All pull requests

Pull requests

• API: Always return usage for non-streaming completions

  #487·blaakstone opened Sep 24, 2026·

  ·

• Add opt-in prefill progress streaming (prompt_progress SSE chunks)

  #486·eliaskg opened Sep 24, 2026·

  ·
  1

• 🛡️ Fail closed on un-compilable grammar constraints; re-rail whitespace stalls

  #481·sdougbrown opened Sep 17, 2026·

  ·

• Add Spark X2.5 tool call format (spark2_5)

  #480·dfein38347g opened Sep 17, 2026·

  ·
  1

• API: Add llama.cpp-compatible timings to completion responses

  #478·midagedev opened Sep 15, 2026·

  ·

• Log the recurrent-state slot allocation at load; document max_batch_size for recurrent models

  #467·matthematics1137 opened Sep 10, 2026Contributor·

  ·

• Allow ExLlamaV3 on Turing (sm_75), keep Volta rejected

  #463·vcruz305 opened Sep 4, 2026·

  ·

• API: Stream partial tool_calls argument deltas for qwen3_coder formats

  #462·tengotengo opened Sep 2, 2026·

  ·
  1

  3

• Feature/metrics inspector

  #444·klement opened Aug 1, 2026·

  ·
  2

• feature/metrics

  #443·klement opened Aug 1, 2026·

  ·

• feature/anthropic messages api

  #442·klement opened Aug 1, 2026·

  ·

• docs(sampling): clarify that sampler override presets resolve from the working directory (#419)

  #435·Anai-Guo opened Jul 24, 2026Contributor·

  ·
  1

• Lora endpoint

  #434·suspicious-pineapple opened Jul 21, 2026Contributor·

  ·

• [refactor] chore: migrate from typing.{List,Union,Optional,Dict,Set} to builtin syntax

  #394·AlpinDale opened Oct 31, 2025Member·

  ·
  67

• feat: log the current commit hash

  #233·AlpinDale opened Nov 4, 2024Member·

  ·
  1

• Update model.py, to time the model loading time, to create a metric for system tuning to aim for, to optimize the model loading speed

  #222·awatuna opened Oct 18, 2024·

  ·

• Docs

  #214·SecretiveShell opened Sep 29, 2024Contributor·

  ·

• Pydantic config p2

  #213·SecretiveShell opened Sep 28, 2024Contributor·

  ·

• rewrite API keys implementation

  #190·SecretiveShell opened Sep 6, 2024Contributor·

  ·
  1

• Fix logprobs when multiple tokens are returned at once.

  #141·zewt opened Jun 24, 2024Contributor·

  ·
  2

• Added support for multiple API keys, generating OAI-like tokens by default

  #99·kir-gadjello opened May 1, 2024·

  ·