Commit Graph

  • de9673ac3f tokenizer: add byte fallback for SentencePiece BPE encoding (#15232) v0.20.0 Daniel Hiltgen 2026-04-02 13:04:45 -07:00
  • 96b202d34b Add support for gemma4 (#15214) Daniel Hiltgen 2026-04-02 11:33:33 -07:00
  • 79865e6c5a app: use the same client for inference and other requests (#15204) Devon Rifkin 2026-04-02 11:07:50 -07:00
  • 5ab10d347a app: add launch page for a simple way to launch integrations (#15182) Parth Sareen 2026-04-02 10:31:19 -07:00
  • 7fdd90f597 convert: support new Gemma4 audio_tower tensor naming (#15221) v0.20.0-rc1 Daniel Hiltgen 2026-04-02 09:30:05 -07:00
  • 8d846fdbc0 Merge pull request #42 from ollama/jmorganca/gemma4-ggml-improvements v0.20.0-rc0 jmorganca/gemma4-audio-replacements Daniel Hiltgen 2026-04-02 07:16:06 -07:00
  • f3536a356e use 4096 kvcache.NewSWAMemCache jmorganca 2026-04-02 01:56:36 -07:00
  • c89280fb0c Format Gemma4 MoE block field alignment jmorganca 2026-04-02 01:43:14 -07:00
  • eb5434d7fb Remove redundant comments in gemma4 vision model jmorganca 2026-04-02 01:33:46 -07:00
  • 2b949a11d9 convert: align gemma4 audio tensor renames with llama.cpp jmorganca 2026-04-02 01:28:10 -07:00
  • 6b013002fc gemma4: initialize clamps after backend load jmorganca 2026-04-02 01:12:05 -07:00
  • 5e622289c5 gemma4: use full SWA memory for better cache reuse jmorganca 2026-04-02 00:40:32 -07:00
  • 9c8bcecdb2 cmd: simplify audio input to dropped file attachments jmorganca 2026-04-02 00:40:27 -07:00
  • 1cbe7950d6 gemma4: fix MoE fused gate_up split and multiline tool-call arg parsing jmorganca 2026-04-02 00:01:44 -07:00
  • 95073400fc gemma4: rewrite renderer to match HF Jinja2 template exactly Daniel Hiltgen 2026-04-01 22:35:26 -07:00
  • c29932c631 gemma4: add per_expert_scale to MoE router and fix moe_intermediate_size config Daniel Hiltgen 2026-04-01 17:34:20 -07:00
  • 1ce101c9a0 gemma4: update converter for new weight drop naming Daniel Hiltgen 2026-04-01 15:44:47 -07:00
  • 5a7928ed38 gemma4: add OpenAI audio transcription API and input_audio support Daniel Hiltgen 2026-04-01 15:34:05 -07:00
  • 7fdc051091 gemma4: fix renderer to emit BOS token Daniel Hiltgen 2026-04-01 15:33:45 -07:00
  • 5bad871241 gemma4: add parser, renderer, and integration test plumbing Daniel Hiltgen 2026-04-01 15:31:08 -07:00
  • 82437d620a gemma4: add transcribe command (ollama transcribe MODEL) Daniel Hiltgen 2026-04-01 08:38:54 -07:00
  • 570c53859d gemma4: add audio input support for run command Daniel Hiltgen 2026-04-01 08:33:45 -07:00
  • ebd70f73b7 gemma4: add OpenAI audio API support and capability detection Daniel Hiltgen 2026-03-31 22:21:23 -07:00
  • eb5df80733 integration: add gemma4 audio tests including OpenAI API coverage Daniel Hiltgen 2026-03-30 17:48:40 -07:00
  • 356c0b8e34 gemma4: add audio support with USM conformer encoder Daniel Hiltgen 2026-03-31 21:54:08 -07:00
  • ea3c6a3cbe gemma4: add Gemma 4 GGML model support Daniel Hiltgen 2026-03-30 17:38:56 -07:00
  • f6b69f3f28 integration: improve vision test robustness and add thinking tests Daniel Hiltgen 2026-03-30 14:58:08 -07:00
  • e38b606e8b bench: add prompt calibration, context size flag, and NumCtx reporting Daniel Hiltgen 2026-03-30 14:21:11 -07:00
  • a8292dd85f launch: replace deprecated OPENAI_BASE_URL with config.toml profile for codex (#15041) Eva H 2026-04-01 08:43:23 -07:00
  • cb0033598e tokenizer: add SentencePiece-style BPE support (#15162) Daniel Hiltgen 2026-03-31 17:00:36 -07:00
  • 4d14b0ff92 mlx: respect tokenizer add_bos_token setting in pipeline (#15185) Daniel Hiltgen 2026-03-31 16:46:30 -07:00
  • d9cb70c270 docs: update pi docs (#15152) Parth Sareen 2026-03-31 16:37:55 -07:00
  • 31f968fe1f cmd: set OpenCode default model in config (#15127) v0.19.0-rc2 v0.19.0 Jeffrey Morgan 2026-03-29 12:11:36 -07:00
