← back to report

ollama/ollama

// all events — 2026-10-31

Type AI Score Description Actor Reason Date
commit 1.00 decision: allow empty state for vllm-d3 (#19173) Commit message contains explicit AI assi 2026-10-23
pr 1.00 server: give MLX loads the llama-server memory pre-flight Explicitly states that it was generated 2026-10-31
pr 1.00 llm: size the predicted KV cache per layer Explicitly states it was generated with 2026-10-27
pr 1.00 mlxrunner: abort generation when a token repeats past the li Explicitly states that Claude Code helpe 2026-10-29
pr 1.00 README: add Wavefunction to community integrations Explicitly states it was generated with 2026-10-30
pr 1.00 docs: replace deprecated zod-to-json-schema with Zod v4 nati Explicitly states the contribution was d 2026-03-09
pr 1.00 types/model: ignore null sampler values in generation_config PR body explicitly mentions AI collabora 2026-10-29
pr 1.00 server: handle NaN values in embedding responses Explicitly states the contribution was d 2026-03-09
pr 1.00 server: never evict a CPU-only runner to satisfy a GPU memor PR body explicitly mentions AI collabora 2026-10-25
pr 1.00 create: respect explicitly empty system prompts Explicitly states that AI assistance was 2026-10-13
pr 1.00 llm: don't count CPU-resident input tensors in the VRAM pred PR body explicitly mentions AI collabora 2026-10-29
pr 1.00 readme: add Autopilot to community integrations The description explicitly says it was g 2026-10-29
pr 1.00 server: report vision only for projectors with a vision enco Explicitly states it was generated with 2026-10-27
pr 1.00 app/wintray: open log folder correctly when the path contain Explicitly states it was generated with 2026-10-06
pr 1.00 README: add Paw Pulse to Observability & Monitoring PR body explicitly mentions AI collabora 2026-10-28
pr 1.00 mlx: add Kolibri 1 support This is explicitly an automated Codex re 2026-10-04
pr 1.00 server: keep interrupted downloads resumable across restarts Explicitly identifies the change as gene 2026-10-27
pr 1.00 llm: retry llama-server startup with eager CUDA module loadi Explicitly states it was generated with 2026-10-13
pr 0.62 fix: proper media ordering in Embedding for multimodal model Generic polished summary structure and r 2026-10-27
pr 0.28 mlx: add Windows support and CPU backend for MLX runner Highly polished and structured, but deta 2026-10-28
pr 0.15 Claude/quirky mccarthy yumum8 Very unusual title, but it provides too 2026-10-28
pr 0.12 README: add Clairvoyance to community integrations Short promotional copy may be copied mar 2026-10-29
pr 0.10 README: add self-host to community integrations Personal motivation, concrete features, 2026-10-28
commit 0.08 mlx: add Flint model support (#19181) Detailed domain-specific change summary 2026-10-21
pr 0.08 README: (Community Integration) Add BashBrunch Awkward phrasing and concrete project de 2026-10-30
pr 0.08 mlxrunner: support draft_num_predict=0 to disable drafting Technical and concise, with concrete beh 2026-10-29
pr 0.08 Add claude Docker tag to release pipeline Very terse change description, but no cl 2026-10-29
pr 0.08 proxy: harden Codex Guardian auto-review against namespaced Technical, specific wording with informa 2026-10-29
pr 0.08 docs: add Vauta to community integrations Specific integration details and natural 2026-10-28
pr 0.08 README: add Quock to Mobile community integrations Brief, specific integration request with 2026-10-28
pr 0.07 mlx: support concurrent generation Detailed benchmark and implementation sp 2026-10-22
pr 0.06 openai: preserve readable Responses input content Technical API-specific discussion with c 2026-10-21
commit 0.05 api: populate show renderer/parser fields and add template_s Concise, domain-specific technical summa 2026-10-18
commit 0.05 launch: suppress Claude Code connector warning for Ollama (# Brief technical commit with a normal co- 2026-10-09
commit 0.05 launch: avoid inherited service-tier warnings in Codex (#188 Terse, domain-specific commit-style titl 2026-10-08
commit 0.05 respect CLAUDE_CODE_AUTO_COMPACT_WINDOW user setting Brief configuration change phrased like 2026-10-08
pr 0.05 server: reject clips the model has no encoder for Contains concrete code paths, tests, and 2026-10-29
pr 0.05 model: add MiMo V2.6 model support Very brief model-support announcement pr 2026-10-30
pr 0.05 cmd: show help when running ollama without arguments Concise, specific CLI behavior and clean 2026-10-23
