v0.33.0

0.33.0 (2026-10-14)

Features

• provider: add GMI Cloud as a provider and official partner (#2279) (e736c28)
• sessions: add /sessions to resume or delete saved conversations (#2262) (fc3be25)

Bug Fixes

• statusline: reset context usage after full compaction (#2275) (fc82603)

v0.32.0

0.32.0 (2026-10-12)

Features

• add AnySearch as an explicit web search backend (#2261) (5bc160f)
• gateway: add API Route hybrid gateway (#2225) (9a2910d)
• publish as @twigpine/openclaude (#2269) (50d4058)
• search: make Exa the default web search backend and add /search (#2257) (790c009)

Bug Fixes

• api: validate retry count environment values (#2172) (76b12a6)
• hooks: match the .sh suffix case-insensitively on Windows (#2254) (53c7ed9)
• openai-shim: keep terminal usage from a repeated finish_reason (#2252) (10254f4)
• openai-shim: strip store for Anthropic models on OpenAI-compatible routes (#2242) (#2247) (0bef4d2)

Performance Improvements

• read: dedup touched-but-identical files and covered sub-ranges (#2259) (5ff58c4)

v0.31.0

0.31.0 (2026-09-22)

Features

• launcher: support --max-old-space-size-percentage (#2219) (055e5d5)
• providers: add Command Code hybrid gateway (#2196) (5b9daef)
• query: configure idle timeout (#2233) (876fb79)
• skills: add verify for revocations and eyebrow drift checks (#2215) (eb3c590)
• skills: enforce registry revocations.json on install (#2187) (0ea8eef)
• web-search: add Ollama backend (#2234) (70d97fd)
• zai: add GLM-5.3-Flash Coding Plan support (#2185) (aceacf0)

Bug Fixes

• agent: forward subagent permission prompts to parent session (#2235) (a07b06c)
• api: validate API_TIMEOUT_MS in the native Anthropic clients (#2169) (35d7edc)
• bg: clean up old completed session artifacts (#2167) (70fe2de)
• chrome: launch the packaged CLI entrypoint in npm installs (#2146) (bfb44ff)
• ci: restore main validation (#2236) (5fb5abb)
• ctx: let explicit overrides beat discovered context windows (#2082) (3451187), closes #2081
• frontmatter: anchor the closing --- to the start of a line (#2027) (cbbe10f)
• hooks: preserve compound Bash commands on Windows (#2217) (0c2043f)
• identify OpenCode Go requests (#2203) (3609d75)
• image: allow large/metadata-less screenshots through on resize failure (#1964) (#1970) (2e687df)
• mcp: resolve server names by own-property, not the prototype chain (#1983) (c3930ca)
• openai-shim: request streaming usage from compatible local providers (#2144) (decad0e)
• plugins: keep Windows marketplace cache on copy ENOENT (#2220) (e2b021d)
• profiles: apply context limit to all profile models (#2201) (0b68e57)
• sdk: preserve async generator session context (#2204) (0abfca3)
• xai: validate OAuth state before settling callbacks (#2228) (d16318a)

Performance Improvements

• bg: avoid full scans for exact session IDs (#2166) (55e9b24)

v0.30.0

0.30.0 (2026-08-31)

Features

• providers: add focused LLMTR hybrid gateway (#2150) (1e56d4e)
• providers: live model lists for OpenRouter and OpenGateway (#2084) (bb6d66f)

Bug Fixes

• api: scope Anthropic attribution to compatible request paths (#2147) (30090c7)
• bg: identify sessions with persisted process markers (#2163) (ca7c3ef)
• effort: preserve known model exclusions when force-enabled (#2148) (69aca78)
• integrations: keep managed AIMLAPI attribution over caller headers (#2179) (aaccb57)
• openai-shim: drop synthetic tool-results marker and guard echoes (#2039) (#2153) (54f963d)
• plugins: anchor marketplace hostPattern against lookalike hosts (#2177) (5f1ab9b)
• settings: preserve concurrent updates (#2137) (34536c6)
• settings: stop proto-named permission rules from aborting validation (#2170) (8db8830)
• tui: proper Unicode/IME input handling for composed sequences (#2018) (#2154) (e802626)

v0.29.1

0.29.1 (2026-08-19)

Performance Improvements

• stop busting the prompt cache and slim per-turn context (#2142) (31ac8a6)

v0.29.0

0.29.0 (2026-08-19)

Features

• aimlapi: passwordless onboarding and resumable card top-up (3/3) (#2032) (fb9102c)
• cost: support exact custom model pricing (#2131) (09eba26)
• gateway: add Concentrate AI provider with dynamic model discovery (#2140) (084bc53)
• merge knowledge graph + conversation arc into memdir (#1811) (c461a03)
• partners: add ApiSmart, refresh Novita AI logo (#2121) (575b407)
• partners: add Concentrate and Exa to partner roster (#2141) (6e35903)
• xai: add Grok 4.6/4.5 to catalog, xAI provider, and gateways (#2117) (7cae408)
• zai: expand Coding Plan catalog support (#2127) (ea65516)

