Issues

• [Feature] Reuse the prefix KV across chat turns

  LogitAI/Gemstone#211 · Filter by author b-re-w opened on Oct 17, 2026

• [Decision] Spotlight-style global search overlay

  LogitAI/Gemstone#207 · Filter by author b-re-w opened on Oct 17, 2026

• [Decision] Notifications: what would the app notify about?

  LogitAI/Gemstone#206 · Filter by author b-re-w opened on Oct 17, 2026

• [Feature] Theme selection in the settings

  LogitAI/Gemstone#205 · Filter by author b-re-w opened on Oct 17, 2026
  1

• [Feature] Full Markdown rendering for replies

  LogitAI/Gemstone#204 · Filter by author b-re-w opened on Oct 17, 2026

• [Feature] Native desktop executable: verify the GraalVM build and hand it to compose-multiplatform-extended (G8, S3.4)

  LogitAI/Gemstone#191 · Filter by author b-re-w opened on Oct 17, 2026
  1

• [Feature] Offline operation: chat without a network (G3, S3.2)

  LogitAI/Gemstone#190 · Filter by author b-re-w opened on Oct 17, 2026

• Python 3.15t: Native dependencies without cp315t wheels

  LogitAI/Gemstone#178 · Filter by author b-re-w opened on Oct 17, 2026
  2

• Python: Make free-threaded 3.15t the standard interpreter for torchnative

  LogitAI/Gemstone#177 · Filter by author b-re-w opened on Oct 17, 2026
  1

• [Feature] Cross-device sync (scope to decide)

  enhancement
  LogitAI/Gemstone#167 · Filter by author b-re-w opened on Oct 3, 2026

• [Feature] GPU (CUDA) inference as a cuda extra

  enhancement
  LogitAI/Gemstone#164 · Filter by author b-re-w opened on Oct 3, 2026 · M4 Performance and coverage (after November)

• [Feature] Single transformers engine replacing the GGUF/BIN/GPTQ backends

  enhancement
  LogitAI/Gemstone#84 · Filter by author b-re-w opened on Oct 2, 2026 · M1 Single torchnative engine
  3

• [Feature] Fast paged-attention kernel

  enhancement
  LogitAI/Gemstone#74 · Filter by author b-re-w opened on Oct 2, 2026 · M4 Performance and coverage (after November)

• [Feature] 4-bit quantised weights

  enhancement
  LogitAI/Gemstone#67 · Filter by author b-re-w opened on Oct 2, 2026 · M4 Performance and coverage (after November)
  1

• [Feature] Replace the inference backends with a single torchnative serving system

  enhancement
  LogitAI/Gemstone#57 · Filter by author b-re-w opened on Oct 2, 2026 · M3 Usable Ollama replacement

• [Feature] Implement on-device Python model execution

  LogitAI/Gemstone#42 · Filter by author b-re-w opened on Aug 6, 2025

• [Chore] Report Jetpack Compose Web/Desktop input bugs

  LogitAI/Gemstone#41 · Filter by author b-re-w opened on Aug 6, 2025
  1