Commits · afdc1d284f52f186a88e7402f74fb1cb9c139df1 · nexlab / coderai · GitLab

Oct 07, 2026
- longcat: the quantised-encoder estimate was 3x too optimistic (v0.2.79) · a24c8b1f
  Stefy Lanza (nextime / spora ) authored Oct 07, 2026
  a24c8b1f
- longcat: quantise the text encoder, so nothing has to leave the card (v0.2.77) · 3b384924
  Stefy Lanza (nextime / spora ) authored Oct 07, 2026
  3b384924
Oct 06, 2026
- longcat: size the load by what stays on the card, and install accelerate (v0.2.71) · 97a5302c
  Stefy Lanza (nextime / spora ) authored Oct 06, 2026
  97a5302c
Oct 05, 2026
- free disk: find the weights engines keep outside the caches (v0.2.43) · 9ba73bb3
  Stefy Lanza (nextime / spora ) authored Oct 05, 2026
  9ba73bb3
- longcat: VRAM accounting, a draining releaser, and video in the front queue (v0.2.32) · 409880f3
  Stefy Lanza (nextime / spora ) authored Oct 05, 2026
  409880f3
Oct 02, 2026
- ocr: olmOCR-2 as an alternative engine, and stop the Surya-2 crash loop (v0.2.24) · 75200c0b
  Stefy Lanza (nextime / spora ) authored Oct 02, 2026
  75200c0b
Sep 19, 2026
- cluster: nodes as engines, one GGUF over llama.cpp RPC, vLLM/SGLang... · f5ca2d82
  Stefy Lanza (nextime / spora ) authored Sep 19, 2026
  f5ca2d82
Sep 16, 2026
- admin: make 'runs only on RunPod' a visible placement choice; host backend gets a form · 362ff475
  Stefy Lanza (nextime / spora ) authored Sep 16, 2026
  362ff475
Sep 14, 2026
- runpod: send local text adapters to the pod that can receive them · a746e0fe
  Stefy Lanza (nextime / spora ) authored Sep 14, 2026
  a746e0fe
- text: LoRA and QLoRA on text models, locally and on RunPod · 34d08890
  Stefy Lanza (nextime / spora ) authored Sep 14, 2026
  34d08890
Sep 13, 2026
- llm: make every text model remotizable, GGUF included · 6c6646c4
  Stefy Lanza (nextime / spora ) authored Sep 13, 2026
  6c6646c4
Sep 12, 2026
- manager: don't evict a model to make room for itself · 2a8bb4c2
  Stefy Lanza (nextime / spora ) authored Sep 12, 2026
  2a8bb4c2
- manager: match a sibling config by its alias, not through the alias map · 5c284c97
  Stefy Lanza (nextime / spora ) authored Sep 12, 2026
  5c284c97
- manager: thread the requested id through the TEXT load path too · 32c1d758
  Stefy Lanza (nextime / spora ) authored Sep 12, 2026
  32c1d758
- manager: keep the requested id so sibling configs stay distinguishable · 484c88f4
  Stefy Lanza (nextime / spora ) authored Sep 12, 2026
  484c88f4
- manager: size a model by the config that was requested, not a same-path sibling · c5ae720e
  Stefy Lanza (nextime / spora ) authored Sep 12, 2026
  c5ae720e
Sep 10, 2026
- video: serve MiniMax-H3 through an isolated worker, and clone voices in the studio · 4abca4a9
  Stefy Lanza (nextime / spora ) authored Sep 10, 2026
  4abca4a9
Sep 06, 2026
- runpod(phase2): serverless request path + per-model config + no-local-download · 17604ab4
  Stefy Lanza (nextime / spora ) authored Sep 06, 2026
  17604ab4
Sep 05, 2026
- runpod(phase1): account config + GraphQL client + admin key/GPU-catalog plumbing · 7c057af2
  Stefy Lanza (nextime / spora ) authored Sep 05, 2026
  7c057af2
Sep 02, 2026
- audio(whisper-server): convert mp4/any input with our ffmpeg (fix stuck-queue hang) · 717ab1ad
