Commit History

  1. Optimize AJ-Mini for faster responses: reduce tokens, add fast test endpoints

    AJ STUDIOZ
    bdeabc1
  2. Improve AJ-Mini: proper system prompt, better response extraction, increased tokens

    AJ STUDIOZ
    d2bfde4
  3. Remove API key requirement for public AJ-Mini access

    AJ STUDIOZ
    5f69e9a
  4. Update timestamps to Indian Standard Time (IST)

    AJ STUDIOZ
    70c23c6
  5. Rebrand to AJ-Mini v1.0 - unlimited free backup model by AJ STUDIOZ

    AJ STUDIOZ
    2f940ac
  6. Revert to TinyLlama for stability - works reliably on HF Spaces

    AJ STUDIOZ
    acca413
  7. Upgrade to AJ-Coder v1.0 - Fine-tuned by AJ STUDIOZ based on Qwen2.5-Coder-1.5B

    AJ STUDIOZ
    a618657
  8. Optimize inference speed: reduce max_tokens, add torch.no_grad, greedy decoding

    AJ STUDIOZ
    4502e0b
  9. Load model directly in Space - Unlimited FREE usage!

    AJ STUDIOZ
    bad0554
  10. Switch to TinyLlama for free tier compatibility

    AJ STUDIOZ
    0a9b18d
  11. Switch to direct HTTP requests with new router endpoint

    AJ STUDIOZ
    ee32985
  12. Fix: Explicitly use new HuggingFace router endpoint

    AJ STUDIOZ
    b52d3d8
  13. Update huggingface-hub to 0.26.2 for new router endpoint

    AJ STUDIOZ
    05890d1
  14. Update to deepseek-ai/deepseek-coder-1.3b-instruct model

    AJ STUDIOZ
    fb60ef4
  15. Revert to HuggingFace InferenceClient - Cloud-based solution for low-spec systems

    AJ STUDIOZ
    4534aef

  1. Fix: Switch to InferenceClient to resolve deprecated API endpoint

    AJ STUDIOZ
    5528539
  2. Switch to TinyLlama for reliable inference

    AJ STUDIOZ
    030c76f
  3. Switch to Qwen2.5-Coder for better reliability

    AJ STUDIOZ
    406395b
  4. Add YAML metadata to README

    AJ STUDIOZ
    2485a9a
  5. Initial commit: AJ STUDIOZ DeepSeek API

    AJ STUDIOZ
    d039057