| pr |
1.00 |
model/parsers: preserve Unicode in olmo3 tool arguments |
|
PR body explicitly mentions AI collabora |
2026-10-09 |
| pr |
1.00 |
README: add Creatos to community integrations |
|
Explicitly states it was generated with |
2026-10-09 |
| pr |
1.00 |
README: add AI Character Engine to community integrations |
|
Explicitly discloses that the project wa |
2026-10-09 |
| pr |
1.00 |
mlx: add Kolibri 1 support |
|
The item is explicitly an automated Code |
2026-10-04 |
| pr |
1.00 |
openai: preserve retained reasoning when replaying compactio |
|
PR body explicitly mentions AI collabora |
2026-10-09 |
| pr |
1.00 |
feat: Add/update AGENTS.md for AI agent guidance |
|
Explicitly identifies the content as gen |
2026-10-09 |
| pr |
1.00 |
docs: describe multimodal embedding inputs in OpenAPI |
|
Explicitly states the change was prepare |
2026-10-08 |
| pr |
1.00 |
responses: preserve namespace tool identity |
|
The description explicitly states that C |
2026-05-22 |
| pr |
1.00 |
types/model: ignore null generation defaults |
|
Explicitly states that AI assistance was |
2026-10-08 |
| pr |
1.00 |
openai: close text items before streaming tool calls |
|
PR body explicitly mentions AI collabora |
2026-10-05 |
| pr |
0.95 |
fix: upgrade seroval to 1.6.2 (CVE-2026-104846) |
|
Explicitly attributed to an automated se |
2026-10-09 |
| pr |
0.38 |
cli: adaptive Simplified Chinese localization (zh-CN, Englis |
|
Very polished, exhaustive presentation h |
2026-09-27 |
| pr |
0.35 |
README: add Vessel to community integrations |
|
Polished promotional wording, but no def |
2026-10-09 |
| pr |
0.30 |
README: Restore handy-ollama Chinese tutorial link in READM |
|
Structured and polished, but mostly conv |
2026-10-09 |
| pr |
0.28 |
server: add per-request logging of the full request and resp |
|
Polished explanatory structure and persu |
2026-10-08 |
| pr |
0.22 |
server: reject invalid embeddings before returning success |
|
Highly polished validation summary, but |
2026-09-12 |
| pr |
0.20 |
server: preserve imported GGUF blobs after validation |
|
Standardized and formal wording is mildl |
2026-09-12 |
| pr |
0.12 |
README: Re-add handy-ollama to community integrations |
|
Specific restoration details, with norma |
2026-03-26 |
| pr |
0.12 |
server: do not hand over a tool call the model did not finis |
|
Specific debugging details and informal |
2026-08-04 |
| pr |
0.12 |
app: skip macOS startup from Trash |
|
Detailed macOS implementation and test r |
2026-10-09 |
| pr |
0.12 |
Add AgentBridge to Frameworks & Agents community integration |
|
Specific integration context is present, |
2026-09-28 |
| pr |
0.12 |
Remove llama.cpp compatibility patches and migrate legacy GG |
|
Compact, domain-specific migration summa |
2026-10-08 |
| pr |
0.10 |
openai: accept custom tool calls in Responses history |
|
Highly domain-specific change summary wi |
2026-10-09 |
| pr |
0.10 |
fix(parser): treat unmatched FROM globs as not-exist on Wind |
|
Detailed domain-specific explanation and |
2026-09-25 |
| pr |
0.10 |
model/renderers: close the empty thought after a tool respon |
|
Unusually detailed renderer references o |
2026-10-08 |
| pr |
0.10 |
launch: fill cloud model metadata gaps from show on config w |
|
Polished but highly specific behavior, f |
2026-08-17 |
| commit |
0.08 |
server: hide duplicate and downgrade guards from list (#1887 |
|
Specific llama.cpp, GGUF, manifest, and |
2026-10-08 |
| commit |
0.08 |
