feat(altair): implement line-loss-training

[github-actions]

Implementation: line-loss-training - python/altair

Implements the python/altair version of line-loss-training.

File: plots/line-loss-training/implementations/python/altair.py

Parent Issue: #2860

---

🤖 impl-generate workflow

[claude]

AI Review - Attempt 1/3

Image Description

Light render (plot-light.png): The plot displays training (blue line) and validation (yellow line) loss curves on a warm off-white background (#FAF8F1). The title "line-loss-training · altair · anyplot.ai" is centered at the top in dark text. Axis labels ("Epoch" on X-axis, "Cross-Entropy Loss" on Y-axis) are clearly readable in dark text. Grid lines are subtle and appropriate. A red diamond marker annotates the optimal stopping point (minimum validation loss at epoch ~26) with accompanying red text label. Both curves are clearly visible with distinct colors, and all text elements are legible against the light background.

Dark render (plot-dark.png): The same plot structure appears on a warm near-black background (#1A1A17). Title and axis labels are now in light text, clearly readable against the dark surface. Grid lines remain subtle. The data curves maintain identical colors to the light render (blue training, yellow validation), and the annotation diamond and text remain red. All text is readable without dark-on-dark failures. The color consistency between renders is correct (only chrome should flip, not data colors).

Score: 81/100

Category	Score	Max
Visual Quality	27	30
Design Excellence	10	20
Spec Compliance	15	15
Data Quality	15	15
Code Quality	9	10
Library Mastery	5	10
Total	81	100

Visual Quality (27/30)

• VQ-01: Text Legibility (8/8) - All titles, labels, and annotations readable in both themes
• VQ-02: No Overlap (6/6) - No overlapping elements; legend and annotation positioned clearly
• VQ-03: Element Visibility (6/6) - Lines, points, and annotation marker all clearly distinguishable
• VQ-04: Color Accessibility (1/2) - Uses non-Okabe-Ito colors; palette not CVD-optimized per spec
• VQ-05: Layout & Canvas (4/4) - Proportions well-balanced, elements within canvas bounds
• VQ-06: Axis Labels & Title (2/2) - Descriptive labels with units (Cross-Entropy Loss)
• VQ-07: Palette Compliance (0/2) - CRITICAL FAILURE: First series uses #306998 instead of #009E73; second series uses #FFD43B instead of #D55E00; annotation color not theme-adaptive

Design Excellence (10/20)

• DE-01: Aesthetic Sophistication (4/8) - Generic styling with Python default colors, no custom palette design
• DE-02: Visual Refinement (2/6) - Minimal customization; basic grid and view configuration
• DE-03: Data Storytelling (4/6) - Clear visualization of overfitting with effective annotation at optimal stopping point

Spec Compliance (15/15)

• SC-01: Plot Type (5/5) - Correct line chart implementation
• SC-02: Required Features (4/4) - Both loss curves, legend, and optimal point annotation present
• SC-03: Data Mapping (3/3) - Epoch to X-axis, loss to Y-axis; all data visible
• SC-04: Title & Legend (3/3) - Correct format with proper legend labels

Data Quality (15/15)

• DQ-01: Feature Coverage (6/6) - Shows training/validation curves with overfitting pattern
• DQ-02: Realistic Context (5/5) - Realistic neural network training data, neutral domain
• DQ-03: Appropriate Scale (4/4) - Sensible loss values and epoch range

Code Quality (9/10)

• CQ-01: KISS Structure (3/3) - Simple, linear code without unnecessary functions
• CQ-02: Reproducibility (2/2) - Uses np.random.seed(42) for deterministic output
• CQ-03: Clean Imports (2/2) - Only necessary imports (altair, numpy, pandas)
• CQ-04: Code Elegance (2/2) - Appropriate complexity, no simulated UI features
• CQ-05: Output & API (0/1) - CRITICAL FAILURE: Saves as plot.png instead of plot-{THEME}.png; no theme-adaptive configuration

Library Mastery (5/10)

• LM-01: Idiomatic Usage (4/5) - Good use of Altair's layer composition and encoding
• LM-02: Distinctive Features (1/5) - Multiple layers used but in standard patterns

Score Caps Applied

• None

Strengths

• Realistic training data with proper overfitting behavior (validation diverges from training)
• Effective annotation highlighting optimal stopping point with visual marker
• Clear, readable rendering in both light and dark themes
• Good font sizing for visibility at large canvas resolution
• Proper legend structure and axis labeling

Weaknesses

• Palette violation: Uses Python colors (#306998, #FFD43B) instead of Okabe-Ito (#009E73 first series, #D55E00 second)
• Missing theme adaptation: Code does not read ANYPLOT_THEME environment variable or implement theme-adaptive chrome
• Wrong output files: Saves as plot.png/plot.html instead of plot-{THEME}.png/plot-{THEME}.html
• Non-adaptive annotation color: Hardcoded #E63946 red for annotation text doesn't follow theme tokens

