release: v0.8.2#480
Merged
Merged
Conversation
Stop publishing Orbit task records on orbit-cli.com
Planned-By: codex
…n [ORB-00284] Planned-By: codex Implemented-By: codex
Clean up redundant low-value unit tests
Code-quality cleanup for ORB-00276 v2 state SQLite migration
orbit-knowledge was built as a versioned store (content-addressed objects + mutable refs + locks) but agents only use it as a derived index. Drift between the JSON object store and the SQLite sidecar ships stale results, and the lock protocol can't coordinate same-branch worktrees. Propose orbit-graph: per-worktree SQLite, regenerable from disk, <=10k LOC. V1 is read-only. V2 write surface (Rename/ReplaceBody/Move with a working-graph overlay) is sketched in 3_vision.md but deliberately not in the V1 contract. Spec lives at specs/GRAPH_SPEC.md. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Planned-By: codex
Allocate ADRs for orbit-graph design decisions
… [ORB-00297] Planned-By: codex
Scaffold tools/graph-equiv binary for v1↔v2 equivalence checks
…00296] Planned-By: codex
Establish bench/baselines.json for orbit-graph perf gate
…yp… [ORB-00298] Create orbit-graph-extract crate skeleton (Extractor trait + types) Planned-By: codex
Create orbit-graph-extract crate skeleton (Extractor trait + types)
Publish semantic companion assets in release workflow
…owledge and outline the migration plan. Emphasize that the decision to phase out orbit-knowledge will depend on effectiveness measurements, not a predetermined outcome.
…tr… [ORB-00300] Move Selector parser from orbit-knowledge to orbit-graph-extract Planned-By: codex
Move Selector parser from orbit-knowledge to orbit-graph-extract
[codex] Port rust extractor into orbit-graph-extract [ORB-00301]
Port c + markdown + config extractors into orbit-graph-extract
…csharp-extractors [codex] Port Java, Kotlin, and C# extractors [ORB-00304]
Add min_confidence parameter to orbit-graph impact and trace queries
…CP… [ORB-00329] orbit-graph polish batch: golden tests, callees coverage, MCP schema parity, UX warts Planned-By: codex
orbit-graph polish batch: golden tests, callees coverage, MCP schema parity, UX warts
Planned-By: codex
Detached-HEAD DB layout: ADR + implement decision
…-extraction [ORB-00330] Extract Rust clap commands for trace
Selective adoption from the league design mock without backend schema changes. - Add narrative tagline above the leaderboard (canonical-agent leader callouts; skipped when window total < 10 created tasks to avoid misleading on quiet runs) - Replace the cramped duel matrix with a CSS-grid reskin (proportional w/l bars, per-agent identity colors) - Add an insights panel with 5 rules (leader / watch / cold / surprise / coverage) that each self-skip when the underlying signal is missing - Add a 5-segment window selector (1h/24h/7d/30d/all), shipped inert (aria-disabled) — phase 2 (ORB-00337) wires the backend - Introduce 8 per-agent color tokens (codex/claude/gemini/grok + low variants) in :root No data-contract changes; ScoreboardSummary v5 untouched.
… c… [ORB-00338] orbit-graph cutover: equivalence + perf eval, ADR, swap legacy call sites Planned-By: codex
orbit-graph cutover: equivalence + perf eval, ADR, swap legacy call sites
Wires the phase-1 window selector to a real `?window=` query.
orbit-store:
- New `ScoreboardWindow { Hour, Day, Week, Month, All }` with
`duration()`, `as_str()`, `from_str()` (canonical: 1h / 24h / 7d /
30d / all). `ScoreboardInputs.window` defaults to `All`.
- Schema v5 -> v6 (additive): top-level `window` + `window_since`.
- `generate_summary_with_inputs` gates snapshot reads on `window == All`
(pr / task_review / tokens / planning_duels JSON) and filters
timestamped slices (tasks / learnings / adrs / frictions / job_runs)
against `since`. `recent_7d` keeps its fixed 7d boundary by design.
- +4 unit tests: snapshot zeroing, audit pass-through, tasks filter,
enum round-trip.
orbit-core:
- `OrbitRuntime::generate_scoreboard_summary` now takes
`Option<ScoreboardWindow>`; threads `since` into the audit SQL
cutoffs (`audit_tool_call_counts_*`, `audit_top_tool_calls`).
orbit-dashboard:
- `?window=` via axum `Query`: missing -> `all`, unknown -> HTTP 400
with `{"error": "..."}`. Denials honor the same cutoff.
- HTML: drop `aria-disabled` + phase-1 disabled tooltip.
- JS: delegated click handler attached during render (guarded by
`dataset.wired`) refetches `/api/scoreboard?window=<seg>` and
re-renders. Boot fetch uses `?window=24h` to match the default
highlighted segment.
- UX cleanup (user-requested): drop the bottom "Attribution Cleanup"
matrix and its non-canonical columns (admin/agent/human/pro). The
canonical four-family matrix is now the only leaderboard. Removes
dead code: OTHER_SCOREBOARD_COLUMNS, CANONICAL_SCOREBOARD_SET,
scoreboardSignalForColumns.
- +4 handler tests in new api/tests/scoreboard.rs.
