Commit History

  1. fix: format history as string to prevent Llama 3 hallucinating assistant role during contextualization

    Dus Tran
    ce8ae9c
  2. fix: prevent LLM from hallucinating queries during contextualization

    Dus Tran
    428313f
  3. fix: preload models on startup and increase LLM timeout

    Dus Tran
    9a01e3a
  4. fix: soften strict prompt for keyword queries

    Dus Tran
    3ff2322
  5. fix: handle 2D numpy array from multilingual reranker

    Dus Tran
    bd6c976
  6. fix: switch to multilingual reranker

    Dus Tran
    d6fdc60
  7. fix: dockerignore and memory init

    Dus Tran
    c8a642a
  8. chore: remove IS_RENDER and gemini imports

    Dus Tran
    fbac3e6

  1. feat(rag): enable rerank model on render

    Dus Tran
    463f005
  2. feat(rag): use bkai huggingface model universally

    Dus Tran
    0bbf4ee
  3. fix(prompt): allow LLM to respond to chit-chat naturally

    Dus Tran
    3f20a80
  4. fix(rag): update prompt and reduce temperature to prevent hallucinations

    Dus Tran
    3380923
  5. fix: use models/gemini-embedding-2 with 768 dimensions for compatibility with local database

    Dus Tran
    dc6ff05
  6. fix: correct GoogleGenerativeAIEmbeddings class name

    Dus Tran
    f475e72
  7. fix: make heavy ML libraries import conditionally on Render to prevent OOM/timeouts

    Dus Tran
    af847aa
  8. fix: strip carriage returns from database url

    Dus Tran
    525c11d
  9. feat: tích hợp supabase database và cấu hình github actions ci

    Dus Tran
    ea19adc

  1. “…I”

    Dus Tran
    feb5410

  1. “…I”

    Dus Tran
    cf573aa
  2. hoàn thiện Multimodal RAG chatbot với FastAPI Backend & Web UI

    Dus Tran
    44ed2c7