Version updated for https://github.com/remyxai/outrider to version v1.7.40. This action is used across all versions by 2 repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. Action Summary Outrider is a GitHub Action that automates the process of evaluating new methods against an existing codebase. It helps teams validate and compare new implementations by scheduling or dispatching evaluations on demand, using Anthropic Opus or z.ai GLM-5.2 as model backends for cost-effective evaluation. The action provides draft PRs with self-reviews, issues when preflight, validators, or self-review routes the paper to discussion, a branch-only mode, and ensures no duplicate work by not re-recommending papers once referenced.

Version updated for https://github.com/remyxai/outrider to version v1.7.34. This action is used across all versions by 2 repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. Action Summary This GitHub Action automates the process of validating and comparing new machine learning methods against existing codebases. It schedules runs weekly or allows for ad-hoc dispatches based on user inputs, providing self-review notes, issues, and a selection narrative in the step summary. The action uses pluggable model backends like Anthropic Opus and z.ai GLM-5.2 to evaluate candidates, with options for branch-only mode and cost considerations.

Version updated for https://github.com/remyxai/outrider to version v1.7.33. This action is used across all versions by 2 repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. Action Summary This GitHub Action, Outrider, automates the process of validating and comparing new methods against an organization’s existing codebase. It uses Anthropic Opus or z.ai GLM-5.2 as model backends to evaluate arXiv methods (or design-doc leads) against real call sites, providing a self-review and issue routing feature. It supports branch-only mode and avoids duplicate work by not re-recommending papers once they have been referenced. The action is designed to streamline the integration of new models into production environments while ensuring compliance with code review metrics.

Version updated for https://github.com/remyxai/outrider to version v1.7.28. This action is used across all versions by 2 repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. Action Summary Outrider is a GitHub Action that automates the validation and comparison of new methods against an organization’s codebase. It provides draft PRs with a self-review, handles preflight checks and issues, supports branch-only mode for exploring multiple candidates without committing to any one, and allows quickstart installation via CLI tools. The action uses Anthropic Opus or z.ai GLM-5.2 as model backends, with configurable costs based on the backend used.

Version updated for https://github.com/remyxai/outrider to version v1.7.24. This action is used across all versions by 2 repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. Action Summary This GitHub Action automates the validation and comparison of new methods against an organization’s codebase by integrating them into real call sites, providing a self-review process and handling issues when preflight, validators, or self-review routes the paper to discussion. It supports various model backends like Anthropic Opus and z.ai GLM-5.2 for different scenarios, including exploration and branch-mode exploration. The action is designed to streamline the workflow by running coding agents in ephemeral runners and handling multiple candidates per week efficiently.

Version updated for https://github.com/remyxai/outrider to version v1.7.23. This action is used across all versions by 2 repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. Action Summary This GitHub Action automates the validation and comparison of new research methods against an existing codebase. It helps teams measure changes by scheduling runs or triggering them on demand, using Anthropic Opus or z.ai’s GLM models as backends. The action supports draft PRs with self-reviews, issues for preflight checks, and branch-only mode for exploration without committing to a single candidate.

Version updated for https://github.com/remyxai/outrider to version v1.7.20. This action is used across all versions by 2 repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. What’s Changed Fix Pre-PR fidelity gate’s remediation pass now inherits the coding session’s self_review mode context. Under v1.7.19 and earlier, the remediation pass (after _attempt_pre_pr_fidelity_patch applied edits) called _run_pre_pr_fidelity_check without self_review, silently defaulting to Mode 1 with no substitutions or scope-outs. Any Mode-2 refinement that triggered the fidelity gate on the first pass then had its documented substitutions wrongly re-flagged as fabrication on the second pass — hard-to-shift outcome for legitimate Mode-2 diffs.

Version updated for https://github.com/remyxai/outrider to version v1.7.16. This action is used across all versions by 2 repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. What’s Changed What’s new start-from-ref — new workflow input that names a branch/tag/SHA on the target fork. When set, the coding session begins with that ref’s diff already applied to the workspace, so lead-content phrasing like “add tests for X” or “refactor Y” acts on real state instead of being advisory.

Version updated for https://github.com/remyxai/outrider to version v1.7.15. This action is used across all versions by 2 repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. What’s Changed Closes REMYX-195 with a mode-aware pre-PR fidelity gate and lands the lead-content input for spec-driven dispatches beyond arXiv paper implementations. Mode-aware fidelity gate The pre-PR fidelity gate now routes by the coding session’s cited implementation mode instead of applying one strict method-vs-diff comparison to every output:

Version updated for https://github.com/remyxai/outrider to version v1.7.14. This action is used across all versions by 2 repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. What’s Changed Rewrites the preflight and INVOCATION.md prompts to enumerate three legitimate implementation modes instead of a binary port-or-Issue choice. Three modes Direct port — implement the paper’s method as-described. Requires the repo to host the paper’s full infrastructure. Adapted port — implement the paper’s core mechanism at full fidelity while substituting auxiliary components (learned estimators, bespoke optimizers, benchmark suites) with target-native equivalents (parameter-free proxies, existing library functions, scope cuts). Inspired experiment — take the paper’s core insight or framing and implement a target-native experiment drawing on it. The PR applies the paper’s idea rather than reproducing its method. Route to Issue only when all three modes fail. The coding session’s self-review must cite which mode was used and, for Modes 2/3, name the specific substitutions or reframed insight.

