DESIGNOSFORGE v1.6.0

DESIGNOSFORGE v1.6.0 Release Notes

DESIGNOSFORGE v1.6.0 upgrades the system from prompt governance into training-aware design orchestration.

Highlights

• Adds AestheticMemoryIndex for corpus audit, memory-index generation, and case recommendation.
• Upgrades PromptPacket to v1.6 with project-context lock, case-memory selection, and failure-memory sections.
• Adds project-context coverage across all current caption batches.
• Separates commercial projects, academic discipline competitions, Culture China research, and public cultural communication.
• Generates lora_training_sandbox/aesthetic_corpus/aesthetic_memory_index.json for batch-level memory selection.
• Adds CLI commands:
  • python -m app.cli lora audit-corpus
  • python -m app.cli lora build-memory-index
  • python -m app.cli lora recommend --domain exhibition-board --context academic-discipline-competition

Migration Notes

Prompt builders and downstream workflows should expect PromptPacket v1.6 sections 01_TASK_BRIEF through 18_REVISION_PROTOCOL.

After adding new case images or captions, run:

PYTHONPATH=. python -m app.cli lora audit-corpus
PYTHONPATH=. python -m app.cli lora build-memory-index

Do not use public reference images for weight training until rights are cleared.

Validation

• python -m app.cli capabilities
• python -m app.cli lora audit-corpus
• python -m app.cli lora recommend --domain exhibition-board --context academic-discipline-competition
• python -m app.cli run "做一个品牌 VI 方案" --prompt-packet
• pytest -q
• tools/validate_source_skill.py

DESIGNOSFORGE v1.5.1

DESIGNOSFORGE v1.5.1 Release Notes

Highlights

DESIGNOSFORGE v1.5.1 reserves the LoRA aesthetic training space needed for future case-image and reference-image learning.

• Adds a design-domain taxonomy for UI, poster, exhibition board, VI/brand, environmental art, packaging, typography, infovis, web, and short-video AIGC.
• Adds style axes such as minimal premium, editorial grid, Swiss modern, commercial product, soft luxury, tech futurism, cultural contemporary, environmental competition, experimental typography, and technical infovis.
• Adds python -m app.cli lora init-aesthetic-space to generate corpus folders, domain manifests, and taxonomy snapshots.
• Keeps real images out of git by default while tracking manifests, captions, quality reviews, and .gitkeep placeholders.
• Adds documentation for rights status, rejected examples, comparison sets, and quality labels.

Why It Matters

Future LoRA tuning should not mix every image into one unstructured dataset. DESIGNOSFORGE now separates design domains and style axes so aesthetic preferences can be optimized intentionally.

The reserved corpus supports:

• real successful case images
• reference images
• rejected images for failure contrast
• before/after comparison sets
• captions and prompt annotations
• quality reviews for visual cleanliness, text accuracy, layout order, and reference fidelity

Validation

• python -m app.cli lora init-aesthetic-space
• python tools/validate_source_skill.py
• pytest -q
• text health audit with mojibake_count: 0

DESIGNOSFORGE v1.5.0

DESIGNOSFORGE v1.5.0 Release Notes

Highlights

DESIGNOSFORGE v1.5.0 is a quality jump focused on visual discipline and prompt reliability.

• Open source: released as an MIT-licensed Codex agent/skill system.
• Cleaner images: anti-fragmentation rules prevent scattered tiny elements, dirty texture noise, random icons, and overfilled backgrounds.
• Stronger prompts: PromptPacket v1.5 adds aesthetic thesis, composition hierarchy, grid/density, exact text rules, QA gates, and revision protocol.
• Better text: exact visible text, spelling locks, no pseudo-text, and no mojibake are now first-class gates.
• Better layout: grid, margin, reading path, density ceiling, and no-overlap concerns are explicit.
• Better release process: GitHub validation workflow, PR body, release notes, and source skill validator are included.

Developer Notes

• New module: app.core.aesthetic_quality
• New CLI: python -m app.cli quality audit "<text>"
• New CLI: python -m app.cli quality guardrails "<text>"
• New workflow: .github/workflows/validate.yml
• New validator: tools/validate_source_skill.py
• New install guide: docs/CODEX_INSTALL.md

Upgrade Guidance

Use release/1.5.0 as the PR branch. After GitHub remote binding, push the branch and tag:

git push -u origin release/1.5.0
git push origin v1.5.0

Open a draft PR with docs/PR_BODY_v1.5.0.md, then publish release notes from docs/RELEASE_NOTES_v1.5.0.md.