feat(gcs): update image URLs to use staging and remove history references

- Changed image URLs to point to staging instead of production
- Removed history folder references from the promotion process
- Updated documentation to reflect the latest storage practices

Author: MarkusNeusinger

.github/workflows/impl-generate.yml

@@ -661,15 +661,14 @@ jobs:
           ISSUE: ${{ steps.issue.outputs.number }}
           PR_NUMBER: ${{ steps.pr.outputs.pr_number }}
         run: |
-          # Use versioned production URLs (permanent, in history/ folder)
-          # These will be available after impl-merge promotes from staging
-          PNG_URL="https://storage.googleapis.com/pyplots-images/plots/${SPEC_ID}/${LIBRARY}/history/v0.png"
+          # Use staging URLs (available immediately after upload)
+          PNG_URL="https://storage.googleapis.com/pyplots-images/staging/${SPEC_ID}/${LIBRARY}/plot.png"
           HTML_URL=""
 
           # Set HTML_URL for interactive libraries
           case "$LIBRARY" in
             plotly|bokeh|altair|highcharts|pygal|letsplot)
-              HTML_URL="https://storage.googleapis.com/pyplots-images/plots/${SPEC_ID}/${LIBRARY}/history/v0.html"
+              HTML_URL="https://storage.googleapis.com/pyplots-images/staging/${SPEC_ID}/${LIBRARY}/plot.html"
               ;;
           esac
 

.github/workflows/impl-merge.yml

@@ -165,27 +165,18 @@ jobs:
 
           STAGING="gs://pyplots-images/staging/${SPEC_ID}/${LIBRARY}"
           PRODUCTION="gs://pyplots-images/plots/${SPEC_ID}/${LIBRARY}"
-          HISTORY="${PRODUCTION}/history"
 
-          # Get created date from metadata for version naming (YYYY-MM-DD)
-          CREATED_DATE=$(yq '.created' "plots/${SPEC_ID}/metadata/${LIBRARY}.yaml" 2>/dev/null | cut -d'T' -f1 || echo "$(date +%Y-%m-%d)")
-          echo "Created date: ${CREATED_DATE}"
-
-          # 1. Copy from staging to production (current)
+          # Copy from staging to production
           gsutil -m cp -r "${STAGING}/*" "${PRODUCTION}/" 2>/dev/null || echo "No staging files to promote"
 
-          # 2. Copy versioned to history/ folder (using date instead of version number)
-          gsutil cp "${STAGING}/plot.png" "${HISTORY}/${CREATED_DATE}.png" 2>/dev/null || true
-          gsutil cp "${STAGING}/plot.html" "${HISTORY}/${CREATED_DATE}.html" 2>/dev/null || true
-
           # Make production files public
           gsutil -m acl ch -r -u AllUsers:R "${PRODUCTION}/" 2>/dev/null || true
 
           # Clean up staging
           gsutil -m rm -r "${STAGING}/" 2>/dev/null || true
 
           rm -f /tmp/gcs-key.json
-          echo "::notice::Promoted images to production + history/${CREATED_DATE}"
+          echo "::notice::Promoted images to production"
 
       - name: Update issue labels
         if: steps.check.outputs.should_run == 'true' && steps.issue.outputs.number != ''

CLAUDE.md

@@ -241,7 +241,7 @@ Example: `plots/scatter-basic/` contains everything for the basic scatter plot.
 - **`plots/{specification-id}/`**: Plot-centric directories (spec, metadata, implementations together)
   - `specification.md`: Library-agnostic specification (Description, Applications, Data, Notes)
   - `specification.yaml`: Spec-level metadata (tags, created, issue, suggested, updates)
-  - `metadata/{library}.yaml`: Per-library metadata (preview_url, quality_score, history)
+  - `metadata/{library}.yaml`: Per-library metadata (preview_url, quality_score, review feedback)
   - `implementations/{library}.py`: Library-specific implementations
 - **`prompts/`**: AI agent prompts for code generation, quality evaluation, and tagging
   - `templates/`: Spec and metadata templates
@@ -447,7 +447,7 @@ uv run python -c "from core.database import is_db_configured; print(is_db_config
 - Everything in `plots/{specification-id}/`:
   - `specification.md` - specification description
   - `specification.yaml` - spec-level metadata (tags, created, issue, etc.)
-  - `metadata/{library}.yaml` - per-library metadata (quality score, generation history)
+  - `metadata/{library}.yaml` - per-library metadata (quality score, review feedback)
   - `implementations/*.py` - library implementations
 
 **What's NOT Stored in DB**:

docs/architecture/database.md

@@ -410,9 +410,7 @@ GROUP BY s.id;
 ```
 1. New version of implementation
    → UPDATE impls SET preview_url=..., quality_score=..., updated_at=NOW()
-
-2. Old GCS images moved to history/
-   → Database references latest URLs
+   → GCS images overwritten with latest version
 ```
 
 ### Deletion Flow

docs/architecture/repository.md

@@ -258,8 +258,7 @@ gs://pyplots-images/
 ├── plots/{spec-id}/{library}/           # Production (after merge)
 │   ├── plot.png                         # Full-size optimized image
 │   ├── plot_thumb.png                   # Thumbnail (600px width)
-│   ├── plot.html                        # Optional (interactive libraries)
-│   └── history/                         # Version history
+│   └── plot.html                        # Optional (interactive libraries)
 │
 └── staging/{spec-id}/{library}/         # Temp (during review)
     └── plot.png, plot.html

docs/workflow.md

@@ -14,7 +14,7 @@ pyplots is a **community-driven, AI-powered platform** that automatically discov
 
 ### Key Principles
 
-1. **Images in GCS, Code in GitHub**: Plot PNGs stored in Google Cloud Storage with version history, source code version-controlled
+1. **Images in GCS, Code in GitHub**: Plot PNGs stored in Google Cloud Storage, source code version-controlled
 2. **Multi-Version Support**: All plots tested across Python 3.11+ (3.11, 3.12, 3.13, 3.13 primary)
 3. **Hybrid Automation**: AI handles routine tasks, humans approve critical decisions
 4. **Standard Datasets**: Use well-known datasets (pandas iris, seaborn tips, kaggle) for realistic previews
@@ -369,11 +369,10 @@ Via **GitHub Issue Labels**:
    - Avoid redundant evaluations
 
 2. **Efficient Storage**:
-   - GCS versioning: all versions kept permanently for history tracking
-   - Path structure: `plots/{spec-id}/{library}/{variant}/v{ISO-timestamp}.png`
-   - Thumbnails: `v{timestamp}_thumb.png` (400px width) for gallery views
+   - Path structure: `plots/{spec-id}/{library}/plot.png`
+   - Thumbnails: `plot_thumb.png` (600px width) for gallery views
    - Images never in git repository
-   - Before/after comparison in Issues for updates
+   - Only latest version stored (no version history)
 
 3. **Smart Scheduling**:
    - Event-based maintenance (not daily scheduled)