Commits

Commit history

1. Commits on Jul 26, 2026

   • fix(_internals): clean up native resources on initialization failures

     Show description for 998b7b5
     JamePeng

     committed
     998b7b5View commit details

     Copy full SHA for 998b7b5

     View code at this pointBrowse repository at this point

   • fix(embedding): respect n_seq_max when streaming embedding batches

     Show description for 29ad350
     JamePeng

     committed
     29ad350View commit details

     Copy full SHA for 29ad350

     View code at this pointBrowse repository at this point

   • refactor(batch): improve sequence capacity validation guidance

     Show description for 0d3ae93
     JamePeng

     committed
     0d3ae93View commit details

     Copy full SHA for 0d3ae93

     View code at this pointBrowse repository at this point

   • feat(embedding): modernize the built-in Llama embedding API

     Show description for 93ebd83
     JamePeng

     committed
     93ebd83View commit details

     Copy full SHA for 93ebd83

     View code at this pointBrowse repository at this point
2. Commits on Jul 21, 2026

   • Bump version to 0.3.44

     Show description for ebf6099
     JamePeng

     committed
     ebf6099View commit details

     Copy full SHA for ebf6099

     View code at this pointBrowse repository at this point
3. Commits on Jul 20, 2026

   • fix(ggml): preload bundled OpenMP runtime before loading ggml-base

     Show description for 0eb2615
     JamePeng

     committed
     0eb2615View commit details

     Copy full SHA for 0eb2615

     View code at this pointBrowse repository at this point
4. Commits on Jul 17, 2026

   • Bump version to 0.3.43

     Show description for caa72e2
     JamePeng

     committed
     caa72e2View commit details

     Copy full SHA for caa72e2

     View code at this pointBrowse repository at this point

   • patch(Gemma4ChatHandler): Synchronize huggingface gemma4 latest chat template

     Show description for a641283
     JamePeng

     committed
     a641283View commit details

     Copy full SHA for a641283

     View code at this pointBrowse repository at this point
5. Commits on Jul 16, 2026

   • feat(llama_ext): support optional llama-ext.h API bindings

     Show description for 978ff32
     JamePeng

     committed
     978ff32View commit details

     Copy full SHA for 978ff32

     View code at this pointBrowse repository at this point

   • fix(ctypes): validate argument types before binding shared library functions

     Show description for 6c163c2
     JamePeng

     committed
     6c163c2View commit details

     Copy full SHA for 6c163c2

     View code at this pointBrowse repository at this point

   • feat(ctypes): handle missing optional symbols gracefully

     Show description for afafdd5
     JamePeng

     committed
     afafdd5View commit details

     Copy full SHA for afafdd5

     View code at this pointBrowse repository at this point

   • fix(mtmd): preserve subclass chat format during MTMD initialization

     Show description for a804dc2
     JamePeng

     committed
     a804dc2View commit details

     Copy full SHA for a804dc2

     View code at this pointBrowse repository at this point
6. Commits on Jul 15, 2026

   • feat(ctypes): support ABI-compatible symbol aliases

     Show description for 849f02a
     JamePeng

     committed
     849f02aView commit details

     Copy full SHA for 849f02a

     View code at this pointBrowse repository at this point

   • refactor(embedding): rename llama_cpp import alias to llama_cpp_lib

     Show description for 859a99d
     JamePeng

     committed
     859a99dView commit details

     Copy full SHA for 859a99d

     View code at this pointBrowse repository at this point

   • Update Submodule vendor/llama.cpp 7f575c3..aff6eb6

     Show description for 5334cb9
     JamePeng

     committed
     5334cb9View commit details

     Copy full SHA for 5334cb9

     View code at this pointBrowse repository at this point
7. Commits on Jul 14, 2026

   • refactor(mtmd): cache Generic MTMD chat template resolution

     Show description for 9d6e598
     JamePeng

     committed
     9d6e598View commit details

     Copy full SHA for 9d6e598

     View code at this pointBrowse repository at this point

   • Update Submodule vendor/llama.cpp e3546c7..7f575c3

     Show description for 1762647
     JamePeng

     committed
     1762647View commit details

     Copy full SHA for 1762647

     View code at this pointBrowse repository at this point
8. Commits on Jul 12, 2026