pr 0.05 llm: add Intel oneAPI (SYCL) GPU backend Detailed platform-specific build and pac 2026-08-08
pr 0.04 fix: increase model name component length limit from 80 to 1 Concise issue-focused changelog with con 2026-09-25
pr 0.04 mlxrunner: implement load progress reporting and stall detec Specific MLX loading details and informa 2026-08-18
pr 0.04 README: add RedAmon to community integrations Specific integration details and concise 2026-10-29
pr 0.04 server: prevent expiring runners from being resurrected Specific scheduler race details and hone 2026-07-27
pr 0.04 cmd: clear the session system prompt when loading a model Detailed regression scenarios, platform 2026-10-28
pr 0.04 envconfig: ignore whitespace around comma-separated list ent Concrete environment-variable behavior, 2026-10-28
pr 0.04 mlx: add video input support Highly domain-specific implementation de 2026-10-28
pr 0.04 mlx: add Nemotron 3.5 Bolt 74B-A9B support Uses specific implementation terminology 2026-10-22
commit 0.03 llm: LFM3 GGUF fixes and llama-server speech (#19344) Detailed domain-specific changelog with 2026-10-28
commit 0.03 dflash: add the DFlash2 draft model (#19149) Detailed domain-specific explanation use 2026-10-26
commit 0.03 Support GA Gradium Phonon checkpoints (#19227) Specific technical changes and issue ref 2026-10-22
commit 0.03 ci: switch to released WoA CTK (#19185) Informal wording, typo, and specific CUD 2026-10-21
commit 0.03 mlx: deduplicate CUDA runtime payloads (#17956) Specific CUDA packaging details and ters 2026-10-20
commit 0.03 mlx: preserve generation panics during cleanup (#18886) Specific panic and cache behavior with c 2026-10-13
commit 0.03 Add Kev decision scoring to MLX (#18976) Terse change summary with MLX-specific t 2026-10-13
commit 0.03 mlx: compact restored recurrent state after speculative roll Detailed but domain-specific technical c 2026-10-12
commit 0.03 server: reuse verified manifests and loaded models in infere Thorough engineering changelog with conc 2026-10-12
pr 0.03 macos: best-effort support for Intel Macs on macOS 12 (Monte Detailed domain-specific changes and cav 2026-10-30
pr 0.03 anthropic: keep all thinking blocks of an assistant message Specific bug reproduction, implementatio 2026-10-30
pr 0.03 openai: make compatibility response IDs unique Specific technical details and test resu 2026-10-30
pr 0.03 cmd/launch: use thinking metadata for OpenCode settings Concise implementation details and repos 2026-10-30
pr 0.03 openai: keep tool calls on assistant messages with empty arr Specific regression context and test com 2026-10-19
pr 0.03 Fix long System One requests for MLX decision models Terse, domain-specific summary with conc 2026-10-30
pr 0.03 openai: keep multipart images with user text Concrete bug reproduction, implementatio 2026-10-22
pr 0.03 api: bound thinking with a token budget, per request or per Detailed model-specific reasoning and im 2026-08-04
pr 0.03 app/ui: honor cancellation while waiting for server Specific audit follow-up, platform testi 2026-09-15
pr 0.03 server: bound direct URL resolution attempts Concrete timeout regression, fixture cha 2026-09-14
pr 0.03 mlx: sort long top-k rows in chunks Highly specific performance details and 2026-10-29
pr 0.03 llm: fix llama-server memory accounting for multi-model load Dense project-specific debugging details 2026-08-18
pr 0.03 openai: preserve assistant reasoning in content arrays Specific regressions, package tests, and 2026-10-29
pr 0.03 chat: preserve the most recent user message during truncatio Template-like headings contain concrete 2026-09-28
pr 0.03 fix(llm): request flash attention explicitly for architectur Highly specific debugging details, refer 2026-07-30
pr 0.03 server: accept the sha256- digest spelling when creating a b Uses precise implementation details, fai 2026-10-27
pr 0.03 cmd: drop the multiline draft when Ctrl+C interrupts it Informal bug reproduction and precise PT 2026-10-28
pr 0.03 middleware: forward errors that arrive after an OpenAI strea Specific bug behavior, issue references, 2026-10-12
pr 0.03 mlxrunner: continuous batching for OLLAMA_NUM_PARALLEL Template structure aside, the summary an 2026-07-22
pr 0.03 app/ui: preserve unsent draft message when switching chats Specific implementation details, test na 2026-08-30
pr 0.03 app/ui: keep idle chat streams alive and surface failures Detailed technical description with pers 2026-10-27
commit 0.02 mlx: sort long top-k rows in chunks (#19392) Concise technical explanation with speci 2026-10-30