Version updated for https://github.com/remyxai/outrider to version v1.7.13. This action is used across all versions by 2 repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. What’s Changed Follow-up to v1.7.12. The PDF-text discovery path shipped in v1.7.12 depended on pdftotext, which turned out not to be installed on ubuntu-latest runners as of mid-2026 — the fallback fired but silently returned empty. This release makes it actually work.

Version updated for https://github.com/remyxai/outrider to version v1.7.10. This action is used across all versions by 2 repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. What’s Changed Reorders the two attribution gates in _retry_license_via_arxiv_html so the stronger evidence (README references the paper’s arxiv id or ≥2 title words) runs first, with title-overlap demoted to a tie-breaker among survivors. Fixes silent license_class=no-code-link classifications on acronym-named repos whose acronym isn’t literally in the paper title.

Version updated for https://github.com/remyxai/outrider to version v1.7.9. This action is used across all versions by 2 repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. What’s Changed v1.7.8 was tagged on the wrong commit and did not actually contain the enrichment fix; v1 pointer was updated but the deployed code path in fast-paths still skipped _enrich_candidate_licenses. This release ships the real fix: _enrich_candidate_licenses(candidates, target) is now called on the pin-arxiv and search-method fast-paths (gated on REMYX_LICENSE_GATE, idempotent, best-effort). REMYX-190 evidence: https://github.com/remyxai/VQASynth/issues/105 opened with license_class=unknown despite WnQinm/Annotator having a clearly readable BSD-3-Clause LICENSE — the fast-path never called the enrichment step.

Version updated for https://github.com/remyxai/outrider to version v1.7.6. This action is used across all versions by 2 repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. What’s Changed The v1.7.5 cocoindex install step ran ccc --version as a sanity check, but the typer app doesn’t expose a –version flag so it raised a red “No such option ‘–version’” error box in the Actions UI. The || true suppression meant the step still succeeded, but the visible error box was misleading.

Version updated for https://github.com/remyxai/outrider to version v1.7.4. This action is used across all versions by 2 repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. What’s Changed HF Hub checkpoint preflight for architecture-add Issues (REMYX-178) (8a70186) arxiv HTML retry: title-overlap threshold + project-page one-hop + README verification (c6c021f) README: add lerobot #9 (ECoT reasoning supervision) to Examples (85fa29b) Coding-agent prompt: guidance on auto-format scope (be8720d) README: drop the “Recommended” ENVIRONMENTS.md section (82b9549) Recommend ENVIRONMENTS.md + cocoindex as the default setup (9ca24e1) License detection: retry via arxiv HTML on unfavorable buckets (615af84) README: restore smellslikeml/peft #5 as the primary Outrider artifact (e8fb3a0) README: swap smellslikeml/peft #5 for the upstream draft huggingface/peft #3382 (28cfc14) README: add huggingface/peft #3382 (upstream draft) to Examples (bc70d94)

Version updated for https://github.com/remyxai/outrider to version v1.7.0. This action is used across all versions by 2 repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. What’s Changed Behavior changes that replace categorical shape-label guards with measurement-based decisions, and give the selection stage access to workflow-attached tooling (ENVIRONMENTS.md). Behavior changes Substitution guard removed. shape in ("replacement", "simplification") → auto-Issue short-circuit no longer fires before implementation. Replacement/simplification runs proceed; the path allowlist + check_integration() catch broken diffs on measured evidence. ENVIRONMENTS.md at selection. Loader runs early and threads the body into select_recommendation’s prompt, so the selection agent has workflow-attached tooling (AST-search skills, MCP servers) while verifying candidates. Empty ENVIRONMENTS.md = unchanged behavior. Self-review orphan surfaced, not vetoed. When is_orphan=true, the PR ships with a prominent warning in the body instead of being downgraded to Issue. Upstream measurement-based gates already catch scaffold-shaped diffs. Confabulation check. Extracts path-like tokens from selection_reasoning and verifies each against the workdir. Step-summary shows N of M verified; a 0 of N verified line surfaces confidently-wrong reasoning. Compatibility Backwards-compatible for runs without an ENVIRONMENTS.md file (loader no-ops). No config changes needed.

Version updated for https://github.com/remyxai/outrider to version v1.6.34. This action is used across all versions by 2 repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. What’s Changed When the agentic selection call fails (429, timeout, unparseable output), Outrider’s fallback picks the highest-relevance candidate. Ties on relevance were previously broken by list position — Python’s max() returns the first element at the max value.

Version updated for https://github.com/remyxai/outrider to version v1.6.31. This action is used across all versions by 2 repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. What’s Changed The self-review call now honors the claude-timeout workflow input — completing the per-stage timeout consolidation begun in v1.6.28 (preflight), v1.6.29 (audit), v1.6.30 (selection). After this release, claude-timeout is the single budget knob across every Claude-Code stage in the chain.