  • b7bda92d52 model: add qwen3-next compatibility for legacy ssm_in projections (#15133) Jeffrey Morgan 2026-03-29 11:50:47 -07:00
  • 8e54823fd3 revert context length warnings change (#15121) Parth Sareen 2026-03-28 16:43:59 -07:00
  • 7c8da5679e launch: improve multi-select for already added models (#15113) Parth Sareen 2026-03-28 13:44:40 -07:00
  • 6214103e66 launch: auto-install pi and manage web-search lifecycle (#15118) Parth Sareen 2026-03-28 13:06:20 -07:00
  • 9e7cb9697e mlx: fix vision capability + min version (#15106) v0.19.0-rc1 Patrick Devine 2026-03-27 17:09:28 -07:00
  • 3824e380a8 server: preserve raw manifest bytes during pull (#15104) Bruce MacDonald 2026-03-27 15:42:31 -07:00
  • c9b2dcfc52 anthropic: fix empty inputs in content blocks (#15105) Devon Rifkin 2026-03-27 15:41:27 -07:00
  • b00bd1dfd4 launch: skip context length warning for MLX models and show model name (#15102) Parth Sareen 2026-03-27 15:01:33 -07:00
  • 522d553b03 server: preserve raw manifest bytes during pull fix-manifest-digest-on-pull Bruce MacDonald 2026-03-27 14:30:51 -07:00
  • ac83ac20c4 anthropic: fix KV cache reuse degraded by tool call argument reordering Jesse Gross 2026-03-09 16:24:57 -07:00
  • e7ccc129ea app: fix false "out of date" model warnings (#15101) Bruce MacDonald 2026-03-27 14:15:10 -07:00
  • 69ed0c2729 parsers: qwen3.5 streaming tool-call parsing and add regression test (#15098) Jeffrey Morgan 2026-03-27 14:04:14 -07:00
  • b720a264a6 app: fix false "out of date" model warnings Bruce MacDonald 2026-03-27 13:35:18 -07:00
  • 1cefa749aa model/parsers: close think block if tool block starts in Qwen3.5 (#15022) Alfredo Matas 2026-03-27 19:28:34 +01:00
  • aec2fef95d ci: harden cuda include path handling (#15093) v0.19.0-rc0 Daniel Hiltgen 2026-03-27 07:57:07 -07:00
  • 366625a831 launch: warn when server context length is below 64k for local models (#15044) v0.18.4-rc1 Eva H 2026-03-27 00:15:53 -07:00
  • 516ebd8548 ci: include mlx jit headers on linux (#15083) Daniel Hiltgen 2026-03-26 23:10:07 -07:00
  • f567abc63f tui: update chat title (#15082) Parth Sareen 2026-03-26 18:06:53 -07:00
  • 7a2306087b wip hoyyeva/vscode-improve Eva Ho 2026-03-26 19:55:13 -04:00
  • 1adfc27f04 launch/vscode: prefer known vs code paths over code on PATH (#15073) Eva H 2026-03-26 15:06:28 -07:00
  • 4a2b9f9dbc launch: hide cline integration (#15080) Parth Sareen 2026-03-26 14:33:43 -07:00
  • e46b67a6cc launch: hide vs code (#15076) v0.18.4-rc0 Parth Sareen 2026-03-26 13:52:50 -07:00
  • c000afe76c doc: update vscode doc (#15064) Eva H 2026-03-26 13:45:48 -07:00
  • 9d7b18f81e mlxrunner: combine setStateRaw and setStateDetached into setState Jesse Gross 2026-03-25 15:24:57 -07:00
  • 4f5999fd3f mlxrunner: schedule periodic snapshots during prefill Jesse Gross 2026-03-24 16:55:49 -07:00
  • ac5f0dbb6a mlxrunner: improve eviction and LRU tracking Jesse Gross 2026-03-24 16:55:27 -07:00
  • 8b8bcf0952 launch: set default model as active selection in vscode copilot chat Eva Ho 2026-03-26 14:50:13 -04:00
  • d1151e18a1 mlx: fix KV cache snapshot memory leak Jesse Gross 2026-03-25 10:47:59 -07:00
  • ebbce136c7 ggml: force flash attention off for grok rick 2026-03-25 02:51:46 +01:00
  • 26b9f53f8e api/show: overwrite basename for copilot chat (#15062) v0.18.3-rc2 v0.18.3 Devon Rifkin 2026-03-25 14:02:22 -07:00
  • 7575438366 cmd: ollama launch vscode (#15060) Eva H 2026-03-25 13:37:02 -07:00
  • 7d7c90d702 tui: add left arrow back navigation in model selector (#14940) Eva H 2026-03-25 11:53:48 -07:00
  • 4fda69809a ci: fix windows cgo compiler error (#15046) v0.18.3-rc1 Daniel Hiltgen 2026-03-24 16:45:36 -07:00
