All pull requests

Pull requests

• fix(usage): sum duration_ms when adding Usage objects

  #690·pcbeingused333 opened Nov 3, 2026Contributor·

  ·

• feat: add Runcrate as an OpenAI-compatible provider

  #689·heydryft opened Nov 3, 2026·

  ·

• fix(requesty): implement api_embed_execute so embed works

  #688·pcbeingused333 opened Nov 3, 2026Contributor·

  ·

• Keep named criteria lists on a suite, layered by key

  #686·TonsOfFun opened Nov 2, 2026Contributor·3 tasks done·

  ·

• Verify a RubyLLM model or judge can be built before a run

  #685·TonsOfFun opened Nov 2, 2026Contributor·4 tasks done·

  ·

• Publish with less glue: run_id from the report, absolute receipt urls, from_env

  #684·TonsOfFun opened Nov 2, 2026Contributor·3 tasks done·

  ·

• Replay each scenario a configurable number of trials

  #683·TonsOfFun opened Nov 2, 2026Contributor·4 tasks done·

  ·

• Assert a report's pass gates from TestHelper and RSpec

  #682·TonsOfFun opened Nov 2, 2026Contributor·2 tasks done·

  ·

• Skip the judge for a replay that already failed its scenario's expectations

  #681·TonsOfFun opened Nov 2, 2026Contributor·3 tasks done·

  ·

• Select scenarios by key strictly, so a misspelled key fails the run

  #680·TonsOfFun opened Nov 2, 2026Contributor·3 tasks done·

  ·

• Build the regex for a count or a share a scenario expects

  #679·TonsOfFun opened Nov 2, 2026Contributor·3 tasks done·

  ·

• Let a suite declare its pass gates per group

  #676·TonsOfFun opened Nov 2, 2026Contributor·5 tasks done·

  ·

• fix(usage): include cache tokens in Anthropic total_tokens

  #675·pcbeingused333 opened Nov 2, 2026Contributor·

  ·

• Fail a run as lost when its worker stops

  #674·yorzi opened Nov 2, 2026Contributor·14 of 19 tasks·

  ·

• Keep an agent's own tool observer under Capture

  #673·yorzi opened Nov 2, 2026Contributor·11 of 19 tasks·

  ·

• Cancel a run only while it is still in progress

  #672·yorzi opened Nov 2, 2026Contributor·11 of 19 tasks·

  ·

• docs(mcp): add scoped local code inspection with Sem

  #671·rs545837 opened Nov 2, 2026·

  ·

• fix(model-capabilities): strip top_k for thinking-first models

  #669·pcbeingused333 opened Oct 31, 2026Contributor·

  ·

• Let the Anthropic provider load the anthropic gem 2.x

  #667·TonsOfFun opened Oct 31, 2026Contributor·

  ·

• Spike: a System One judge (Jev, or a local OpenJevX) for scenario evaluations

  #660·TonsOfFun opened Oct 30, 2026Contributor·6 tasks·

  ·

• Update anthropic requirement from ~> 1.12 to ~> 2.3

  dependenciesruby
  #640·dependabot[bot] opened Oct 27, 2026Bot·

  ·

• Add optional approved MemCode memory actions for Rails agents

  #518·vivekgupta-memcode opened Oct 1, 2026·

  ·