   • patch(Llama): Increase chunk preview limit to 128 in Llama.eval exception

     Show description for e417924
     JamePeng

     committed
     e417924View commit details

     Copy full SHA for e417924

     View code at this pointBrowse repository at this point

   • Bump version to 0.3.42

     Show description for e522cec
     JamePeng

     committed
     e522cecView commit details

     Copy full SHA for e522cec

     View code at this pointBrowse repository at this point
9. Commits on Jul 11, 2026

   • fix(loader): improve Windows DLL search path handling and diagnostics

     Show description for 516ec3f
     JamePeng

     committed
     516ec3fView commit details

     Copy full SHA for 516ec3f

     View code at this pointBrowse repository at this point

   • Update Submodule vendor/llama.cpp a935fbf..c92e806

     Show description for 6c954f4
     JamePeng

     committed
     6c954f4View commit details

     Copy full SHA for 6c954f4

     View code at this pointBrowse repository at this point

   • fix(ggml): load ggml-base before ggml library

     Show description for 36bdecf
     JamePeng

     committed
     36bdecfView commit details

     Copy full SHA for 36bdecf

     View code at this pointBrowse repository at this point
10. Commits on Jul 10, 2026

    • fix(mtmd): validate MTMD inputs before tokenization

      Show description for 7f59a86
      JamePeng

      committed
      7f59a86View commit details

      Copy full SHA for 7f59a86

      View code at this pointBrowse repository at this point

    • feat(LlamaBatch): add mixed token embedding batch support

      Show description for 11364d1
      JamePeng

      committed
      11364d1View commit details

      Copy full SHA for 11364d1

      View code at this pointBrowse repository at this point
11. Commits on Jul 9, 2026

    • feat: add embedding rows to LlamaBatch

      Show description for a5ceecb
      JamePeng

      committed
      a5ceecbView commit details

      Copy full SHA for a5ceecb

      View code at this pointBrowse repository at this point

    • fix(_internals): harden LlamaBatch token writes

      Show description for 0fdbc3d
      JamePeng

      committed
      0fdbc3dView commit details

      Copy full SHA for 0fdbc3d

      View code at this pointBrowse repository at this point
12. Commits on Jul 8, 2026

    • fix: validate eval tokens before native decode

      Show description for 6c7b16f
      JamePeng

      committed
      6c7b16fView commit details

      Copy full SHA for 6c7b16f

      View code at this pointBrowse repository at this point
13. Commits on Jul 7, 2026

    • fix(types): make assistant message name optional

      Show description for b4c74bf
      JamePeng

      committed
      b4c74bfView commit details

      Copy full SHA for b4c74bf

      View code at this pointBrowse repository at this point

    • Bump version to 0.3.41

      Show description for 169d5e1
      JamePeng

      committed
      169d5e1View commit details

      Copy full SHA for 169d5e1

      View code at this pointBrowse repository at this point

    • refactor(mtmd): extract prompt rendering and media marker normalization

      Show description for d3f0ac9
      JamePeng

      committed
      d3f0ac9View commit details

      Copy full SHA for d3f0ac9

      View code at this pointBrowse repository at this point
14. Commits on Jul 4, 2026

    • fix(vocab): update vocab_type to use self.vocab and add None checks

      Show description for 79a8635
      JamePeng

      committed
      79a8635View commit details

      Copy full SHA for 79a8635

      View code at this pointBrowse repository at this point
15. Commits on Jul 2, 2026

    • Update Submodule vendor/llama.cpp 4f31eed..fdb1db8

      Show description for 8b38e72
      JamePeng

      committed
      8b38e72View commit details

      Copy full SHA for 8b38e72

      View code at this pointBrowse repository at this point
16. Commits on Jun 30, 2026

    • refactor(mtmd): extract `mtmd_tokenize` into `_mtmd_tokenize` standalone helper

      Show description for 46708d1
      JamePeng

      committed
      46708d1View commit details

      Copy full SHA for 46708d1

      View code at this pointBrowse repository at this point
17. Commits on Jun 26, 2026

    • Update Submodule vendor/llama.cpp 1191758..3fc4e10

      Show description for b74cabb
      JamePeng

      committed
      b74cabbView commit details

      Copy full SHA for b74cabb

      View code at this pointBrowse repository at this point
18. Commits on Jun 21, 2026

    • feat(speculative): Improve ngram-map draft selection and accept feedback

      Show description for acc3fe2
      JamePeng

      committed
      acc3fe2View commit details

      Copy full SHA for acc3fe2

      View code at this pointBrowse repository at this point