v0.8.0

Provider-neutral backlogs. Backlog items no longer bind to Claude by default, and
a new loop flag lets a Claude-aliased backlog run portably under any agent without
editing it — closing the #38 failure mode where a model: "opus" item silently
halted the loop under a non-Claude agent. Additive minor bump.

Added

• rauf loop run --no-model (alias --model none) — ephemeral per-run model
  override that makes the loop ignore each backlog item's model field for that
  run (the new ignoreItemModel loop option). Resolution drops to
  --model > project default > provider default, so a backlog whose items carry
  Claude-only tier aliases (opus/sonnet/…) runs portably under a non-Claude
  --agent without a persistent edit to backlog.json. Also accepted on the
  POST /loop/start body for server-mode parity. (#38)

Changed

• author-backlog skill is provider-neutral by default — item model is now
  omitted unless the user explicitly opts into a Claude tier, keeping authored
  backlogs agent-portable. Tier aliases are documented as Claude-only and
  agent-binding. (#38)
• review-backlog skill flags items carrying Claude-only model aliases as a
  portability concern (new "Claude-bound model alias" anti-pattern). (#38)

Full Changelog: v0.7.0...v0.8.0

v0.7.0

The agent-agnostic epic — rauf's loop runner is no longer Claude-only. A pluggable
provider layer (packages/loop/src/providers/) lets the loop drive any CLI coding
agent via presets + a generic adapter, with agent selection, availability
pre-checks, and a hardened process-group lifecycle. Additive minor bump.

Added

• LLM-agnostic provider architecture in packages/loop — providers/
  (registry, presets, generic-CLI + CLI-agent adapters, shared types), an
  agent-selection resolver, and a process-group lifecycle for clean
  child-process teardown. The runner resolves and launches the configured agent
  by precedence and classifies its outcome provider-agnostically.
• .gitattributes — LF normalization (* text=auto eol=lf) + export-ignore
  for dev-only trees (specs/, tests/, .github/, test-sandbox/).
• npm-publishability prep on the packages the installer's rauf@0.6.0 pin
  targets (publishConfig / files / bin) — machinery only; no publish is
  executed (the npx rauf@0.6.0 path is documented as "available once rauf 0.6.0
  is published").
• Optional npm-publish.yml — workflow_dispatch-only publish machinery,
  outside the PR gate (not run by this feature).

Changed

• README — added a labeled cross-agent section linking feature-forge's
  cross-agent install story (loop-runner framing retained).

Full Changelog: v0.6.0...v0.7.0

v0.6.0

Phase 4 of the rauf UX/DX overhaul — web/CLI recovery parity, a shared status
vocabulary, and a ratified agent contract. Additive minor bump (no
minRunnerVersion change, no feature-forge lockstep).

Added

• Web recovery parity with the CLI: reset, resume, review, unblock,
  and validate are now exposed as web server routes with matching status-page
  controls, so the dashboard can drive the same loop-recovery operations the CLI
  offers.
• Shared status label-map across CLI and web — REVIEWING and
  PAUSED_USAGE_LIMIT badges and a "Needs Human" label render identically in
  both surfaces.
• rauf update --check — report-only drift audit that prints whether a
  project's artifacts are stale (tool-version lag or dead hash keys) and exits
  non-zero if so, writing nothing. Makes fleet-wide staleness scriptable.

Changed

• status exit codes are aligned with the unified scheme via a shared
  statusExitCode mapping, so the web DerivedStatus and CLI agree on outcome
  semantics.
• Agent-contract documentation finalized and the UX-overhaul canon ratified
  (canon-conformance review: GO, 0 blockers).
• rauf update now prunes stale artifact-hash keys from the marker (e.g. the
  legacy ralph.sh/ralph-status.sh/ralph-add.sh hashes carried over from a
  pre-rename install) instead of preserving them indefinitely.
• rauf migrate documentation sharpened as a legacy one-shot (it renames
  structure but does not backfill artifacts — follow with rauf update; non-rauf
  config references to .ralph are reported but not auto-rewritten).

