Skip to content

v1.2.4 — Opus 4.7 effort hints + docs cleanup

Choose a tag to compare

@mosandlt mosandlt released this 18 Apr 17:30
· 23 commits to main since this release

🇬🇧 English below · 🇩🇪 Deutsch darunter


🇬🇧 English

Maintenance + one small feature — targeted at Claude Opus 4.7 users.

Opus 4.7 per-mode effort hints

When the active model is claude-opus-4-7, every outbound request now attaches an output_config.effort hint derived from the mode you're using. The hint tells the model how much reasoning to invest, so simple tasks stay snappy and complex ones get the full treatment.

Mode Effort
Plus (light cleanup) low
Emoji (decoration) low
Rage (tone shift) medium
Business (structure + style) high
Office (document analysis) high
Prompt (intent → prompt) xhigh

Threading goes through LLMRouter → AnthropicClient. Sonnet, Haiku, OpenAI, Ollama, and every custom OpenAI-compatible profile are unaffected. No UI change — the hint is derived automatically from the mode you picked.

Docs cleanup (maintenance)

  • CLAUDE.md re-synced with the actual code state: current version, 7-mode list, full file overview, architecture diagram shows the Privacy-wrap + inline-recovery branch, Settings-tab count corrected (7, not 6), obsolete LSUIElement=YES claim removed (activation policy is set programmatically now, since v1.2.1).
  • Release history in CLAUDE.md brought up to date with v1.2.0–v1.2.3.
  • Stray .build/ cache + old v1.0.0 zip removed from the Nextcloud-synced project folder.

Install

  1. Download blitzbot-1.2.4-macos-arm64.zip.
  2. Unzip, drag blitzbot.app to /Applications.
  3. First launch: right-click → Open (ad-hoc signed, Gatekeeper requires the manual confirmation once).
  4. If macOS 15+ flags the app as damaged: xattr -dr com.apple.quarantine /Applications/blitzbot.app.

🇩🇪 Deutsch

Wartung + ein kleines Feature — gezielt für Claude-Opus-4.7-Nutzer.

Opus 4.7 per-Modus Effort-Hints

Wenn das aktive Modell claude-opus-4-7 ist, setzt jeder Request jetzt automatisch einen output_config.effort-Hint, abgeleitet aus dem Modus. Der Hint sagt dem Modell, wieviel Reasoning es investieren soll — einfache Tasks bleiben schnell, komplexe bekommen die volle Tiefe.

Modus Effort
Plus (leichtes Glätten) low
Emoji (Dekoration) low
Rage (Tonwechsel) medium
Business (Struktur + Stil) high
Office (Dokument-Analyse) high
Prompt (Idee → Prompt) xhigh

Threading läuft durch LLMRouter → AnthropicClient. Sonnet, Haiku, OpenAI, Ollama und jedes custom OpenAI-kompatible Profil bleiben unverändert. Keine UI-Änderung — der Hint wird automatisch aus dem Modus abgeleitet.

Doku-Cleanup (Wartung)

  • CLAUDE.md mit dem echten Code-Stand synchronisiert: aktuelle Version, 7-Modi-Liste, komplette Datei-Übersicht, Architektur-Diagramm zeigt Privacy-Wrap + Inline-Recovery-Branch, Settings-Tab-Zahl korrigiert (7, nicht 6), veraltete LSUIElement=YES-Aussage entfernt (Activation-Policy wird seit v1.2.1 programmatisch gesetzt).
  • Release-Historie in CLAUDE.md auf v1.2.0–v1.2.3 nachgezogen.
  • Stray .build/-Cache + alte v1.0.0-Zip aus dem Nextcloud-Projektordner entfernt.

Installation

  1. blitzbot-1.2.4-macos-arm64.zip herunterladen.
  2. Entpacken, blitzbot.app nach /Programme ziehen.
  3. Erster Start: Rechtsklick → Öffnen (ad-hoc signiert, Gatekeeper braucht diese einmalige Bestätigung).
  4. Falls macOS 15+ meldet „beschädigt": xattr -dr com.apple.quarantine /Applications/blitzbot.app.