chore: bump harbor 0.13 -> 0.19 - #76
Merged
Merged
Conversation
API surface nasde drives is unchanged (JobConfig.model_validate, Job.create, job.run(), AgentConfig import_path/kwargs/skills/mcp_servers/env, Codex CODEX_FORCE_AUTH_JSON contract). The ADR-012 native skill-injection moved upstream from BaseInstalledAgent to per-agent implementations with the same HOME-scoped targets. Unlocks ~10 new built-in agents (pi, grok-build, langgraph, deerflow, eve, mimo, computer-1, acp, vibe) and git-reference skills (org/name@ref); drops pandas/numpy/pyarrow/mcp from the tree. Gates: 436 tests green, pip-audit clean, smoke e2e on ddd-threshold-discount (claude-vanilla) with non-null cache-aware token/cost economics on a fresh 0.19 trajectory (score 0.85, 560k tokens, $0.71). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01YbqvAPmMpaQSacRLqvoQQE
szjanikowski
pushed a commit
that referenced
this pull request
Sep 12, 2026
Move [Unreleased] under 0.6.0 (2026-09-11), add the missing CHANGELOG entry for the calibration round 2 engine changes (#75: agent diff as judge input, per-task dimensions, precheck.sh hook, eval --eval-model/--eval-backend) and the [#76] link-ref on the Harbor 0.19 entry. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UwjBNRxR1vkNGBcETRzdf2
szjanikowski
added a commit
that referenced
this pull request
Sep 12, 2026
* chore: release v0.6.0 Move [Unreleased] under 0.6.0 (2026-09-11), add the missing CHANGELOG entry for the calibration round 2 engine changes (#75: agent diff as judge input, per-task dimensions, precheck.sh hook, eval --eval-model/--eval-backend) and the [#76] link-ref on the Harbor 0.19 entry. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UwjBNRxR1vkNGBcETRzdf2 * changelog: tighten the 0.6.0 entries Trim the 0.6.0 section to concise, user-facing entries: one bullet per feature with the value it brings, no implementation history or internal verification notes. Same facts, 76 lines -> 47. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UwjBNRxR1vkNGBcETRzdf2 --------- Co-authored-by: Szymon Janikowski <szymon.janikowski@itlibrium.com> Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Why
We were 6 minor versions behind (pin
>=0.13,<0.14, upstream 0.19.0). The bump unblocks the next roadmap step — adding pi (pi.dev) as a 4th agent family (Harbor 0.19 ships a built-inpiagent) — and brings ~10 new built-in Harbor agents plus git-reference skills (org/name@ref).What changed
harbor[daytona,modal,e2b,runloop,gke]>=0.19,<0.20(specific minor pin, as before)uv.lockre-resolved — dependency tree shrinks: pandas, numpy, pyarrow, mcp, dill, multiprocess, xxhash no longer pulled in[Unreleased]Compatibility audit (0.19 vs 0.13)
JobConfig.model_validate/Job.create/job.run()AgentConfig(import_path,kwargs,skills,mcp_servers,env)harbor.models.trajectories.step.Step(Opik monkeypatch)CODEX_FORCE_AUTH_JSON/CODEX_AUTH_JSON_PATH)BaseInstalledAgent._build_register_skills_commandto per-agent implementations; same$HOME/.agents/skills/~/.gemini/skillstargetsVerification
pip-audit(CI-equivalent invocation): no known vulnerabilitiesddd-threshold-discount×claude-vanilla, 1 eval repetition — trial + verifier + judge completed, score 0.85, and cache-aware token/cost economics fully populated on a fresh 0.19 ATIF trajectory (560k tokens, $0.71, cache reads 487.5k / writes 57.7k) — the highest-risk surface (trajectory shape, ADR-011/014) holdsnasde harborCLI pass-through andnasde pricing showworkMerge order
Merge this before PR #75 (calibration round 2); #75 then rebases on main (trivial CHANGELOG conflict at most).
🤖 Generated with Claude Code
https://claude.ai/code/session_01YbqvAPmMpaQSacRLqvoQQE