Bug Fixes

• api: resolve swarm-field tool names by own-property (#2123) (f553d08)
• bg: preserve detached session terminal outcomes (#2133) (108a413)
• code-reviewer: require inline diff input and preserve read-only search in embedded-search builds (#2102) (645d596)
• extend Ling 3.0 Tiny :free availability window to Aug 17 (ee64d80)
• Ling 3.0 Tiny :free window back to Aug 13 (official promo end) (6277bfa)
• mcp: paginate discovery list operations (#2132) (e9e6beb)
• web: link release notes to GitHub (#2114) (c40d663)
• websearch: reject non-positive WEB_CUSTOM env overrides (#2124) (6c7a12b)

v0.28.0

0.28.0 (2026-08-11)

Features

• cli: add --yolo alias for --dangerously-skip-permissions (#2097) (95eeb0b)
• codex: move codexplan default to GPT-5.6 Sol (#2051) (9540946)
• integrations: add ApiSmart OpenAI-compatible gateway provider (#2109) (ff91642)
• opengateway: add Ling 3.0 Tiny :free — Day-0 launch, free until August 13 (#2112) (7b03ad1)
• opengateway: free retirement — paid Ling id, dual Nemotron, Macaron Venti (#2108) (54b9cd8)

Bug Fixes

• bash: convert BRE interval braces when previewing sed edits (#1955) (eb1de5b)
• cost: guard model-cost lookup against prototype-member model ids (#2064) (c327805)
• input: preserve text in DEL-coalesced chunks (#2091) (1bf8076)
• mcp: serialize OAuth and XAA refresh across processes (#2093) (6465a51)
• model-picker: eliminate O(n²) catalog rebuild lag in /model (#2078) (248424f)
• query: use monotonic watchdog deadlines (#2110) (16e332e)
• release: synchronize web changelog entries (#2088) (7743cf2)
• repl: make local interactive max-turns configurable (#2086) (b0cbfe1)
• repomap: resolve file language by real extension, own-property only (#2100) (41d2f3b)
• session: make transcript replacements crash-safe (#2094) (d834904)
• web: add v0.27.0 changelog entry and clarify release-data ownership (#2075) (63fda83)

Performance Improvements

• cli: enable Node module compile cache (#2092) (d427a4b)

v0.27.0

0.27.0 (2026-07-30)

Features

• auth: opt-in loopback proxy hosts that keep subscription (OAuth) auth (#2050) (3925f27)
• integrations: add Ling 3.0 Flash free to the Opengateway catalog (#2057) (3c5856a)
• opengateway: add Macaron V1 Tall to the gateway catalog (#2067) (e636f7d)
• ui: single-row centered startup logo with ANSI Shadow wordmark (#2053) (12994f2)
• web: replace favicon/logo with Ember Block O brand mark (#2065) (56a9201)
• web: v0.26 refresh — buddy page, changelog, partners, provider … (#2060) (0f76b54)

Bug Fixes

• agents: allow subagents from multi-repo parent sessions (#2063) (8df37c7)
• minimax: mark MiniMax-M2.7 as text-only input (#2068) (5cac15c)
• query: do not trip tool-failure guard on parallel same-turn failures (#2048) (9d5b77d)
• sdk: report a permission timeout as a timeout (#2028) (580a6b1)
• stats: stop over-counting totalDays by one (#1953) (83440a6)
• ui: keep SpinnerModeGlyph visible inside status parens (#2047) (10a9190)
• web: make web/ build standalone — stop importing the repo-root p… (#2061) (c2030bb)

v0.26.0

0.26.0 (2026-07-27)

Features

• aimlapi: add passwordless client methods and response-shape guards (2/N) (#2020) (022f057)

Bug Fixes

• apply ultrathink effort to provider requests (#2046) (53d3cf7)
• bridge: truncate derived session titles on grapheme boundaries (#1982) (a6b3d7a)
• cache-probe: omit unsupported cache request fields (#2044) (a652013)
• commands: insert slash-command argument text literally, not as regex refs (#1966) (62d15d4)
• fs: tolerate EPERM from mkdir on Windows drive roots (#2026) (5f6c608)
• memory: match nested directories on path boundaries, not name prefixes (#1974) (0ff1d1c)
• query: keep long-running tools active (#2022) (c23b6e1)
• ui: show streaming token count immediately (#2030) (01a01fb)

v0.25.0

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_stream rejection 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)

Performance Improvements

• tools: preserve UTF-8-safe head and tail in persisted previews (#1960) (507ba4b)