model: add multimodal embeddings (#18820) |
|
Specific architecture and API details su |
2026-10-06 |
| pr |
0.08 |
Fix windows unmatched wildcard |
|
Technical, issue-focused description wit |
2026-09-21 |
| pr |
0.08 |
chat: always preserve the most recent user message during tr |
|
Detailed, repository-specific diagnosis |
2026-08-20 |
| pr |
0.08 |
launch: include installed models in Codex /model picker |
|
Concrete Codex behavior and model metada |
2026-10-08 |
| pr |
0.08 |
readme: add Golem to community frameworks & agents |
|
Concise, domain-specific integration des |
2026-10-06 |
| pr |
0.08 |
llm: recognize Harmony header whitespace in thinking grammar |
|
Technical specificity and nuanced scope |
2026-10-07 |
| pr |
0.07 |
mlx: preserve generation panics during cleanup |
|
Specific panic, cleanup, cache, and regr |
2026-10-08 |
| pr |
0.07 |
fix(llm): only feed content-carrying events to the repeat de |
|
Detailed, source-backed debugging narrat |
2026-07-23 |
| pr |
0.06 |
server: skip pull requires check for builds without a releas |
|
Specific failure analysis, toolchain det |
2026-10-08 |
| pr |
0.06 |
compatmigrate: fix Statfs space math on FreeBSD |
|
Specific FreeBSD type details and target |
2026-10-07 |
| pr |
0.06 |
server: proxy cloud usage and balance APIs |
|
Concise changelog with concrete API and |
2026-10-07 |
| 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 |
| commit |
0.05 |
mlx: match publisher tokenizer semantics (#18779) |
|
Detailed tokenizer terminology and focus |
2026-10-04 |
| pr |
0.05 |
docs: restructuring |
|
Brief, conventional documentation change |
2026-10-06 |
| pr |
0.05 |
Bug fix : app/webview: define NOMINMAX before including wind |
|
Specific compiler details and reproducti |
2026-09-26 |
| pr |
0.05 |
api: bound thinking with a token budget, per request or per |
|
Though structured, it contains highly sp |
2026-08-04 |
| pr |
0.05 |
x/transfer: bound authentication retries |
|
Specific Go retry behavior and terse tec |
2026-09-15 |
| pr |
0.05 |
llm: report transport errors instead of JSON errors on trunc |
|
Detailed reproduction, error behavior, t |
2026-10-07 |
| commit |
0.04 |
Set CLAUDE_CODE_MAX_CONTEXT_TOKENS to whatever the models co |
|
Concise technical commit message without |
2026-10-07 |
| commit |
0.04 |
Reduce model lookup and MLX decision request overhead (#1880 |
|
Specific implementation details and revi |
2026-10-06 |
| commit |
0.04 |
nemotron_h: add MLX vision support (#17714) |
|
Detailed domain-specific changes and rev |
2026-09-17 |
| commit |
0.04 |
ci: harden release artifact uploads (#18516) |
|
Specific release sequencing, retry timin |
2026-09-17 |
| commit |
0.04 |
create: add server-side MLX imports and drop GGUF conversion |
|
Detailed implementation summary with con |
2026-09-15 |
| pr |
0.04 |
thinking: split Mistral [THINK] reasoning tags instead of le |
|
Highly domain-specific reproduction and |
2026-10-08 |
| pr |
0.04 |
parsers: drop an unmatched thinking close tag instead of lea |
|
Highly specific debugging context and in |
2026-09-07 |
| pr |
0.04 |
llm: send assistant thinking to the chat template |
|
Precise code-path analysis and nuanced t |
2026-09-07 |
| pr |
0.04 |
llama: end a spent reasoning budget at a line, not mid-word |
|
Specific compatibility, patch-order, and |
2026-09-03 |
| pr |
0.04 |
mlx: compact restored recurrent state after speculative roll |
|
Specific MLX details and measured result |
2026-10-05 |
| pr |
0.04 |