…340] Planned-By: codex
…er… [ORB-00339] orbit-graph plugin MCP: expose all 7 tools, remove dead pack reference Planned-By: codex
orbit.graph.show: return source as string, not raw bytes
orbit-graph plugin MCP: expose all 7 tools, remove dead pack reference
[codex] Fix CI failures on agent-main
…y … [ORB-00344] Roll back ORB-00338 cutover: restore orbit-knowledge as primary graph backend Planned-By: codex
Repair two CI breaks left by the orbit-graph cutover rollback: - orbit-core: drop graph_tool_sync_from_linked_worktree_attributes_to_worktree_branch and its now-orphaned worktree fixture/helpers. The rollback removed the orbit.graph.sync tool, but this integration test (added in 28a72d3, outside the reverted cutover commit) still called it and panicked the suite. - graph-equiv: parse current orbit-graph-cli callees output. V2CalleeEntry was reverted to expect `from_span`, but the (un-reverted) orbit-graph CalleeEdge emits a one-based `line`; read it directly. Fixes the "missing field from_span" backend errors in the equivalence harness. Also fix a stale orbit-tools crate-doc reference to a non-existent `orbit-types` crate (it is orbit-common). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Roll back ORB-00338 cutover: restore orbit-knowledge as primary graph backend
Trims six tools off the agent MCP allowlist and switches them to
`register_inactive` so they're hidden from the default `orbit tool list`
surface while staying reachable via `orbit tool run` / `runtime.run_tool`:
- `orbit.task.reject` — task rejection is a human/operator decision.
- `orbit.friction.{list,show,resolve}` — triage workflows for operators;
agents file friction via `add`.
- `orbit.learning.comment.list` — agents discover learnings via
`orbit.search`; comment listing is review-time / operator work.
- `orbit.learning.upvote` — upvote telemetry is an operator concern.
Also drops the duplicate `orbit.review-thread.*` alias added in
ORB-00273. The canonical `orbit.task.review_thread.*` form is what the
README, the `orbit-review-task` skill, the MCP allowlist, audit /
redaction policy, and `OrbitBuiltinAction` mapping all reference; the
alias had no callers outside its own tests.
Test guards updated:
- `EXPECTED_INACTIVE_TOOL_NAMES` gains the six trimmed names; the
count assertion moves 21 → 27.
- `friction_list_is_exposed_to_mcp_dispatch` flips to assert
`NotFound` (renamed to `friction_list_is_not_exposed_to_mcp_dispatch`).
- `friction_surface_supports_artifact_triage` is rewritten to split the
active set (`add`, `tags`, `update`) from the registered-but-inactive
set (`list`, `show`, `resolve`, `stats`), still verifying all six are
reachable via `all_schemas()`.
Groundhog tools and `orbit.learning.list` stay as-is: Groundhog is
load-bearing for the engine's checkpoint-attempt loop; learning.list is
already `register_inactive`.
Apply the dashboard redesign mock to Scoreboard, Threads, and Knowledge using existing dashboard payloads and actions only. Validation: node --check for dashboard JS, cargo test -p orbit-dashboard, make ci-fast, and local browser smoke check.
…ge… [ORB-00347] Make default_crew and [duel].candidates dynamic from detected agents at orbit init Planned-By: claude
Make default_crew and [duel].candidates dynamic from detected agents at orbit init
Bump workspace crates, plugin manifest, and npm proxy to 0.8.0 and refresh Cargo.lock. Minor bump for the breaking agent MCP surface trim + dropped orbit.review-thread.* alias (ORB-00346, ORB-00348); adds the 0.8.0 CHANGELOG entry (1 breaking change, 3 highlights).
The v0.8.0 release failed to publish: orbit-search-companion could not link on the ubuntu-22.04 runners because its prebuilt ONNX Runtime (via ort / fastembed) references glibc 2.38+ symbols (__isoc23_*) absent on glibc 2.35. Split the Linux release build so the orbit CLI stays on ubuntu-22.04 (glibc 2.35 floor preserved) while the companion builds on ubuntu-24.04 (glibc 2.39). Fix-forward to 0.8.1; v0.8.0 is not retagged (npm is immutable).
ort-sys 2.0.0-rc.12 ships no ONNX Runtime prebuilt for x86_64-apple-darwin
('can't do xcframework linking'), so orbit-search-companion can't link on
Intel macOS. The Intel-mac release leg now builds the orbit CLI only; the
release publishes companions for macOS-arm64 + Linux x86_64/arm64. The CLI
still ships for Intel mac. With the 0.8.1 glibc fix, this completes recovery
of v0.8.0. Fix-forward to 0.8.2; prior tags are not moved.
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
3 participants
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.
Promotes the v0.8.x release line to the production branch. See CHANGELOG.md.
Net release: v0.8.2 (recovery of the v0.8.0 cut).
orbit.review-thread.*alias dropped ([ORB-00346], [ORB-00348]); highlights: dashboard/scoreboard redesign, dynamic crew/duel composition, companion-asset publishing. The GitHub Release failed to publish (Linux + Intel-mac companion build failures), so it was recovered via two fix-forward patches:v0.8.2 release CI is green (build + publish-release + Homebrew + macOS smoke);
make release-checkagrees across plugin.json / package.json / npm / GitHub Release at 0.8.2. Known limitation: the Linux semantic companion requires glibc ≥2.38 (follow-up filed).🤖 Generated with Claude Code