commit 0.02 openai: preserve readable Responses input content (#19192) Terse, specific commit-style wording wit 2026-10-29
commit 0.02 ci: pin buildx to workaround deadlock (#19373) Terse, issue-specific commit message wit 2026-10-28
commit 0.02 llama.cpp: workaround Metal unary crash (#19372) Terse technical workaround and exact err 2026-10-28
commit 0.02 MLX, MLX-C: version bump - prompt speed up for gemma/flint/l Highly specific implementation details a 2026-10-26
commit 0.02 mlx: preserve small-image detail for compact Gemma 4 vision Terse, domain-specific summary with conc 2026-10-24
commit 0.02 Update gemma renderers (#19241) Detailed technical changelog with concre 2026-10-23
commit 0.02 flint: fix replies with thinking disabled (#19216) Informal, specific notes about Flint beh 2026-10-22
commit 0.02 mlxrunner: add Phonon text-to-speech support (#19166) Technical implementation details and a n 2026-10-22
commit 0.02 llama.cpp update (#19194) Terse update title with no AI-style pros 2026-10-22
commit 0.02 server: read loaded runner models without refMu (#19070) Includes concrete concurrency behavior, 2026-10-20
commit 0.02 mlx: add PPLX Decider v1.2 support (#19072) Terse, domain-specific implementation su 2026-10-17
commit 0.02 llama-server: let llama-server infer the draft speculative t Specific technical rationale and reposit 2026-10-17
commit 0.02 docs: add model grouping to cloud usage (#19069) Specific documentation changes and a nor 2026-10-16
commit 0.02 New model... (#19030) Highly technical changelog with concrete 2026-10-15
commit 0.02 Add decision model integration tests and benchmarks (#18746) Technical test details and project-speci 2026-10-13
commit 0.02 docs: update README to include cloud and privacy language Brief conventional documentation commit 2026-10-12
commit 0.02 standardizing language referring to cloud services Short editorial change description, with 2026-10-08
pr 0.02 app: make ChatGPT disconnect failures retryable Specific implementation summary accompan 2026-10-27
pr 0.02 README: add schemagate to community integrations Brief, specific integration description 2026-10-30
pr 0.02 Add Lirnyk to community integrations Specific product behavior and repository 2026-10-30
pr 0.02 parser: omit values that a Modelfile cannot represent when e Highly specific parser behavior, tests, 2026-10-28
pr 0.02 gemma4: recover a finished tool call that is missing its clo Specific captured failure, parser behavi 2026-08-04
pr 0.02 gemma4: do not answer with a channel name the parser was cut Highly specific streaming-parser behavio 2026-08-08
pr 0.02 server: do not hand over a tool call the model did not finis Concrete agent failure, schema details, 2026-08-04
pr 0.02 qwen3coder: tolerate a dropped closing tag, and stop rewriti Highly specific parser behavior, reprodu 2026-08-21
pr 0.02 llm: stop guessing that a repetitive payload is a runaway Detailed domain-specific reproduction an 2026-08-04
pr 0.02 parsers: drop an unmatched thinking close tag instead of lea Contains concrete parser states, observe 2026-09-07
pr 0.02 llm: send assistant thinking to the chat template Specific API fields, template behavior, 2026-09-07
pr 0.02 fix(llm): mirror GraphSize KV accounting in PredictServerVRA Deep metadata and memory-accounting refe 2026-08-07
pr 0.02 report per-request speculative decoding metrics Specific technical detail and informal p 2026-10-28
pr 0.02 openai: return empty embedding data as an array Detailed Go behavior, issue references, 2026-10-29
pr 0.02 compatmigrate: fix Statfs space math on FreeBSD FreeBSD type details and targeted build 2026-10-07
pr 0.02 openai: reject invalid content types alongside tool calls Detailed edge cases, resource limits, an 2026-10-29
pr 0.02 ci: pin buildx to workaround deadlock Concise technical PR text with concrete 2026-10-28
pr 0.02 llama.cpp: workaround Metal unary crash Specific error text and terse workaround 2026-10-28
pr 0.02 llama.cpp: version bump b11963 Terse version-bump title with no AI-styl 2026-10-28
pr 0.02 transfer: allow public IPv4 destinations through NAT64 Detailed RFC-specific behavior and concr 2026-10-28
pr 0.02 llm: LFM3 GGUF fixes and llama-server speech Technical, implementation-specific chang 2026-10-27
pr 0.02 Add AITerm to Community Integrations (Terminal & CLI) Concise, domain-specific integration des 2026-10-27
pr 0.02 discover: use the newest library version that supports all G Highly specific debugging details and do 2026-10-27
