All pull requests

Pull requests

• Fix Perplexity requests failing after Sonar moved to the Agent API

  #5703·ransomr opened Oct 5, 2026Collaborator·

  ·
  1

• Fix type checking and a test with jsonpath-ng 1.9.0

  qualified
  #5702·ransomr opened Oct 5, 2026Collaborator·

  ·
  1

  

• Sandbox agent bridge: serve unknown model names with the eval's model

  #5701·ransomr opened Oct 5, 2026Collaborator·

  ·
  

• Ignore trailing whitespace in answer("line")

  #5700·DeepanshuPal opened Oct 5, 2026·

  ·
  1

  

• Read Hub-delivered eval.yaml as UTF-8 in HF task loading

  #5688·kittimzhe opened Oct 5, 2026·

  ·
  1

  

• Report max-token truncation as incomplete in the Responses agent bridge

  #5685·theSatvik opened Oct 4, 2026·

  ·
  1

  

• fix: preserve model role attribution in scout workers

  #5684·rasmusfaber opened Oct 4, 2026Contributor·

  ·

• Fix OpenAI and Anthropic requests failing during credential refresh

  qualified
  #5679·rasmusfaber opened Oct 4, 2026Contributor·

  ·
  

• Fix/stream convert preserve error

  #5675·aspire488 opened Oct 4, 2026Contributor·

  ·
  1

  1

  

• fix(usage): preserve cached input tokens across provider bridges

  #5672·aspire488 opened Oct 4, 2026Contributor·

  ·
  1

  3

  

• fix(tools): convert structured validation failures to parsing errors

  #5671·aspire488 opened Oct 4, 2026Contributor·

  ·
  1

  1

  

• Let tasks and samples carry a description, recorded in the log and dataframes

  qualified
  #5667·dragonstyle opened Oct 3, 2026Collaborator·

  ·
  1

  1

  

• Fix Antigravity host tool calls being denied as unproposed by the sandbox agent bridge

  qualified
  #5658·sjawhar opened Oct 2, 2026Contributor·

  ·
  

• Design: Working time under concurrency within a sample

  #5653·ransomr opened Oct 1, 2026Collaborator·

  ·

• Fix tool approval deciding on a different call from the one that runs

  #5652·ransomr opened Oct 1, 2026Collaborator·

  ·

• Fix approvers' modify decisions running the wrong call or logging the wrong arguments

  qualified
  #5651·ransomr opened Oct 1, 2026Collaborator·

  ·
  

• Agent bridge follows the eval's configuration for service tier, storage, truncation and tool options

  qualified
  #5650·ransomr opened Oct 1, 2026Collaborator·

  ·
  1

  

• Fix usage undercounted when a provider call sends several billed requests

  #5643·ransomr opened Oct 1, 2026Collaborator·

  ·

• Fix sample limits being ignored when exceeded inside tool calls and MCP sampling

  #5642·ransomr opened Oct 1, 2026Collaborator·

  ·

• Fix HF provider generating concurrent requests with another model's settings

  qualified
  #5641·ransomr opened Oct 1, 2026Collaborator·

  ·
  1

  

• Fix Perplexity responses picking up another request's citations and usage

  qualified
  #5638·ransomr opened Oct 1, 2026Collaborator·

  ·
  

• Add the shared shard-set walk for eval sharding (internal)

  qualified
  #5622·ransomr opened Sep 30, 2026Collaborator·

  ·
  4

  

• Design: compose-prefix verification mode and decisions for the S3 incremental flush design

  qualified
  #5608·ransomr opened Sep 29, 2026Collaborator·

  ·
  

• Record dataset revision in eval log header

  qualified
  #5555·MattFisher opened Sep 25, 2026Contributor·

  ·
  1

  

• Preserve internal state across multiple assistant text blocks

  #5553·deepujain opened Sep 24, 2026Contributor·

  ·
  1

  1