Commit History

2 days ago

  1. Cap llama.cpp prompt threads (n_threads_batch) at the container's cores

    Verified signature
    25470c5
  2. Install the llama.cpp wheel over HTTPS (builder checkout gave the LFS pointer)

    Verified signature
    0ac57bb
  3. Top 3 passages to the LLM: faster answers, citations 100% in eval

    Verified signature
    44c8bd7
  4. Install llama.cpp from a prebuilt glibc wheel (compiling on the builder timed out)

    Verified signature
    57fb762
  5. Docker SDK: compile llama.cpp for x86-64 + AVX2, bake models into the image

    Verified signature
    3dcb34c
  6. v2: cited answers with hybrid retrieval + cross-encoder rerank, Qwen3-1.7B on CPU, retrieval trace

    Verified signature
    a072368

  1. Fix text-generation pipeline and move chat_input to top level

    Verified signature
    134e302
  2. initial commit

    Verified signature
    cca01b4