  Stefy Lanza (nextime / spora ) authored Sep 02, 2026
  717ab1ad
Sep 01, 2026
- audio: CrisperWhisper support (isolated venv, verbatim + word timestamps) · b069137b
  Stefy Lanza (nextime / spora ) authored Sep 01, 2026
  b069137b
- audio: STT backends, diarization, speaker embeddings/ID, word timestamps · 7e5a925c
  Stefy Lanza (nextime / spora ) authored Sep 01, 2026
  7e5a925c
Aug 27, 2026
- vram: don't ratchet down a model's learned GPU-layer split under contention · f54428cf
  Stefy Lanza (nextime / spora ) authored Aug 27, 2026
  f54428cf
- config(multi-config): fix GGUF dispatch + per-config-id measured persistence · f9f2a373
  Stefy Lanza (nextime / spora ) authored Aug 27, 2026
  f9f2a373
- rerank: add /v1/rerank cross-encoder endpoint + surface OCR in /v1/models · e44d41ac
  Stefy Lanza (nextime / spora ) authored Aug 27, 2026
  e44d41ac
Aug 21, 2026
- vllm: serve models from the model list (nvidia/radeon-style), model_id optional · 7dee70a4
  Stefy Lanza (nextime / spora ) authored Aug 21, 2026
  7dee70a4
- vllm: cooperative VRAM co-tenancy + GPU selection (full engine parity) · 4223107e
  Stefy Lanza (nextime / spora ) authored Aug 21, 2026
  4223107e
- Add vLLM as a first-class managed engine backend (LLM + OCR Surya2) · a5d9fe82
  Stefy Lanza (nextime / spora ) authored Aug 21, 2026
  a5d9fe82
Aug 16, 2026
- Add dedicated OCR subsystem + finalize multi-engine backends · 2485403f
  Stefy Lanza (nextime / spora ) authored Aug 16, 2026
  2485403f
Jul 29, 2026
- thermal: CPU-hot pauses ALL engines (incl colibri) — revert per-engine exemption · 8ff477c1
  Stefy Lanza (nextime / spora ) authored Jul 29, 2026
  8ff477c1
Jul 28, 2026
- thermal: engine-level CPU wait also gated on per-engine CPU relevance · c44376a9
  Stefy Lanza (nextime / spora ) authored Jul 28, 2026
  c44376a9
- colibri: GLM-5.2 tool-call parsing + treat colibri as a normal VRAM-eviction citizen · e2081488
  Stefy Lanza (nextime / spora ) authored Jul 28, 2026
  e2081488
- colibri: integrate GLM-5.2 native C engine (driven directly, no colibri Python) · 50f54eb3
  Stefy Lanza (nextime / spora ) authored Jul 28, 2026
  50f54eb3
Jul 24, 2026
- fix: config-key shadow dropped mmproj on GGUF vision model reload · 1126bdbc
  Stefy Lanza (nextime / spora ) authored Jul 24, 2026
  1126bdbc
- thermal: filter impossible GPU temps at the source (gpu_detect + read chokepoint) · 9ff0f709
  Stefy Lanza (nextime / spora ) authored Jul 24, 2026
  9ff0f709
Jul 23, 2026
- thermal: ignore physically impossible GPU temperature readings · 3c9536e1
  Stefy Lanza (nextime / spora ) authored Jul 23, 2026
  3c9536e1
- dinov2.cpp GGUF backend on Vulkan + Tasks-page compute-device badges · 1ca4a83b
  Stefy Lanza (nextime / spora ) authored Jul 23, 2026
  1ca4a83b
Jul 22, 2026
- video: fix VRAM teardown leak + track in-flight loads with reservations · 2ea24b0b
  Stefy Lanza (nextime / spora ) authored Jul 22, 2026
  2ea24b0b
- turboquant: ship turboquant-py + accurate bit-width errors for the library backend · 4d2aeb69
  Stefy Lanza (nextime / spora ) authored Jul 22, 2026
  4d2aeb69
Jul 03, 2026
- lora-train: cross-engine GPU lock so nothing reloads mid-training (fix intermittent OOM) · 18c6a358
  Stefy Lanza (nextime / spora ) authored Jul 03, 2026
  18c6a358