model/parsers: flush qwen3.5 tool parser on an empty final c |
|
Detailed parser behavior and test-case s |
2026-10-08 |
| pr |
0.04 |
model: reject empty scheme prefix in names |
|
Fuzzing details, invariants, and exact t |
2026-10-07 |
| pr |
0.04 |
docs: fix broken download links in app README |
|
Template-like structure is filled with c |
2026-09-04 |
| pr |
0.04 |
docs: fix 6 dead links in README community integrations list |
|
Specific URLs, projects, and HTTP verifi |
2026-10-06 |
| pr |
0.04 |
docs: fix min_p formatting in modelfile |
|
Concise, repository-specific documentati |
2026-10-07 |
| pr |
0.04 |
fix(wintray): open logs dir correctly when path contains spa |
|
Concrete Windows root cause, limitation, |
2026-10-07 |
| commit |
0.03 |
mlx: mitigate high latency after GPU idle (#18807) |
|
Technical description references a concr |
2026-10-05 |
| commit |
0.03 |
mlx: System one support (#18701) |
|
Detailed commit history and explicit co- |
2026-10-03 |
| commit |
0.03 |
server: report only decision capability for decision models |
|
Specific technical explanation using rel |
2026-10-01 |
| commit |
0.03 |
x/transfer, server: tighten redirect handling for registry r |
|
Uses precise security and networking ter |
2026-09-17 |
| commit |
0.03 |
MLX, MLX-C: version bump (#18449) |
|
Specific version and model details indic |
2026-09-15 |
| pr |
0.03 |
launch: suppress Claude Code connector warning for Ollama |
|
Brief, implementation-specific descripti |
2026-10-08 |
| pr |
0.03 |
model: add multimodal embeddings |
|
Terse, domain-specific implementation su |
2026-10-06 |
| pr |
0.03 |
qwen3.5: a tool call can open before the thinking channel is |
|
Highly specific parser behavior, test re |
2026-09-24 |
| pr |
0.03 |
server: reload when two tags share a blob but need different |
|
Detailed reproduction steps and implemen |
2026-09-07 |
| pr |
0.03 |
qwen3coder: tolerate a dropped closing tag, and stop rewriti |
|
Concrete production symptoms and parser |
2026-08-21 |
| pr |
0.03 |
gemma4: do not answer with a channel name the parser was cut |
|
Specific streaming behavior, token detai |
2026-08-08 |
| pr |
0.03 |
gemma4: recover a finished tool call that is missing its clo |
|
Captured failure details, existing-test |
2026-08-04 |
| pr |
0.03 |
README: add MailRules to community integrations |
|
Concise, specific integration note with |
2026-10-09 |
| pr |
0.03 |
launch: avoid inherited service-tier warnings in Codex |
|
Specific configuration behavior and conc |
2026-10-08 |
| pr |
0.03 |
server: don't list migrated models twice |
|
Detailed migration behavior and regressi |
2026-10-07 |
| pr |
0.03 |
llm,fs/gguf: count VRAM per context and predict the KV cache |
|
Dense implementation jargon, measurement |
2026-10-08 |
| pr |
0.03 |
decision: add the Saina Helm encoding |
|
Very detailed domain-specific proposal w |
2026-10-08 |
| pr |
0.03 |
manifest: avoid symlinks on Windows |
|
Specific Windows filesystem and locking |
2026-10-07 |
| pr |
0.03 |
openai: treat a tool message with null content as an empty r |
|
Specific issue overlap, tests, and cavea |
2026-10-07 |
| pr |
0.03 |
server: serve Prometheus metrics at /metrics when OLLAMA_MET |
|
Detailed issue comparison and implementa |
2026-09-17 |
| commit |
0.02 |
manifest: avoid symlinks on Windows (#18852) |
|
Specific Windows, locking, and pruning d |
2026-10-07 |
| commit |
0.02 |
mlx: fix patch for recent mlx update (#18812) |
|
