All pull requests

Pull requests

• fix(chat_format): parse Gemma 4 native tool-call tokens into tool_calls (#2227)

  #2232·Anai-Guo opened May 28, 2026Contributor·1 of 2 tasks·

  ·
  1

  4

• security: fix SSRF in multimodal image URL loading (_load_image)

  #2220·hoangperry opened May 16, 2026·5 tasks done·

  ·

• fix: improve error message when LlamaModel fails to load

  #2187·Anai-Guo opened Apr 21, 2026Contributor·

  ·
  1

  1

• Add chat template for gemma models

  #2183·C00kieFact0ry opened Apr 13, 2026·

  ·
  1

• fix: prevent KV cache corruption on SWA/ISWA models + hot-path perf

  #2180·avion23 opened Apr 12, 2026Contributor·

  ·
  1

• perf: vectorize KV cache prefix matching with numpy

  #2179·nausicaalii opened Apr 11, 2026·4 tasks done·

  ·

• build: disable soname to reduce binary size

  #2177·Bing-su opened Apr 9, 2026·

  ·
  1

• feat: add reasoning_effort to chat completions API

  #2167·abetlen opened Mar 30, 2026Owner·

  ·
  4

• fix: auto-disable mmap when all layers offloaded to GPU (#1964)

  #2147·ljluestc opened Mar 22, 2026·

  ·
  1

• Clear kv cache and reset tokens after chat completion

  #2141·thisisayushg opened Mar 14, 2026·

  ·

• This PR implements the previously stubbed state management methods in the _internals.py module and updates the corresponding API calls in llama.py to use the correct underlying C++ function names.

  #2134·bsides230 opened Mar 5, 2026·

  ·

• feat: Add DeepSeek R1 and distilled model support

  #2131·ljluestc opened Mar 1, 2026·

  ·
  1

• feat: add streaming tool use (rebased #1884 on latest main)

  #2129·XyLearningProgramming opened Feb 23, 2026·

  ·
  1

• chore: bump conda-incubator/setup-miniconda from v3.1.0 to v3.3.0

  #2128·Aiudadadadf opened Feb 22, 2026·

  ·

• feat: support Granite-Docling model

  #2109·dhdaines opened Jan 4, 2026·

  ·
  1

• Fix issue #2096: Handle URLs with embedded HTTP credentials in _load_image

  #2102·nMaroulis opened Dec 10, 2025·

  ·

• chore: update typing-extensions dependency and set github actions setup-python to v6

  #2099·AnvithaCodes opened Nov 28, 2025·

  ·

• Fix: Install correct CUDA toolkit during build

  #2088·chamalgomes opened Nov 12, 2025·

  ·
  1

  1

• Include x64 directory for CUDA DLLs on Windows

  #2083·ajparsons opened Oct 24, 2025·

  ·

• Better Qwen2.5-VL chat template.

  #2066·alcoftTAO opened Sep 7, 2025Contributor·

  ·

• Add timeout and error handling in FastAPI uvicorn server

  #2044·amandwivedi45 opened Jul 22, 2025·

  ·

• Actually create a random seed when using seed = -1 on load

  #2042·m-from-space opened Jul 16, 2025·

  ·

• Improve error message when model file is missing

  #2041·NITHIN0710 opened Jul 9, 2025·

  ·
  1

• ARM Runners support CUDA SBSA

  #2039·johnnynunez opened Jul 7, 2025·

  ·
  1

• Flush libc stdout/stderr in suppress_stdout_stderr

  #2015·AuroraWright opened May 7, 2025·

  ·