router-for-me/CLIProxyAPI v7.2.133

on GitHub

Linux release assets

• CLIProxyAPI_<version>_linux_<arch>.tar.gz is 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.gz is 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.gz is the FreeBSD arm64 build. It is built without CGO and does not support dynamic library plugins.

Changelog

• perf(api): skip inactive request interceptors (970529b)
• perf(openai): reduce websocket transcript merge allocations (f9bd9de)
• test(openai): harden websocket transcript allocation coverage (f8bcd1c)
• fix(openai): preserve duplicate websocket input semantics (49b2f89)
• fix(openai): match websocket item metadata case-insensitively (baa11ed)
• fix(gemini): preserve additionalProperties:false for Antigravity response schemas (61c4fd8)
• Merge pull request #4928 from ramapitecusment/codex/websocket-transcript-allocations-v2 (8b02fed)
• Merge pull request #4360 from router-for-me/perf/skip-inactive-request-interceptors (810d4dd)
• fix(auth): avoid cooldown for request-scoped 401 request-faults (203f5b1)
• fix(claude): keep tool_result blocks first when injecting currentDate (297139c)
• fix(claude): add Anthropic unified rate-limit parsing helpers (ac82bed)
• fix(openai): ensure Responses usage includes token detail fields (e0b4956)
• feat(models): add GPT-5.6 Sol Work Mode model registrations and Codex client config updates (dd21444)
• feat(models): add max_completion_tokens to model definitions and responses (046b59e)
• fix(auth): make session-affinity updates safe for rebound sessions (9169ad5)
• Merge pull request #4994 from router-for-me/models (48d0218)
• fix(gemini): merge conditional schema branches during JSON schema cleanup (f53a2b6)
• fix(openai): propagate environment context and sanitize antigravity generation config in interaction adapters (10afcc8)
• chore(models): remove GPT-5.6 Sol Work Mode registrations from model registry config (c1ff55f)
• fix(claude): deduplicate duplicate tool outputs in OpenAI-to-Claude request conversion (1ecb7df)
• fix(claude,gemini,antigravity): centralize tool-call ID generation and harden signature sanitization (616d1b1)
• perf(translator): batch assemble response arrays and merge multi-choice parts (124dab6)
• fix(gemini): simplify OpenAI chat response content and choices serialization (a581838)
• perf(translator): batch assemble translation arrays before writing JSON (0c58c3c)
• perf(translator): switch response translators to batched raw-array insertion via SetRawArrayItems (8b3b304)
• pref(claude): keep raw Claude tool IDs for deduplication in request translation (7efe0a7)

What's Changed

• perf(openai): reduce websocket transcript merge allocations by @ramapitecusment in #4928
• perf(api): skip inactive request interceptors by @G0d2i11a in #4360
• fix(claude): keep tool_result blocks first when injecting currentDate by @mardausdennis in #4971
• feat(models): add max_completion_tokens to model definitions and responses by @hkfires in #4994

New Contributors

• @ramapitecusment made their first contribution in #4928
• @G0d2i11a made their first contribution in #4360
• @mardausdennis made their first contribution in #4971

Full Changelog: v7.2.132...v7.2.133