0.25.0 (2026-07-20)
Features
- add LongCat as first-class OpenAI-compatible provider (#1986) (8f81e48)
- aimlapi: provider foundation (1/5) — config, catalog, ambient-key gate (#1995) (7674d4d)
- buddy: hero pixel-art companions with signature Enter animations (#1972) (d683e85)
- codex: add GPT-5.6 family models and fix saved-model rehydration (#2014) (cb460e5)
- install: enforce and guard the zero-warning npm install contract (#2019) (ca7a7e0)
- kimi: add Kimi K3 context variants (#1989) (e3fb051)
- onboarding: first-run experience for third-party providers (#1864) (fff83a1)
- perf: tier1 token optimization — universal tool compression, doom loop detection, configurable compaction (#1869) (83d54b0)
- provider: route GPT-5.6 models to the OpenAI Responses API (#1961) (de76950)
- statusline: show token counts in context bar (ctx 74K/200K (37%)) (#1967) (626c487)
Bug Fixes
- api: enforce API_TIMEOUT_MS for OpenAI-compatible headers (#1940) (3808d19)
- api: self-heal
tool_streamrejection from non-Z.AI gateways (#1950) (#1951) (487cae7) - compaction: honor disabled auto-compact under memory pressure (#1999) (630fc2d)
- memdir: enforce entrypoint cap in bytes, not UTF-16 char length (#1918) (7b9e477)
- openai: compress tool history on Responses requests (#1958) (5399a11)
- output-style: resolve style names by own-property (#2023) (722e0c3)
- permissions: enforce read-only plan mode (#1938) (1f20e92)
- provider: support custom Anthropic bearer auth (#1929) (46e8056)
- query: bound per-turn latency growth in long REPL sessions (#1949) (#1952) (a327815)
- repl: add correction context after interruption (#1936) (eb72c77)
- resume: read the session tag from its own entry, not a tool's tag input (#1975) (1d053b2)
- session-title: ignore API error responses (#1992) (86fb6db)