pr 0.02 anthropic: stream thinking after text or tool_use as a new b Detailed implementation and regression c 2026-10-18
commit 0.01 cmd: show help when running ollama without arguments (#19248 Minimal, direct commit message with no A 2026-10-28
commit 0.01 llama.cpp: version bump b11963 (#19360) Simple version-bump commit message. 2026-10-28
commit 0.01 Fix kimi cli tool-only assistant turns (#19312) Terse, informal commit bullets with spec 2026-10-27
commit 0.01 voice: add voice catalog for Phonon (#19242) Terse, domain-specific commit summary wi 2026-10-23
commit 0.01 mlxrunner: spell out symbols, numbers and links for Phonon ( Terse, technical changelog entries with 2026-10-23
commit 0.01 Make sure all launchers know about cloud model metadata (#19 Terse, repository-specific wording is ch 2026-10-22
commit 0.01 openai: preserve failures in Responses streams (#19225) Concise technical commit title with a sp 2026-10-22
commit 0.01 anthropic: reuse resolved thinking for web search (#19098) Brief domain-specific wording does not s 2026-10-22
commit 0.01 build: update jsonparser to v1.1.2 (#19220) Short dependency-update title is typical 2026-10-22
commit 0.01 launch: use cloud model context in Claude Code (#19119) Concise, project-specific commit title w 2026-10-22
commit 0.01 build: update Go toolchain and dependencies (#19200) Standard build-maintenance wording is te 2026-10-22
commit 0.01 mlx: add d3 decision model support (#19196) Brief technical title and benchmark note 2026-10-22
commit 0.01 anthropic: surface errors after streaming starts (#19188) Terse, issue-specific title with no AI-s 2026-10-22
commit 0.01 avoid scary claude warning about auto compact window being t Terse, issue-specific wording strongly s 2026-10-19
commit 0.01 server: cancel remote inference when clients disconnect (#19 Concise technical commit title with a sp 2026-10-19
commit 0.01 openai: accept foreign compaction in Responses history (#190 Concise, domain-specific commit title wi 2026-10-19
commit 0.01 openai: allow Responses effort on non-thinking models (#1901 Brief, natural commit-style description 2026-10-17
commit 0.01 transfer: support proxies on registry redirect targets (#190 Terse, domain-specific implementation su 2026-10-16
commit 0.01 win: prevent orphaned runners (#18500) Detailed Windows implementation text use 2026-10-15
commit 0.01 use cloudPassthroughMiddleware Terse, repository-specific wording is ch 2026-10-14
commit 0.01 revert docs changes for now Brief imperative commit message with no 2026-10-14
commit 0.01 MLX: version bump (#18993) Conventional version-bump message with s 2026-10-14
commit 0.01 openai: accept custom tool calls in Responses history (#1891 Specific commit title and standard co-au 2026-10-14
commit 0.01 docs: Merge pull request #18981 from ollama/cloud_recommende Terse merge commit with repository-speci 2026-10-14
commit 0.01 mlx: port the Metal custom kernels to CUDA (#18657) Detailed, domain-specific CUDA and MLX i 2026-10-13
commit 0.01 launch: allow usage credits without upgrading (#18985) Concise conventional commit message with 2026-10-13
commit 0.01 Update README for improved privacy wording Brief, natural documentation update summ 2026-10-13
commit 0.01 decision: fix JSON state prompts for MLX decision models (#1 Specific implementation references and c 2026-10-13
commit 0.01 cli: add decision model support to `ollama run` (#18962) Terse commit message naming the affected 2026-10-13
commit 0.01 docs: changing model recs to Models page link, updating refe Brief, slightly awkward wording typical 2026-10-13
commit 0.01 gemma4: use MLX SDPA for CUDA prefill with wide head dims (# Specific MLX, CUDA, SDPA, and head-dimen 2026-10-13
commit 0.01 ci: harden download steps with retries (#18883) Informal concise wording with a concrete 2026-10-13
commit 0.01 server: fix multimodal embedding load with small num_ctx (#1 Terse, domain-specific commit title; no 2026-10-13
commit 0.01 Introduce clef and clef flash on cloud Terse feature description is typical of 2026-10-12
commit 0.01 docs: Merge pull request #18963 from ollama/security.md_upda Standard merge commit message with repos 2026-10-12
commit 0.01 docs: update SECURITY.md Brief, specific documentation change wit 2026-10-12
commit 0.01 Revert "docs: restructuring" (#18960) Conventional terse revert commit message 2026-10-12
commit 0.01 MLX, MLX-C: version bump (#18956) Concise technical version-bump commit ti 2026-10-12
commit 0.01 launch: populate Codex model picker with installed Ollama mo Specific feature description using proje 2026-10-12