Issues Found

1. VQ-07 CRITICAL: Color palette not Okabe-Ito compliant

   • Fix: Replace #306998 with #009E73 (training), #FFD43B with #D55E00 (validation)

2. CQ-05 CRITICAL: Incorrect output file handling and missing theme support

   • Fix: Read os.getenv("ANYPLOT_THEME", "light"); implement theme-adaptive colors for background, text, axes; save as plot-{THEME}.png and plot-{THEME}.html

3. Theme-adaptive chrome: Implement full theme token system (INK, INK_SOFT, ELEVATED_BG per altair.md)

   • Fix: Add theme-aware configuration for axis colors, title color, legend styling, annotation text color

AI Feedback for Next Attempt

Implement complete theme-adaptive styling following the altair.md specification. Replace hardcoded colors with Okabe-Ito palette (first series #009E73, second #D55E00). Read ANYPLOT_THEME environment variable and generate both plot-light and plot-dark outputs with proper theme tokens for all chrome elements (axis labels, title, grid, annotation text).

Verdict: REJECTED

[github-actions]

🔧 Repair Attempt 1/4

Applied fixes based on AI review feedback.

Status: Repair completed, re-triggering review...

---

🤖 impl-repair

[claude]

AI Review - Attempt 2/3

Image Description

Light render (plot-light.png): The plot displays training and validation loss curves declining over 50 epochs on a warm off-white background (#FAF8F1). The training loss (teal/blue) and validation loss (yellow) are both visible as thick lines with points marked. The title "line-loss-training · altair · anyplot.ai" is clearly readable in dark text. Axis labels ("Epoch", "Cross-Entropy Loss") and all tick labels are legible. Grid lines are subtle but present. A red diamond marks the minimum validation loss at epoch 26 with an annotation label. All text and data elements are readable against the light background. Legibility: PASS.

Dark render (plot-dark.png): Same plot on warm near-black background (#1A1A17). Title and axis labels are clearly visible in light text. Critical issue: The validation loss curve (yellow #FFD43B) has very poor contrast against the dark background and is barely visible — nearly indistinguishable from the background. The training loss curve (teal) is readable. The annotation text is visible in white/red. The data colors are identical to the light render (only background changed), but the yellow line's luminance is insufficient for dark-theme readability. Legibility: FAIL — the validation loss line violates readability requirements on dark surfaces.

Score: 65/100

Category	Score	Max
Visual Quality	21	30
Design Excellence	6	20
Spec Compliance	14	15
Data Quality	15	15
Code Quality	5	10
Library Mastery	4	10
Total	65	100

Visual Quality (21/30)

• VQ-01: Text Legibility (8/8) — All text readable in both themes
• VQ-02: No Overlap (6/6) — No text collisions
• VQ-03: Element Visibility (2/6) — Validation loss line barely visible on dark render
• VQ-04: Color Accessibility (1/2) — Position 7 yellow poor contrast on dark
• VQ-05: Layout & Canvas (4/4) — Good proportions, nothing cut off
• VQ-06: Axis Labels & Title (2/2) — Descriptive with units
• VQ-07: Palette Compliance (0/2) — Non-Okabe-Ito, violates first-series hard rule

Design Excellence (6/20)

• DE-01: Aesthetic Sophistication (2/8) — Generic blue/yellow colors, no design intent
• DE-02: Visual Refinement (2/6) — Default grid, no customization
• DE-03: Data Storytelling (2/6) — Annotation shows minimum, but weak visual hierarchy

Spec Compliance (14/15)

• SC-01: Plot Type (5/5) — Correct line chart
• SC-02: Required Features (4/4) — Both curves + optimal stopping point marker
• SC-03: Data Mapping (3/3) — Epoch vs. loss, axes correct
• SC-04: Title & Legend (2/3) — Inconsistency: code says "pyplots.ai" but render shows "anyplot.ai"

Data Quality (15/15)

• DQ-01: Feature Coverage (6/6) — Full training/validation lifecycle shown
• DQ-02: Realistic Context (5/5) — Realistic neural network training scenario
• DQ-03: Appropriate Scale (4/4) — Sensible loss values

Code Quality (5/10)

• CQ-01: KISS Structure (3/3) — No unnecessary functions
• CQ-02: Reproducibility (2/2) — np.random.seed(42)
• CQ-03: Clean Imports (2/2) — Only needed imports
• CQ-04: Code Elegance (2/2) — Clear logic
• CQ-05: Output & API (0/1) — FAIL: Saves to plot.png (hardcoded) instead of plot-{THEME}.png; does not read or respect ANYPLOT_THEME environment variable

