v1.7.0

D4Vinci/Scraplingv1.7.0Sep 7, 2026by umputun

Upgrading?

AI Summary

This release updates the default Codex model to `gpt-5.6-sol` and improves Windows compatibility by using forward slashes for paths. It introduces a new feature to archive completed progress logs and resolves multiple issues regarding plan archiving, OAuth persistence, and API limit detection.

Key Highlights

  • Archive completed progress logs instead of truncating them
  • Update default codex model to gpt-5.6-sol
  • Return repo-relative paths with forward slashes for Windows compatibility
  • Detect API Error codes and Claude spend limits
  • Persist OAuth token refreshes from container to host

New Features

  • Archive completed progress logs instead of truncating them

Full Release Notes

**New Features**

- Archive completed progress logs instead of truncating them #449 @umputun

**Improved**

- Update default codex model to gpt-5.6-sol and effort to high #421 @umputun
  Explicit `codex_model` / `codex_reasoning_effort` values in your own config are preserved on upgrade - edit them or comment them out to pick up the new defaults.
- Stop pointing users at gpt-5.5 for codex 76910af
- Describe Windows support as best effort 09fe05e
- Refresh dependencies and fix the e2e playwright install #434 @paskal
- Bump github.com/go-pkgz/notify from 1.4.0 to 1.5.0 #456 @dependabot
- Bump github.com/stretchr/testify from 1.11.1 to 1.12.0 #432 @dependabot
- Bump actions/setup-go from 6 to 7 #416 @dependabot

**Fixed**

- Match claude limit patterns by diagnostic provenance #455 @yeroo
- Return repo-relative paths with forward slashes so the plan file excludes itself on Windows #454 @dimi4ik
- Archive the completed plan on the feature branch #451 @umputun
- Detect codex "Selected model is at capacity" as a limit pattern #448 @umputun
- Let worktree mode start with a dirty source checkout #445 @umputun
- Serve the worktree's plan copy to the dashboard #442 @umputun
- Surface a failed plan archive instead of reporting a clean success #441 @umputun
- Restrict the completed-plan archive commit to the plan paths #437 @umputun
- Stop review agents from spawning subagents of their own #428 @umputun
- Detect claude individual spend limit in default patterns #426 @umputun
- Tighten signal rules in review prompts #424 @umputun
- Enumerate API Error codes in default claude_error_patterns #423 @umputun
- Pass max thinking level through to pi #433 @asalimonov
- Persist OAuth token refreshes from container to host #412 @umputun
- Harden OAuth credential bind mounts and keep older wrappers working 51f0d25
- Keep authenticated status in claude diagnostics and add weekly limit defaults 84100f4

Related Repositories