All pull requests

Pull requests

• docs: A provider stream's mcp_list_tools items match the final response; internal servers' items stay off it

  #586·github-actions[bot] opened Nov 15, 2026Bot·2 of 3 tasks·

  ·
  1

• docs: version-context: drop the since-v1.13.0 count_tokens anchor on the PD getting-started page

  #585·github-actions[bot] opened Nov 15, 2026Bot·2 of 3 tasks·

  ·
  1

• docs: A PD side's own 500 excludes that side's worker alone; retries.md says the pair is excluded together

  #584·github-actions[bot] opened Nov 15, 2026Bot·2 of 3 tasks·

  ·
  1

• docs: A batched embeddings request records one breaker sample: the worst answer by the worker's reading

  #583·github-actions[bot] opened Nov 15, 2026Bot·2 of 3 tasks·

  ·
  1

• docs: A structural_tag response_format rides to engines as sent; the pipeline and API pages compile everything to a...

  #582·github-actions[bot] opened Nov 14, 2026Bot·2 of 3 tasks·

  ·
  2

• docs: --grouped-dp-sticky hash pins a sticky key's request to one engine of a grouped ZMQ worker; docs say never pinned

  #581·github-actions[bot] opened Nov 14, 2026Bot·2 of 3 tasks·

  ·
  2

• docs: PD streams anchor smg_router_ttft_seconds at the prefill dispatch; docs say the clock starts at decode dispatch

  #580·github-actions[bot] opened Nov 14, 2026Bot·2 of 3 tasks·

  ·
  2

• docs: The engine's advertised media limits live in the mm_item_limits worker label; the page never names it

  #579·github-actions[bot] opened Nov 14, 2026Bot·2 of 3 tasks·

  ·
  2

• docs: /generate validates like chat: sampling rules run and a refused body answers the gateway's error envelope

  #578·github-actions[bot] opened Nov 14, 2026Bot·2 of 3 tasks·

  ·
  2

• docs: The SGLang language-side encode dispatch carries video and audio references; the images-only rule is the gRPC...

  #577·github-actions[bot] opened Nov 14, 2026Bot·2 of 3 tasks·

  ·
  2

• docs: The chart mounts the rate-limit and scheduler YAML from ConfigMaps; the Helm section never says how

  #576·github-actions[bot] opened Nov 14, 2026Bot·2 of 3 tasks·

  ·
  2

• docs: GET input_items pages by limit, order and after, and the object read refuses stream=true; docs say one page

  #575·github-actions[bot] opened Nov 14, 2026Bot·2 of 3 tasks·

  ·
  2

• docs: /v1/tokenize and /v1/detokenize refuse in the gateway envelope; an unknown model is 404 model_not_found

  #574·github-actions[bot] opened Nov 14, 2026Bot·2 of 3 tasks·

  ·
  2

• docs: WASM module add and remove failures answer by cause (400/503/504, 409 in-flight); docs say a failed add is 400

  #573·github-actions[bot] opened Nov 14, 2026Bot·2 of 3 tasks·

  ·
  2

• docs: A control-plane mutation runs to its outcome line through client departure and shutdown; a panic logs 500

  #572·github-actions[bot] opened Nov 14, 2026Bot·2 of 3 tasks·

  ·
  2

• docs: Every mcp_tool_result in a Messages MCP answer writes is_error (false on success); the example shows null

  #571·github-actions[bot] opened Nov 14, 2026Bot·2 of 3 tasks·

  ·
  2

• docs: The grpc-server Cargo feature is gone; the dev guide's feature table still calls it a marker

  #570·github-actions[bot] opened Nov 14, 2026Bot·2 of 3 tasks·

  ·
  2

• docs(discovery): the -slurm router image variant carries the Slurm clients

  #519·hello-alexmcc opened Nov 10, 2026Contributor·

  ·
  1