Terse technical title with issue referen |
2026-10-06 |
| commit |
0.02 |
test: split create integration tests out (#18470) |
|
Plain, domain-specific explanation with |
2026-10-05 |
| commit |
0.02 |
pull: allow RCs to pull matching min_version (#18790) |
|
Concise release-version behavior change |
2026-10-04 |
| commit |
0.02 |
mlx: Decision model improvements (#18776) |
|
Terse technical bullets describing concr |
2026-10-04 |
| commit |
0.02 |
openai: split tool results with images instead of adding [im |
|
Specific implementation details and doma |
2026-10-03 |
| commit |
0.02 |
MLX: version bump (#18576) |
|
Terse, domain-specific commit message wi |
2026-09-23 |
| commit |
0.02 |
server: fix getExistingName canonicalizing model name parts |
|
Detailed Go bug analysis uses concrete r |
2026-09-22 |
| commit |
0.02 |
tests: fix metadata deletion test flake (#18493) |
|
Terse, issue-specific wording with concr |
2026-09-17 |
| commit |
0.02 |
llama.cpp: version bump b10969 (#18446) |
|
Terse, repository-specific changelog wit |
2026-09-15 |
| commit |
0.02 |
API: Deprecate typical_p (#18448) |
|
Concise technical change using specific |
2026-09-15 |
| commit |
0.02 |
MLX: version bump (#18235) |
|
Terse version-bump notes and typical rev |
2026-09-14 |
| commit |
0.02 |
llm: keep gemma3n projector off the CPU (#18376) |
|
Concise, domain-specific wording and con |
2026-09-11 |
| pr |
0.02 |
x/mlxrunner/mlx: link against libdl on linux |
|
A concise, technically precise linker di |
2026-08-04 |
| pr |
0.02 |
llm: stop guessing that a repetitive payload is a runaway |
|
The duplicated sentence and concrete deb |
2026-08-04 |
| pr |
0.02 |
ci: harden download steps with retries |
|
Colloquial phrasing and specific CI flak |
2026-10-08 |
| pr |
0.02 |
server: hide duplicate and downgrade guards from list |
|
Concrete legacy-GGUF behavior and issue |
2026-10-08 |
| pr |
0.02 |
Set CLAUDE_CODE_MAX_CONTEXT_TOKENS to model's context length |
|
Direct issue wording and concrete contex |
2026-10-07 |
| pr |
0.02 |
cmd: add a home menu for models and agents |
|
Concise product behavior description use |
2026-10-08 |
| pr |
0.02 |
openai: accept json_object text format in responses |
|
Concise, domain-specific change descript |
2026-10-08 |
| pr |
0.02 |
llm: reuse llama-server HTTP connections for embed load |
|
Concrete implementation details, issue r |
2026-09-12 |
| pr |
0.02 |
docs: adding research agent sample project |
|
Brief and concrete documentation note wi |
2026-10-07 |
| pr |
0.02 |
cmd: remove account step from CLI onboarding |
|
Concrete CLI behavior and exact test com |
2026-10-06 |
| pr |
0.02 |
llm: add support for AIX platform and big-endian GGUF models |
|
Terse platform-specific changelog with c |
2026-10-07 |
| pr |
0.02 |
UX: reject oversized pulls by default |
|
Specific implementation limits and infor |
2026-09-04 |
| commit |
0.01 |
mlx: drop carried metal residency patch now that it is upstr |
|
Terse, domain-specific commit title with |
2026-10-07 |
| commit |
0.01 |
cmd: remove account step from CLI onboarding (#18826) |
|
Short, conventional commit message with |
2026-10-07 |
| commit |
0.01 |
llama: fix clef head reads past 2GiB on windows (#18777) |
|
Highly specific implementation details a |
2026-10-07 |
| commit |
0.01 |
server: proxy cloud usage and balance APIs (#18829) |
|
Concise technical summary naming concret |
2026-10-07 |
| commit |
0.01 |
docs: fix 6 dead links in README community integrations list |