Removed

• rauf update --yes retired from --help — update is non-destructive and
  never prompts (the flag is still tolerated for back-compat).

Full Changelog: v0.5.0...v0.6.0

v0.4.0

Added

• rauf loop run --pause-on-needs-human — opt-in run mode that halts the loop
  (state paused_human, with a distinct non-zero exit code) on the first
  RAUF_NEEDS_HUMAN instead of setting the item aside and continuing, so a
  supervising session can detect the pause. Emits a loop_paused NDJSON event.
• rauf resume --answer <id> "<text>" (repeatable) — inject a human's answer
  into a paused needs-human item and re-queue it; the answer is threaded into the
  item's next prompt and cleared once it completes.
• Machine-observation surfaces (loop run --ndjson event vocabulary and
  status --json DerivedStatus) are now documented as a versioned contract
  in docs/SPEC-BACKLOG-TOOL-CONTRACT.md, with the machine-vs-human surface
  distinction made explicit.
• Web dashboard: a specific empty/error state when the configured root directory
  does not exist (with a Settings link) plus pre-save root validation; and a
  favicon (served in dev and from the compiled binary).

Changed

• Backlog-authoring guidance uses model tier aliases (opus/sonnet) instead
  of pinned IDs, and documents opus[1m] for items that need the 1M context
  window (opt-in via the [1m] suffix; no cost premium on Opus).
• The web server's startup recovery resolves its root via the standard precedence
  (RAUF_ROOT env → config → cwd), honoring an explicit RAUF_ROOT/--root
  override.
• --create-branch, --pause-on-needs-human, and resume --answer are now listed
  in the CLI --help flag tables.
• Purged user-facing ralph leftovers from the web UI (theme localStorage key,
  migrated transparently; command examples).

Fixed

• Loop wedge: item completion is now authoritative — if an item's on-disk
  status is perturbed (e.g. reverted to pending) mid-iteration, the runner
  re-asserts in_progress before marking done and surfaces failures, instead of
  silently failing the invalid pending -> done transition and re-running the
  item indefinitely.
• Server startup recovery (recoverStaleLoops) no longer resets in_progress
  items in projects whose lock is held by a live loop (e.g. a direct-mode
  rauf loop run); only genuinely stale loops are recovered.
• LOG_PATTERNS.needsHuman now matches the runner's actual
  Item <id> needs human input (set aside): <reason> line.
• RAUF_* terminal tokens in the diagnostic "Signal text" log dump are redacted so
  agent prose can no longer plant false signals in a grepped rauf.log.
• rauf resume --answer 001 "..." no longer misreads the answer text as the
  project path in the documented no-path form.
• README: broken images and the loop diagram restored/renamed; the version badge is
  now a dynamic GitHub-release badge; docs builds no longer dirty the working tree.

Full Changelog: v0.3.0...v0.4.0

v0.3.0

First stable release under the rauf name. Promotes 0.3.0-rc.2; the
0.3.0-rc.1 and 0.3.0-rc.2 sections below carry the full per-candidate detail.

Changed (BREAKING) — Ralph is now Rauf

• The tool was renamed from ralph to rauf: binary, the @rauf/* package
  scope, .rauf/ state dir, .rauf.json, RAUF.md, RAUF_ROOT,
  X-Rauf-Request, ~/.rauf/, and the RAUF_* loop signals. See
  MIGRATION.md.

Added

• rauf migrate <path> — in-place migration of a legacy ralph project to
  rauf, with --dry-run, --no-backup, --clean-backups, and --global.

Fixed

• Release binaries for x64 are built with Bun's -baseline runtime so they run
  on every x64 CPU; the previous builds required AVX2 and crashed with SIGILL
  on CPUs without it.

Full Changelog: v0.3.0-rc.2...v0.3.0