All pull requests

Pull requests

• site: add Harvey LAB post-training blog post

  #442·ShubyM opened Nov 6, 2026Collaborator·

  ·
  2

• examples/terraform/text-to-sql: a GKE cluster for the text-to-SQL recipe

  #441·linde opened Nov 5, 2026Collaborator·

  ·

• workers: an experimental temp root on a node's local SSD, for single-…

  #440·droot opened Nov 5, 2026Collaborator·

  ·

• observability: live worker state, served uncached at /inspect/now

  #439·droot opened Nov 5, 2026Collaborator·

  ·

• k8s/deploy/demo: live lights, an idle fleet view, and the recipe on t…

  #438·droot opened Nov 5, 2026Collaborator·

  ·

• delta sync: a sampler reads a sparse delta with parallel reads

  #436·droot opened Nov 5, 2026Collaborator·

  ·

• perf(sampler): size TPU sampler batch limits from the chip's HBM

  #435·hodaaaaaaaaaa opened Nov 5, 2026Collaborator·

  ·

• docs: TPU on GKE setup guide

  #419·hodaaaaaaaaaa opened Nov 3, 2026Collaborator·

  ·

• Run telemetry and the dashboard

  #383·ShubyM opened Oct 26, 2026Collaborator·

  ·

• docs: address the sampling pools design review

  #379·ShubyM opened Oct 23, 2026Collaborator·

  ·

• perf(deploy): pre-pull the worker image on every GPU node

  #373·hodaaaaaaaaaa opened Oct 23, 2026Collaborator·

  ·

• perf: prefer GPU types per job with openrl.gpu_types

  #372·hodaaaaaaaaaa opened Oct 23, 2026Collaborator·

  ·

• perf(trainer): write a LoRA sampler save in the background and answer once it is on disk

  #370·hodaaaaaaaaaa opened Oct 23, 2026Collaborator·

  ·

• perf(trainer): save a LoRA adapter once per step, not after every optimizer step

  #369·hodaaaaaaaaaa opened Oct 23, 2026Collaborator·

  ·

• feat(rl): cross-precision RL off-policy corrections (top-p replay, MIS, score centering)

  #345·sunnythecode opened Oct 20, 2026Collaborator·8 tasks·

  ·

• Require explicit trainer/sampler precision; never fall back

  #344·sunnythecode opened Oct 19, 2026Collaborator·

  ·
  2

• chore(deps-dev): bump the minor-and-patch group across 1 directory with 2 updates

  dependencies
  #333·dependabot[bot] opened Oct 14, 2026Bot·

  ·
  2

• feat: custom trainer image

  #323·ShubyM opened Oct 8, 2026Collaborator·

  ·

• chore(deps): bump astral-sh/setup-uv from 5.4.2 to 10.2.0

  dependencies
  #279·dependabot[bot] opened Sep 30, 2026Bot·

  ·
  1

• chore(deps): bump docker/build-push-action from 5.4.0 to 7.4.0

  dependencies
  #278·dependabot[bot] opened Sep 30, 2026Bot·

  ·
  1

• chore(deps): batch open dependabot PRs

  #270·barney-s opened Sep 25, 2026·

  ·
  1

  2

• chore(deps): bump anyio from 4.13.0 to 4.14.2

  dependenciespython:uv
  #267·dependabot[bot] opened Sep 24, 2026Bot·

  ·

• chore(deps): bump actions/setup-go from 5.5.0 to 7.0.0

  dependencies
  #263·dependabot[bot] opened Sep 23, 2026Bot·

  ·
  1

• chore(deps): bump docker/login-action from 3.7.0 to 4.6.0

  dependencies
  #261·dependabot[bot] opened Sep 23, 2026Bot·

  ·
  1

• chore(deps): bump docker/metadata-action from 5.10.0 to 6.2.0

  dependencies
  #260·dependabot[bot] opened Sep 23, 2026Bot·

  ·
  1