|
Brief conventional documentation commit |
2026-10-06 |
| commit |
0.01 |
MLX: version bump (#18720) |
|
Brief release notes with specific MLX an |
2026-10-06 |
| commit |
0.01 |
llama.cpp: version update (#18761) |
|
Concise version-update commit message. |
2026-10-06 |
| commit |
0.01 |
docs: document image input for decision models (#18757) |
|
Terse, conventional documentation commit |
2026-10-02 |
| commit |
0.01 |
docs: link to available decision models (#18758) |
|
Terse, conventional documentation commit |
2026-10-02 |
| commit |
0.01 |
ci: fix missing build context (#18742) |
|
Terse, conventional commit message with |
2026-10-02 |
| commit |
0.01 |
models: add clef support (#18741) |
|
Brief domain-specific commit title; no s |
2026-10-01 |
| commit |
0.01 |
README: add oxi to community integrations |
|
Terse, repository-specific integration u |
2026-10-01 |
| commit |
0.01 |
llama.cpp: version update (#18577) |
|
Concise technical changelog with specifi |
2026-09-22 |
| commit |
0.01 |
app: avoid System Events for ChatGPT/Codex detection (#18601 |
|
Brief, natural commit title with concret |
2026-09-22 |
| commit |
0.01 |
server: allow registry cross-host redirects among allowliste |
|
Terse, issue-specific commit title with |
2026-09-19 |
| commit |
0.01 |
api: expose model thinking levels and defaults (#18473) |
|
Terse, issue-specific title with no AI-s |
2026-09-18 |
| commit |
0.01 |
app: keep closed macOS windows from reopening on activation |
|
Concise, domain-specific commit-style wo |
2026-09-18 |
| commit |
0.01 |
docs: clarify integrations overview and add assistant icons |
|
Brief conventional documentation change |
2026-09-17 |
| commit |
0.01 |
cli: add first-run onboarding shared with the desktop app (# |
|
Terse conventional feature summary witho |
2026-09-17 |
| commit |
0.01 |
app: add Apps deep link (#18497) |
|
Short, conventional commit-style feature |
2026-09-17 |
| commit |
0.01 |
app: clarify Claude connection dialog (#18498) |
|
Terse, issue-specific commit title with |
2026-09-16 |
| commit |
0.01 |
docs: fix cloud retirements settings link (#18496) |
|
Brief technical commit title is characte |
2026-09-16 |
| commit |
0.01 |
docs: refresh getting started guides (#18450) |
|
Brief conventional commit message with n |
2026-09-15 |
| commit |
0.01 |
mlx: add mlx patch to docker build context (#18440) |
|
Terse, repository-specific commit title |
2026-09-14 |
| commit |
0.01 |
docs: add ChatGPT Desktop integration (#18377) |
|
Brief, conventional documentation commit |
2026-09-14 |
| commit |
0.01 |
app: refresh Apps layout and command copy feedback (#18372) |
|
Concise, conventional commit title with |
2026-09-14 |
| commit |
0.01 |
llm: raise token repeat limit to 100 and return error instea |
|
Terse, domain-specific commit message wi |
2026-09-10 |
| pr |
0.01 |
docs: add missing comma in CLI reference |
|
Terse, specific documentation fix typica |
2026-10-09 |
| pr |
0.01 |
docs: fix copy-pasted alt text on marimo chat image |
|
Brief and concrete documentation correct |
2026-10-09 |
| pr |
0.01 |
docs: fix copy-pasted alt text on Zed dropdown image |
|
Brief, specific fix with relevant access |
2026-10-09 |
| pr |
0.01 |
docs: use https for Oh My Pi image |
|
Terse, practical URL correction. |
2026-10-09 |
| pr |
0.01 |
docs: use https for Hermes Desktop image |
|
Terse, practical URL correction. |
2026-10-09 |
| pr |
0.01 |
llm: include eos tokens in raw generate responses |
