Get up and running with Kimi, GLM, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.
Commits on main
mlx: select Gemma 4 image resolution dynamically (#18603)
b8c3d1f Daniel Hiltgen committed 1 month ago
server: apply structured outputs in a single pass on thinking models
5a0ff31 Jesse Gross committed 2 months ago
llm: apply a format after the thinking in one llama-server request
2ff052b Jesse Gross committed 2 months ago
mlxrunner: constrain a format after the thinking in one generation
1ce2b68 Jesse Gross committed 2 months ago
parsers: report the strings that end a response's thinking
a9d8953 Jesse Gross committed 2 months ago
server: fix getExistingName canonicalizing model name parts independently (#18438)
6ae5088 frob committed 1 month ago
server: allow registry cross-host redirects among allowlisted hosts (#18533)
6383a0f Patrick Devine committed 2 months ago
app: keep closed macOS windows from reopening on activation (#18518)
9a44d9d Eva H committed 2 months ago
x/transfer, server: tighten redirect handling for registry requests (#18512)
dfabde4 Patrick Devine committed 2 months ago
docs: clarify integrations overview and add assistant icons (#18511)
35bdad6 Eva H committed 2 months ago
cli: add first-run onboarding shared with the desktop app (#18495)
7fa8a02 Eva H committed 2 months ago
mlxrunner: Release freed KV buffers during speculative decode
ec3cc23 Jesse Gross committed 2 months ago
mlxrunner: lay out model by contract, checkpoint and construction
66c3238 Jesse Gross committed 2 months ago
mlxrunner: read model manifests through the manifest package
d27fde6 Jesse Gross committed 2 months ago
mlxrunner, cmd, server: fold three small packages into their users
04b450a Jesse Gross committed 2 months ago