  • e4c48e80f5 install: wait for server to be ready after starting on macOS brucemacd/install-server-wait Bruce MacDonald 2026-03-24 15:33:31 -07:00
  • 04b337f58f install: do not remove existing install until after download brucemacd/download-before-remove Bruce MacDonald 2026-02-20 15:04:28 -08:00
  • c9b5da6b0c integration: improve ability to test individual models (#14948) Daniel Hiltgen 2026-03-24 14:28:23 -07:00
  • de5cb7311f mlx: add mxfp4/mxfp8/nvfp4 importing (#15015) v0.18.3-rc0 Patrick Devine 2026-03-24 13:45:44 -07:00
  • 00af64a0ae docs: add more info parth/update-claude-docs ParthSareen 2026-03-24 13:00:18 -07:00
  • 21f0db0d37 docs: update claude code docs heading ParthSareen 2026-03-24 12:17:03 -07:00
  • 40f56cf543 anthropic: handle images in tool_result content blocks parth-anthropic-reference-images-path ParthSareen 2026-03-24 11:43:48 -07:00
  • a60a79a5e1 install: use explicit path to launch Ollama after install brucemac/start-ap-install Bruce MacDonald 2026-03-24 11:07:24 -07:00
  • 95ee7fbd29 mlxrunner: panic on double unpin Jesse Gross 2026-03-20 16:10:19 -07:00
  • ec55536734 mlxrunner: show time since last used in cache dump tree Jesse Gross 2026-03-19 15:24:42 -07:00
  • 77491439c2 mlxrunner: support partial match on pure transformer caches Jesse Gross 2026-03-19 11:20:50 -07:00
  • b166b36cd2 docs: update Claude Code with Telegram guide (#15026) Parth Sareen 2026-03-23 16:31:21 -07:00
  • c2b0bb7a52 mlx: update as of 3/23 (#14789) Daniel Hiltgen 2026-03-23 11:28:44 -07:00
  • 6f26695eae mlx: update upstream mlx version pdevine/mlx-update Patrick Devine 2026-03-21 17:07:47 -07:00
  • 22c2bdbd8a docs: nemoclaw integration (#14962) Bruce MacDonald 2026-03-20 15:27:37 -07:00
  • 6df6d097d9 launch: skip openclaw gateway health check when no daemon install (#14984) Bruce MacDonald 2026-03-20 15:20:14 -07:00
  • d7c176ab91 llm, mlxrunner: fix done channel value consumed by first receiver Jesse Gross 2026-03-18 16:30:43 -07:00
  • 0ff7d724ff mlx: fix subprocess log deadlock Jesse Gross 2026-03-17 17:02:51 -07:00
  • 578c32e42e still more linter stuff pdevine/qwen35_vision Patrick Devine 2026-03-19 17:29:12 -07:00
  • a10d2625ca linters ftw Patrick Devine 2026-03-19 17:20:59 -07:00
  • b960d769ad more linter fixes Patrick Devine 2026-03-19 17:11:43 -07:00
  • 46cb7795e1 add ability to turn on debug request logging (#14106) Devon Rifkin 2026-03-19 17:08:17 -07:00
  • 455a6099d1 gofumpt the linter Patrick Devine 2026-03-19 16:52:35 -07:00
  • 7e6e8377eb mlx: qwen3.5 vision support Patrick Devine 2026-03-17 15:34:56 -07:00
  • 126d8db7f3 parsers: robust xml tool repair (#14961) Bruce MacDonald 2026-03-19 11:24:48 -07:00
  • 3f3a24b418 app: fix desktop app stuck loading when OLLAMA_HOST is an unspecified bind address (#14885) Eva H 2026-03-19 12:57:57 -04:00
  • 96e36c0d90 mlxrunner: share KV cache across conversations with common prefixes Jesse Gross 2026-03-05 15:45:36 -08:00
  • 6f8ddbb26b mlxrunner: fix Slice(0, 0) returning full dimension instead of empty Jesse Gross 2026-03-18 14:10:08 -07:00
  • f9a46b73da [WIP] api/show: fall back to local on cloud 404 drifkin/api-show-fallback Devon Rifkin 2026-03-18 15:56:15 -07:00
  • b5e7888414 cmd/launch: skip redundant config writes when model unchanged (#14941) Eva H 2026-03-18 17:36:52 -04:00
  • eab4d22269 docs: update claude code and openclaw for web search (#14922) Parth Sareen 2026-03-18 14:18:49 -07:00
  • 5759c2d2d2 launch: fix openclaw not picking up newly selected model (#14943) v0.18.2-rc1 v0.18.2 Bruce MacDonald 2026-03-18 13:20:10 -07:00
  • 42b1c2642b docs: update minimax-m2.5 references to m2.7 (#14942) Bruce MacDonald 2026-03-18 12:59:28 -07:00
  • 727d69ddf3 tui: fix signin on headless Linux systems (#14627) Bruce MacDonald 2026-03-18 11:11:17 -07:00