|
First-person testing details and specifi |
2026-10-08 |
| pr |
0.01 |
docs: use American spelling canceled in API reference |
|
Terse documentation wording with a preci |
2026-10-09 |
| pr |
0.01 |
docs: fix IntelliJ brand casing in JetBrains guide |
|
Brief, file-specific casing correction t |
2026-10-09 |
| pr |
0.01 |
docs: use https for Onyx site link |
|
Minimal, concrete link protocol fix. |
2026-10-09 |
| pr |
0.01 |
docs: fix PowerShell casing in troubleshooting guide |
|
Terse correction with relevant code-fenc |
2026-10-09 |
| pr |
0.01 |
docs: fix typo an to and in troubleshooting guide |
|
Short typo fix with a precise source loc |
2026-10-09 |
| pr |
0.01 |
README: update Msty link to msty.ai |
|
Terse, issue-specific update with a mino |
2026-08-26 |
| pr |
0.01 |
fix: honor Codex App num_ctx metadata |
|
Template-like sections contain concise, |
2026-05-30 |
| pr |
0.01 |
server: support slashes in OpenAI retrieve model IDs |
|
Concise technical description with preci |
2026-10-07 |
| pr |
0.01 |
api: include HTTP status in client error messages |
|
Terse, issue-focused wording and concret |
2026-10-07 |
| pr |
0.01 |
renderers: pair functiongemma tool results with their call i |
|
Specific renderer behavior, API details, |
2026-10-08 |
| pr |
0.01 |
README: add oxi to community integrations |
|
Specific product details and explicit au |
2026-10-01 |
| pr |
0.01 |
mlx: drop carried metal residency patch now that it is upstr |
|
Terse, repository-specific commit messag |
2026-10-07 |
| pr |
0.01 |
mlx: route fast quantized matmul on CUDA |
|
Highly specific backend and quantization |
2026-06-25 |
| pr |
0.01 |
llama: fix clef head reads past 2GiB on windows |
|
Extensive platform-specific diagnostics |
2026-10-03 |
| commit |
0.00 |
server: skip local model compatibility migration (#18908) |
|
Terse, specific commit message with no A |
2026-10-09 |
| commit |
0.00 |
routes: fix embed 413 vs 400 routing (#18822) |
|
Terse, issue-linked routing fix typical |
2026-10-06 |
| commit |
0.00 |
openai: keep image positions when merging tool result parts |
|
Terse, code-specific explanation with co |
2026-10-02 |
| commit |
0.00 |
openai: keep empty tool results sent as an empty content arr |
|
Concise bug description uses precise too |
2026-10-01 |
| commit |
0.00 |
openai: keep tool message content parts in one message |
|
Detailed protocol-specific diagnosis and |
2026-09-30 |
| commit |
0.00 |
docs: document System One API (#18702) |
|
— |
2026-09-30 |
| commit |
0.00 |
create: support explicit model capabilities (#18708) |
|
— |
2026-09-29 |
| commit |
0.00 |
llama.cpp: version bump b11232 (#18652) |
|
— |
2026-09-29 |
| commit |
0.00 |
MLX: version bump (#18651) |
|
— |
2026-09-29 |
| commit |
0.00 |
feat: allow ten web searches per response (#18602) |
|
— |
2026-09-28 |
| commit |
0.00 |
mlx: bound pull stall retries and let the watchdog interrupt |
|
— |
2026-09-28 |
| commit |
0.00 |
feat: add System One scoring API (#18606) |
|
— |
2026-09-28 |
| commit |
0.00 |
bench: use HumanEval patch prompts (#17480) |
|
— |
2026-09-26 |
| commit |
0.00 |
llama-server: prepare to remove compatibility patch |
|
Highly domain-specific migration and man |
2026-07-29 |
| commit |
0.00 |
api: deprecate typical_p (#18627) |
|
— |
2026-09-24 |
| commit |
0.00 |
app: isolate cloud-setting tests from Windows user config (# |
|
— |
2026-09-24 |
| commit |
0.00 |
app: defer Settings model discovery (#18598) |
|
— |
