Added AI image generation best practices to CLAUDE.md global configuration
Source type:
obs· Harvested: 2026-05-02 · Original date: 2026-05-01T16:01:24.260Z Metadata:{"project":"apex-ai-institute","type":"change","obs_id":64726}
obs/64726 · change · 2026-05-01T16:01:24.260Z
Added AI image generation best practices to CLAUDE.md global configuration
Claude updated global configuration file (CLAUDE.md) to codify best practices learned from current session’s image generation workflow. New section documents that editorial-style images should be generated via direct OpenAI SDK calls to gpt-image-2 model, bypassing Codex CLI to eliminate unnecessary GPT-5.5 orchestrator overhead. Provides working code template showing OpenAI client initialization and images.generate() usage with recommended size parameter. Documents API key location and highlights quality advantages (watercolor effects, asymmetric layouts, scientific visual motifs) over traditional matplotlib approaches. Cross-references AAI briefing hero image series as proof of concept. This represents proactive self-improvement: capturing session learnings as reusable patterns in global configuration to optimize future workflows.
Concepts: [“what-changed”,“pattern”,“trade-off”]
Facts: [“New section added to CLAUDE.md: “AI 圖像生成 → 直接呼叫 gpt-image-2"",“Recommends direct OpenAI Python SDK usage for editorial images (hero images, illustrations) instead of Codex CLI routing”,“Rationale: eliminates GPT-5.5 orchestrator inference cost when using Codex as intermediary”,“Code template provided: from openai import OpenAI; client.images.generate(model=“gpt-image-2”, prompt, size=“1536x1024”)”,“API key configuration documented: ~/.claude/.env with OPENAI_API_KEY environment variable”,“Quality benefits noted: watercolor rendering, asymmetric composition, scientific motifs superior to matplotlib”,“Reference added: AAI briefing hero series generated using this method”]
[← 回 Alfred Brain Hub]