Skip to content

Pull requests: Luce-Org/lucebox-hub

New pull request New
37 Open 235 Closed
37 Open 235 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

fix(server): support gemma-4's plain-text call:<verb>{} tool-call format
#329 opened Jun 1, 2026 by easel Collaborator • Draft
8
feat(server): soft-close thinking termination via logit-ratio peek
#326 opened Jun 1, 2026 by easel Collaborator Loading…
4 tasks
3
feat(server): add DFlash disk prefix cache for target layer split
#325 opened May 31, 2026 by weicj Collaborator Loading…
12
[codex] Retry dflash generations with no visible output
#324 opened May 31, 2026 by OmarB97 Contributor Loading…
7
Add real-time /status dashboard with SSE push
#322 opened May 31, 2026 by howard0su Contributor Loading…
13
feat(server): support DFlash with mixed-backend target layer split
#321 opened May 31, 2026 by weicj Collaborator Loading…
9
[LuceBox][DFlash][lucebox-pr314-common-empty-fallback][2/n] Default empty spec retry in backend calls
#319 opened May 31, 2026 by OmarB97 Contributor Loading…
[codex] Recover dflash spec-decode agent stalls
#315 opened May 31, 2026 by OmarB97 Contributor Loading…
9
feat(server): add selectable backend IPC payload transport
#312 opened May 30, 2026 by weicj Collaborator Loading…
9
feat(server): reduce layer-split activation memory with backend precision policy
#310 opened May 29, 2026 by weicj Collaborator Loading…
5
feat(dflash): reduce feature mirror memory with dtype policy
#309 opened May 29, 2026 by weicj Collaborator Loading…
2
fix(server): route Qwen3.6/Laguna think-mode reasoning to reasoning_content channel
#308 opened May 29, 2026 by easel Collaborator Loading…
refactor(server): share target layer-split runtime helpers
#306 opened May 29, 2026 by weicj Collaborator Loading…
refactor: extract MoE hybrid mode into common layer for qwen and laguna
#305 opened May 29, 2026 by howard0su Contributor Loading…
2
Add LLM Auto Context Compaction
#304 opened May 29, 2026 by howard0su Contributor • Draft
feat(server): add Laguna target-layer-split adapter
#297 opened May 28, 2026 by weicj Collaborator Loading…
fix(server): support sampled requests in target layer split
#295 opened May 28, 2026 by weicj Collaborator Loading…
3
feat(server): passthrough proxy, piecewise keep-ratio curve, query survival check
#294 opened May 28, 2026 by smpurkis Contributor Loading…
3
feat(server): add Gemma4 draft residency support
#291 opened May 28, 2026 by weicj Collaborator Loading…
2
feat(qwen35moe): pipelined hybrid MoE decode with GPU/CPU overlap
#289 opened May 28, 2026 by howard0su Contributor Loading…
6
feat(lucebox): docker stack + CLI + bench/profile + harness + luce-bench in-tree
#285 opened May 27, 2026 by easel Collaborator Loading…
26
fix(server): Qwen3.6-27B tool calling for claude-code Anthropic path
#276 opened May 25, 2026 by dusterbloom Collaborator Loading…
5 of 7 tasks
4
feat(drafter): ee3 as production default (depends on #274)
#275 opened May 24, 2026 by dusterbloom Collaborator • Draft
6
feat(pflash): prefill compress up to 128k -> 2-12× prefill (content-dependent), decode at parity
#274 opened May 24, 2026 by dusterbloom Collaborator Loading…
38
feat(harness): typed adapters + format-aware session-inject proxy + multi-turn bandit driver
#266 opened May 23, 2026 by dusterbloom Collaborator Loading…
5
ProTip! Follow long discussions with comments:>50.