All pull requests

Pull requests

• fix(streaming): surface MCP metadata on sync final chunk when include_usage

  #49568·yuee3 opened Nov 1, 2026·3 of 4 tasks·

  ·

• feat(providers): add HopBase as a JSON-configured OpenAI-compatible provider

  #49566·victor-zzh opened Nov 1, 2026·

  ·
  2

• fix(a2a): let an agent refuse caller credentials sent through convention headers

  #49564·rameswaran-mohan opened Nov 1, 2026Contributor·4 of 6 tasks·

  ·
  1

• fix(caching): key the response cache on api_base and api_key

  #49563·sunhouy opened Nov 1, 2026·4 of 5 tasks·

  ·
  1

• fix(otel): record failed call duration with error.type in OTel v1

  #49561·yashs33244 opened Nov 1, 2026·4 of 6 tasks·

  ·
  1

• fix(otel)!: bound the V1 hidden_params metric label to model_id

  #49560·yashs33244 opened Nov 1, 2026·4 of 6 tasks·

  ·
  1

• fix(otel): emit openai.request/response.service_tier on v2 LLM spans

  #49559·yashs33244 opened Nov 1, 2026·4 of 6 tasks·

  ·
  1

• fix(otel): restore OpenInference cache and reasoning token details in v2 mapper

  #49558·yashs33244 opened Nov 1, 2026·4 of 6 tasks·

  ·
  1

• fix(otel): emit gen_ai.usage.reasoning.output_tokens on v2 LLM spans

  #49557·yashs33244 opened Nov 1, 2026·4 of 6 tasks·

  ·
  1

• test(spend): cover key worker cancellation and Redis replay

  #49556·yujonglee-berri opened Nov 1, 2026Contributor·5 tasks done·

  ·
  5

  

• test(integration): backport five integration flake fixes to rc/1.108.0

  #49555·yuneng-berri opened Nov 1, 2026Contributor·1 of 6 tasks·

  ·

• test(integration): fix four flaky integration tests

  #49554·yuneng-berri opened Nov 1, 2026Contributor·1 of 6 tasks·

  ·
  1

• fix(proxy): backport the /v1/messages retry spend log fix and team delete test fix to rc/1.108.0

  run-ci
  #49553·yuneng-berri opened Nov 1, 2026Contributor·1 of 6 tasks·

  ·

• fix(google_genai): emit tool messages for functionResponse parts in model turns

  #49552·rejasupotaro opened Nov 1, 2026·5 of 6 tasks·

  ·

• feat(routerhub): add compatible Chat Completions provider

  #49550·UCANGETNOINFO opened Nov 1, 2026·3 of 6 tasks·

  ·
  1

• feat(scim): place SCIM groups in the organization a configured group attribute names

  #49549·jesus-berri opened Nov 1, 2026Contributor·4 of 6 tasks·

  ·
  6

  

• fix(cost-map): correct three novita prices to the pricing page

  #49548·Liutasil501 opened Nov 1, 2026·

  ·
  4

• fix(cost-map): correct four deepinfra prices to the pricing page

  #49547·Liutasil501 opened Nov 1, 2026·

  ·
  1

• refactor(ui): move admin and settings text onto i18n

  #49546·yuneng-berri opened Nov 1, 2026Contributor·4 of 6 tasks·

  ·
  3

• feat(proxy): add POST /key/revoke so anyone who finds a published key can turn it off

  #49545·yuneng-berri opened Nov 1, 2026Contributor·5 of 6 tasks·

  ·
  4

• feat(spend-logs): opt-in latest_turn mode that stores only the current turn of stateless requests

  #49544·jesus-berri opened Nov 1, 2026Contributor·4 of 6 tasks·

  ·
  3

  

• feat: add EmpirioLabs models to model_prices_and_context_window.json

  #49539·Adam-Dalloul opened Oct 31, 2026Contributor·

  ·
  1

• fix(responses): carry logprobs across the chat completions bridge

  #49536·mateo-berri opened Oct 31, 2026Contributor·6 tasks done·

  ·
  17

  

• feat(providers): add Turion decisions and extraction provider

  #49535·balys opened Oct 31, 2026·4 of 6 tasks·

  ·

• fix(main)!: carry the full per-call price entry on the logging object instead of writing litellm.model_cost

  #49532·kerry-berri opened Oct 31, 2026Contributor·4 of 5 tasks·

  ·
  16