux build. It supports dynamic library plugins and is built against a GLIBC 2.17 baseline.CLIProxyAPI_<version>_linux_<arch>_no-plugin.tar.gzis the portable Linux build for musl-based or older systems such as OpenWrt. It does not support dynamic library plugins.
FreeBSD release assets
CLIProxyAPI_<version>_freebsd_aarch64_no-plugin.tar.gzis the FreeBSD arm64 build. It is built without CGO and does not support dynamic library plugins.
Changelog
- feat(translator): support cache write tokens in claude responses (893abbab)
- fix(antigravity): support native thinking signatures without prefixes in claude translator (15231e9f)
- fix(antigravity): retry model fetching per endpoint and prioritize daily base url (c2834b68)
- fix(antigravity): preserve unsigned gemini thinking blocks with trailing carriers (8deeb4ac)
- fix: harden concurrent session handling, listener lifecycle, and token accounting (02c02cda)
- fix(antigravity): strip tool config, labels, and session id in token counting (d0fb44ca)
- fix(antigravity): bypass quota cooldowns and credit hints when cooling is disabled (272c1cff)
- feat(registry): add claude fable 5.1 and gemini 3.8 flash models (dacae582)
- chore(registry): remove "minimal" level from dynamic_allowed definitions in models (bdcccfb8)
- fix(auth): validate access token expiration and retain valid credentials on refresh failure (9812b1e7)
- fix(home): map user_credits_insufficient to 402 and user_period_limit_exceeded to 429 (f416175f)
- fix(auth): enforce minimum cooldown floor and track attempted credentials on 429 (18e01a76)
- fix(claude): upgrade default Claude Code baseline and fingerprint to 2.1.258 (df7e04ea2)
- fix(models): preserve home model capability metadata (1ecf0cb6)
- feat(auth): use home model capabilities for thinking (6ff680e9)
- refactor(kimi): use request thinking helper (c6dd8214)
- fix(codex): clear unsupported reasoning levels (cdda333c)
- Merge pull request #5436 from router-for-me/models (63fdd77f)
- feat(routing): make subagent session affinity configurable via session-affinity-subagents (d577e630)
What's Changed
- fix(models): apply Home-provided model capabilities during execution by @hkfires in https://github.com/router-for-me/CLIProxyAPI/pull/5436
Full Changelog: https://github.com/router-for-me/CLIProxyAPI/compare/v7.2.147...v7.2.148