Commits

Branch selector

 main

User selector

All users

All time

Commit History

Commits on Mar 7, 2026

• feat(server): add audio endpoints and multimodal embedding input

  Show description for 95075a7
  farhan-syah

  committed
  95075a7

  Copy full SHA for 95075a7

• feat(loader): add vision projector loader for multimodal GGUF models

  Show description for ef03e96
  farhan-syah

  committed
  ef03e96

  Copy full SHA for ef03e96

• feat(engine): add multimodal inference executor for vision and audio input

  Show description for 72a5b11
  farhan-syah

  committed
  72a5b11

  Copy full SHA for 72a5b11

• chore(deps): add image crate for vision input decoding

  farhan-syah

  committed
  109ff6d

  Copy full SHA for 109ff6d

• feat(server): wire multimodal generation through chat completions handler

  Show description for d8529e3
  farhan-syah

  committed
  d8529e3

  Copy full SHA for d8529e3

• feat(server): implement audio decoding for PCM16 and WAV formats

  Show description for 1086651
  farhan-syah

  committed
  1086651

  Copy full SHA for 1086651

• feat(loader): add vision and audio fields to model config structs

  Show description for c561d89
  farhan-syah

  committed
  c561d89

  Copy full SHA for c561d89

• feat(server): support OpenAI multimodal message content format

  Show description for 819d30c
  farhan-syah

  committed
  819d30c

  Copy full SHA for 819d30c

• feat(server): add embeddings, rerank, responses, Anthropic, and SLO endpoints

  Show description for 37d5453
  farhan-syah

  committed
  37d5453

  Copy full SHA for 37d5453

• feat(engine): add data parallel, cache routing, embedding, and bench modules

  Show description for 1ecfb58
  farhan-syah

  committed
  1ecfb58

  Copy full SHA for 1ecfb58

• feat(loader): expand GGUF architecture support and fix AttentionConfig fields

  Show description for 083df85
  farhan-syah

  committed
  083df85

  Copy full SHA for 083df85

Commits on Mar 6, 2026

• fix(engine): resolve device type mismatch in prefix cache construction

  Show description for 106155c
  farhan-syah

  committed
  106155c

  Copy full SHA for 106155c

• feat(engine): gate tensor parallelism behind dedicated nccl feature

  Show description for 16b34a2
  farhan-syah

  committed
  16b34a2

  Copy full SHA for 16b34a2

• feat(distributed): add distributed runtime module and wire top-level entry point

  Show description for 1ca8c65
  farhan-syah

  committed
  1ca8c65

  Copy full SHA for 1ca8c65

• feat(server): integrate continuous batching scheduler and add LoRA management endpoint

  Show description for 678ccf4
  farhan-syah

  committed
  678ccf4

  Copy full SHA for 678ccf4

• feat(cli): extract command definitions and add distributed inference modes

  Show description for d5d01f0
  farhan-syah

  committed
  d5d01f0

  Copy full SHA for d5d01f0

• feat(engine): add speculative decoding, LoRA, MoE offload, grammar, and tensor parallelism

  Show description for 5c0649c
  farhan-syah

  committed
  5c0649c

  Copy full SHA for 5c0649c

• refactor(loader): extract API loader and split safetensors loaders module

  Show description for 0aea40a
  farhan-syah

  committed
  0aea40a

  Copy full SHA for 0aea40a

• refactor(tokenizer): extract TokenizerTrait and splintr adapter into separate files

  Show description for 863b3f9
  farhan-syah

  committed
  863b3f9

  Copy full SHA for 863b3f9

• feat(config): extend inference and generation config with advanced options

  Show description for aaaf1b9
  farhan-syah

  committed
  aaaf1b9

  Copy full SHA for aaaf1b9

• feat(deps): add distributed feature flag and pin boostr to registry version

  Show description for ea96fb1
  farhan-syah

  committed
  ea96fb1

  Copy full SHA for ea96fb1

Commits on Mar 5, 2026

• feat(engine,config): add prefix cache support for paged attention

  Show description for 1b5ae44
  farhan-syah

  committed
  1b5ae44

  Copy full SHA for 1b5ae44

• feat(server): add fill-in-the-middle endpoint at POST /v1/infill

  Show description for f6730fe
  farhan-syah

  committed
  f6730fe

  Copy full SHA for f6730fe

• feat(server): add OpenAI-compatible tool calling to chat completions

  Show description for e25ee04
  farhan-syah

  committed
  e25ee04

  Copy full SHA for e25ee04

• feat(tokenizer): add special_token_id lookup to TokenizerTrait

  Show description for 9812307
  farhan-syah

  committed
  9812307

  Copy full SHA for 9812307

• refactor(server): extract startup logic into dedicated module

  Show description for 0388c5e
  farhan-syah

  committed
  0388c5e

  Copy full SHA for 0388c5e

• refactor(engine): split warmup and generate_text into dedicated submodules

  Show description for 3c1b52d
  farhan-syah

  committed
  3c1b52d

  Copy full SHA for 3c1b52d

• feat(metrics): add decode slot and token budget utilization gauges

  Show description for 2e800ac
  farhan-syah

  committed
  2e800ac

  Copy full SHA for 2e800ac

• feat(server): add inference slot management API

  Show description for 3f0e2f8
  farhan-syah

  committed
  3f0e2f8

  Copy full SHA for 3f0e2f8

• refactor(executor): replace ad-hoc debug logs with structured phase tracing

  Show description for f65ce1b
  farhan-syah

  committed
  f65ce1b

  Copy full SHA for f65ce1b

• test(api): add integration tests for server API endpoints

  Show description for fac466e
  farhan-syah

  committed
  fac466e

  Copy full SHA for fac466e

• refactor(server): split handlers into completion, chat, and generation modules

  Show description for fa2011c
  farhan-syah

  committed
  fa2011c

  Copy full SHA for fa2011c

• feat(config): add advanced sampling parameters to GenerationConfig

  Show description for 12aa057
  farhan-syah

  committed
  12aa057

  Copy full SHA for 12aa057

• refactor(engine): split executor into focused submodules

  Show description for 0f44760
  farhan-syah

  committed
  0f44760

  Copy full SHA for 0f44760

• fix(loader): eliminate VarMap memory leaks via Box::leak

  Show description for 0f2e450
  farhan-syah

  committed
  0f2e450

  Copy full SHA for 0f2e450