2026-09-24 |
| commit |
0.00 |
app: sync macOS update menu and icon at startup (#18622) |
|
— |
2026-09-24 |
| commit |
0.00 |
mlxrunner: Update XGrammar to 0.2.7 for structured outputs |
|
— |
2026-09-23 |
| commit |
0.00 |
launch: default Claude Code to client-side auto checks (#185 |
|
— |
2026-09-23 |
| commit |
0.00 |
xgrammar: add standalone CMake project for native library (# |
|
— |
2026-09-23 |
| commit |
0.00 |
docs: document thinking control discovery (#18501) |
|
— |
2026-09-23 |
| commit |
0.00 |
mlx: speed up Qwen 3.8 prompt processing (#18550) |
|
— |
2026-09-23 |
| commit |
0.00 |
mlx: select Gemma 4 image resolution dynamically (#18603) |
|
— |
2026-09-23 |
| commit |
0.00 |
server: apply structured outputs in a single pass on thinkin |
|
Detailed domain-specific explanation lac |
2026-09-14 |
| commit |
0.00 |
llm: apply a format after the thinking in one llama-server r |
|
Technical implementation details and cav |
2026-09-15 |
| commit |
0.00 |
mlxrunner: constrain a format after the thinking in one gene |
|
Specific MLX grammar behavior and limita |
2026-09-14 |
| commit |
0.00 |
parsers: report the strings that end a response's thinking |
|
Parser-specific terminology and concrete |
2026-09-14 |
| commit |
0.00 |
mlxrunner: Release freed KV buffers during speculative decod |
|
Detailed MLX diagnosis with concrete mea |
2026-09-17 |
| commit |
0.00 |
mlxrunner: lay out model by contract, checkpoint and constru |
|
Detailed refactoring rationale uses spec |
2026-09-16 |
| commit |
0.00 |
mlxrunner: lay out nn one layer kind per file |
|
Specific package and interface reference |
2026-09-16 |
| commit |
0.00 |
mlxrunner: read model manifests through the manifest package |
|
Concrete manifest and loader details out |
2026-09-16 |
| commit |
0.00 |
mlxrunner, cmd, server: fold three small packages into their |
|
Highly specific package consolidation de |
2026-09-16 |
| commit |
0.00 |
mlx, mlxrunner: move the MLX engine out of x/ |
|
Detailed path list and build-system refe |
2026-09-16 |
| commit |
0.00 |
tokenizer, ml: remove dead code |
|
Concise cleanup summary cites concrete r |
2026-09-16 |
| commit |
0.00 |
cmd: launch the MLX runner without the engine dispatcher |
|
Specific Cobra behavior and command chan |
2026-09-16 |
| commit |
0.00 |
cmd: remove built-in agent (#18393) |
|
Terse commit title with no AI-style phra |
2026-09-11 |
| commit |
0.00 |
mlxrunner: release the buffers weight loading leaves in the |
|
Specific allocator, model, and memory fi |
2026-09-10 |
| commit |
0.00 |
mlx: drop the empty-handle checks outside the bindings |
|
Precise internal lifecycle details and t |
2026-09-04 |
| commit |
0.00 |
mlx: scope array lifetimes instead of pinning and sweeping |
|
Highly specific implementation and bench |
2026-09-04 |
| commit |
0.00 |
mlxrunner: wait for a killed runner to exit before the sched |
|
Highly specific concurrency and signal-h |
2026-09-09 |
| commit |
0.00 |
mlxrunner: bound MLX loads by system free memory while other |
|
Uses detailed Apple silicon and schedule |
2026-09-10 |
| commit |
0.00 |
mlxrunner: evict the active conversation's own checkpoints u |
|
Specific implementation behavior, model |
2026-09-09 |
| commit |
0.00 |
mlxrunner: keep the reused head of a cached edge safe from e |
|
Detailed cache and recurrent-layer behav |
2026-09-10 |
| pr |
0.00 |
server: skip local model compatibility migration |
|
Terse, specific commit message with natu |
2026-10-09 |