Commit History

  1. remove wiki

    1b0a90a
  2. fix wikipedia search url

    088248d
  3. add tools support

    e1c507f

  1. changing response way llm

    8a9ea80

  1. changing response way

    9ca42fc
  2. Update system model to Gemma 3 1B Instruct and humanize responses

    9eed65c

  1. fix: improve question matcher keyword normalization, expanded stop words, and use full XML context in LLM fallback

    069b0f0
  2. feat: add LLM-based query restructuring and grammar correction stage

    d15b317
  3. feat: implement semantic chunking, dual-retrieval cross-lingual search, and direct Q&A bypass routing

    42e9ea7
  4. Fix Hindi/Hinglish language output alignment using retrieval translation pipeline

    12aa926
  5. fix: add threading lock synchronization to prevent concurrent model inference crashes

    9af21b3
  6. feat: optimize response time by offloading translation to deep-translator and reducing ctx/retrieval k size

    147efdf
  7. data: update source documents to hindi_interview_questions_fixed.pdf

    72e81d0
  8. chore: revert embedding model to all-MiniLM-L6-v2 since query translation handles English retrieval

    9456ff5
  9. feat: implement multilingual query translation and language-specific prompting

    48bc44c
  10. feat: return custom sorry message if context is missing or not provided

    01f2bc0
  11. feat: implement multilingual RAG pipeline for English, Hindi, and Hinglish

    f09b23d
  12. Optimize Hugging Face Space: add eager model loading, reduce max tokens, fix stop tokens, limit CPU threads

    09bc714
  13. feat: implement complete RAG-based AI engine with Ollama fallback and vector search support

    b597dd6

  1. Optimize startup speed and query latency

    d641e1d
  2. add new documents

    717df55
  3. add README.md with Hugging Face Space metadata

    5567085
  4. Initial clean release

    0a96660