Issues

• Web demo: Hugging Face Space for researcher self-serve access

  nmrenyi/mamai#63 · Filter by author nmrenyi opened on Jun 3, 2026

• RAG: investigate query rewriting before retrieval (eval-gated)

  nmrenyi/mamai#62 · Filter by author nmrenyi opened on Jun 3, 2026

• Track Gemma 4 E4B NPU artifact for Snapdragon (QNN backend)

  nmrenyi/mamai#58 · Filter by author nmrenyi opened on May 15, 2026

• Manual review of low-safety responses before expanding pilot

  priority:P1
  nmrenyi/mamai#50 · Filter by author nmrenyi opened on Apr 15, 2026

• RAG context injection causes consistent performance regression across all GGUF models

  Evaluationpriority:P1
  nmrenyi/mamai#49 · Filter by author nmrenyi opened on Apr 15, 2026

• eval: reconsider deployment model — gemma3n-e4b outperforms gemma4-e4b on every open-ended metric

  Evaluationpriority:P1
  nmrenyi/mamai#48 · Filter by author nmrenyi opened on Apr 15, 2026

• eval: run each dataset as a separate cluster job for true GPU parallelism

  Evaluation
  nmrenyi/mamai#47 · Filter by author nmrenyi opened on Apr 15, 2026

• RagStream: guarantee terminal event on lifecycle/coroutine cancellation

  nmrenyi/mamai#45 · Filter by author nmrenyi opened on Apr 12, 2026

• Build a doc-grounded benchmark from the bundled clinical guideline corpus

  enhancementEvaluationpriority:P1
  nmrenyi/mamai#42 · Filter by author nmrenyi opened on Apr 11, 2026
  1

• Add retrieval-only evaluation for the bundled guideline corpus

  enhancementEvaluationpriority:P1
  nmrenyi/mamai#41 · Filter by author nmrenyi opened on Apr 11, 2026
  1

• Prototype agentic guideline search with LiteRT-LM tool calling

  enhancementpriority:P1
  nmrenyi/mamai#40 · Filter by author nmrenyi opened on Apr 11, 2026
  1

• Query rewriting: reformulate user input before retrieval to improve RAG quality

  enhancementEvaluation
  nmrenyi/mamai#36 · Filter by author nmrenyi opened on Apr 4, 2026

• Explore Gemma 4 tool use / function calling capabilities

  enhancement
  nmrenyi/mamai#35 · Filter by author nmrenyi opened on Apr 4, 2026

• RAG evaluation strategy: frameworks, metrics, gaps, and potential contributions

  enhancementEvaluation
  nmrenyi/mamai#34 · Filter by author nmrenyi opened on Apr 4, 2026

• Integrate RAGAS and DeepEval for end-to-end RAG evaluation (retrieval + generation)

  enhancementEvaluationhelp wanted
  nmrenyi/mamai#33 · Filter by author nmrenyi opened on Apr 4, 2026

• Extract RagPipeline into a standalone Flutter plugin for on-device RAG

  enhancementhelp wanted
  nmrenyi/mamai#32 · Filter by author nmrenyi opened on Apr 4, 2026
  3

• Improve PDF highlight text matching reliability

  nmrenyi/mamai#30 · Filter by author nmrenyi opened on Mar 27, 2026

• Verify Swahili (sw) translations with a medical professional

  nmrenyi/mamai#29 · Filter by author nmrenyi opened on Mar 25, 2026

• Evaluate the Swahili ability of the model

  enhancement
  nmrenyi/mamai#27 · Filter by author nmrenyi opened on Mar 15, 2026

• Improve model download robustness

  enhancement
  nmrenyi/mamai#25 · Filter by author nmrenyi opened on Mar 15, 2026

• Add confidence/uncertainty signal to the UI

  enhancement
  nmrenyi/mamai#24 · Filter by author nmrenyi opened on Mar 15, 2026

• Experiment with retrieval parameters (top_k, threshold)

  Evaluation
  nmrenyi/mamai#22 · Filter by author nmrenyi opened on Mar 15, 2026

• Re-examine the whole retrieval module

  nmrenyi/mamai#21 · Filter by author nmrenyi opened on Mar 15, 2026
  1

• Evaluation: consider temperature=0 for MCQ benchmarks

  Evaluation
  nmrenyi/mamai#18 · Filter by author nmrenyi opened on Mar 3, 2026

• Create user test cases for real user evaluation

  Evaluation
  nmrenyi/mamai#8 · Filter by author nmrenyi opened on Feb 18, 2026 · v0.1 basic offline conversational features