pr 0.01 Configure the systemd service via a separate file. Brief, informal description with a speci 2024-07-10
pr 0.01 Add AutoTrust decision model to MLX Short, concrete feature description with 2026-10-30
pr 0.01 x/mlxrunner/mlx: link against libdl on linux Concise, domain-specific linker diagnosi 2026-08-04
pr 0.01 qwen3.5: a tool call can open before the thinking channel is Includes precise parser state behavior a 2026-09-24
pr 0.01 llama: end a spent reasoning budget at a line, not mid-word Specific compatibility-patch lifecycle a 2026-09-03
pr 0.01 server: reload when two tags share a blob but need different Specific scheduler internals, reproducti 2026-09-07
pr 0.01 README: add Amical to community integrations Informal integration note includes speci 2026-10-29
pr 0.01 Add kde-ollama-gui to the AI tools list Short, direct title without AI-style boi 2026-10-28
pr 0.01 README: add Voltip to community integrations Includes specific product details and an 2026-10-28
pr 0.01 app: fix Go doc comments that name nonexistent methods Concise, concrete code references and a 2026-09-07
pr 0.01 docs: fix broken links and punctuation in README and import Brief, repository-specific link and punc 2026-09-04
commit 0.00 feat(launch): use model output limits for Claude Code (#1941 Terse, conventional commit message with 2026-10-30
commit 0.00 app: make ChatGPT disconnect failures retryable (#19336) Terse, domain-specific commit message; n 2026-10-30
commit 0.00 mlx: add LFM3 and d1 models (#19333) Terse, specific model-support title with 2026-10-27
commit 0.00 cmd/launch: fix OpenCode v2 launch and config parsing (#1931 Concise commit message with a normal hum 2026-10-27
commit 0.00 llama.cpp: version bump b11901 (#19311) A minimal version-bump message with no A 2026-10-26
commit 0.00 llama.cpp: version bump b11845 (#19250) Minimal conventional version-bump commit 2026-10-23
commit 0.00 launch: use allocated context for OpenCode (#19193) Concise, repository-specific commit mess 2026-10-23
commit 0.00 qwen3coder: preserve literal function closing tags in parame Concise technical commit message naming 2026-10-23
commit 0.00 llama.cpp: version bump b11738 (#19143) Terse version-bump message with no AI-st 2026-10-20
commit 0.00 openai: finish reasoning before client tool calls (#19146) Terse, domain-specific change summary. 2026-10-20
commit 0.00 test A minimal commit message is strongly con 2026-10-16
commit 0.00 use identifiers for --ask mode question keys Terse, specific commit message with no A 2026-10-16
commit 0.00 cleanup Terse conventional commit message with n 2026-10-15
commit 0.00 gguf: add the q2_0 tensor and file types (#18967) Specific technical change and issue refe 2026-10-15
commit 0.00 be more precise with wording Brief informal wording is characteristic 2026-10-15
commit 0.00 clef cloud launch Terse project-specific launch note with 2026-10-15
commit 0.00 llama.cpp: version bump b11669 (#19032) Conventional version-bump commit with a 2026-10-15
commit 0.00 openai: clarify Responses authentication errors (#19004) Concise conventional commit title with a 2026-10-15
commit 0.00 proxy: pass through ChatGPT upstream access errors (#19007) Concise, domain-specific commit message 2026-10-15
commit 0.00 server: stop HTML-escaping cloud proxy request bodies (#1900 Concise technical commit title describin 2026-10-15
commit 0.00 app: clarify cloud payment errors Short, domain-specific commit summary wi 2026-09-28
commit 0.00 typo in docs Extremely terse and informal commit mess 2026-10-12
commit 0.00 mlx: read candidate logits in float32 for quantized heads Concise technical commit title with doma 2026-10-12
commit 0.00 mlx: clip long laya states instead of rejecting them Terse, domain-specific commit summary wi 2026-10-12
commit 0.00 llm: support multimodal embeddings through llama-server Concise technical wording and project-sp 2026-10-10
commit 0.00 llama: update to b11521 and use native decision support Specific version reference and technical 2026-10-09
commit 0.00 decision: align request validation and option descriptions Brief, technical commit title with concr 2026-10-10
commit 0.00 server: skip local model compatibility migration (#18908) Terse, specific commit message with no A 2026-10-09
pr 0.00 launch: use model output limits for Claude Code Informal, terse wording and specific pro 2026-10-30
pr 0.00 docs: fix manual tar.zst extraction to use zstd (#17860) Contains terse documentation details and 2026-08-22