Library Mastery (4/10)

• LM-01: Idiomatic Usage (3/5) — Basic Altair API, no advanced patterns
• LM-02: Distinctive Features (1/5) — Generic usage, no advanced Altair techniques

Score Caps Applied

• None (VQ-07 = 0 creates implicit ceiling via palette violation)

Strengths

• Clear annotation of optimal stopping point with visual marker
• Realistic training data showing overfitting behavior (diverging curves)
• Good use of line composition with both lines and points for visibility
• Data is correctly mapped and scaled

Weaknesses

• CRITICAL: Palette non-compliance — Uses custom colors (#306998, #FFD43B) instead of Okabe-Ito; violates hard rule: first series must be #009E73
• CRITICAL: Dark-theme readability failure — Yellow validation loss line nearly invisible on dark background (poor luminance contrast)
• API violation — Does not implement theme-adaptive output (should save plot-{THEME}.png; should read ANYPLOT_THEME)
• Missing theme colors — No use of theme tokens (PAGE_BG, INK, INK_SOFT) from style guide
• Title inconsistency — Code has "pyplots.ai" but should be "anyplot.ai"
• Generic styling with no visual customization beyond defaults

Issues Found

1. VQ-07 CRITICAL: Palette non-compliance

   • Problem: Custom colors (#306998 for training, #FFD43B for validation) instead of Okabe-Ito canonical order. Hard rule violated: first series must be #009E73 (brand green).
   • Fix: Change color scale to: Training = (Okabe-Ito pos 1), Validation = (Okabe-Ito pos 2)

2. VQ-03 CRITICAL: Dark-theme visibility failure

   • Problem: Yellow validation loss line (#FFD43B) has insufficient contrast on dark background (#1A1A17); barely visible
   • Fix: See issue Add workflow diagram for new prototype discovery #1 — Okabe-Ito pos 2 (vermillion #D55E00) has much better dark-theme contrast

3. CQ-05 CRITICAL: Output & API violation

   • Problem: Saves to plot.png (hardcoded). Code does not read ANYPLOT_THEME or save to plot-{THEME}.png
   • Fix: Implement theme-adaptive logic per prompts/library/altair.md; save to f'plot-{THEME}.png'

4. SC-04 MEDIUM: Title domain

   • Problem: Line 91 has "pyplots.ai"; should be "anyplot.ai"
   • Fix: Change to "anyplot.ai"

5. Design LOW: Missing theme-adaptive chrome

   • Problem: No use of configure_axis, configure_title, configure_legend for theme colors
   • Fix: Add PAGE_BG, INK, INK_SOFT tokens per altair.md template

AI Feedback for Next Attempt

To achieve APPROVED status:

1. Replace colors with Okabe-Ito palette (Training: #009E73, Validation: #D55E00). This fixes both palette compliance (VQ-07) and dark-theme visibility (VQ-03).
2. Implement theme-adaptive output using ANYPLOT_THEME environment variable; save to plot-{THEME}.png and plot-{THEME}.html.
3. Add theme-aware configure calls for background, axis colors, label colors, and legend styling per the altair.md style guide template.
4. Fix title domain to "anyplot.ai".
5. Consider removing top/right spines and refining grid opacity for visual polish.

Verdict: REJECTED

The implementation fails critical requirements: Okabe-Ito palette compliance (hard rule), dark-theme readability, and theme-adaptive API. The dark render's validation loss line is nearly invisible due to yellow-on-dark contrast failure. Requires substantial fixes to pass review.

[github-actions]

🔧 Repair Attempt 2/4

Applied fixes based on AI review feedback.

Status: Repair completed, re-triggering review...

---

🤖 impl-repair

[claude]

AI Review - Attempt 3/3

Image Description

Light render (plot-light.png): Warm off-white background (#FAF8F1 appearance). Title "line-loss-training · altair · anyplot.ai" is clearly visible at the top. X-axis labeled "Epoch" (1-50) with readable dark tick labels. Y-axis labeled "Cross-Entropy Loss" (0-2.8) with readable dark text. Two lines displayed: teal/green line showing training loss and orange line showing validation loss, both with visible point markers. Red diamond marker at epoch 25 indicating minimum validation loss with text annotation. Legend on right shows "Curve Type" with both series labeled. Grid lines subtle and visible. All text is readable with good contrast against light background - PASS.

Dark render (plot-dark.png): Dark near-black background (matching #1A1A17 specification). Title remains visible with light text. Axis labels and tick labels readable in light color against dark background. Training loss line appears as teal/green and validation loss as orange - colors are identical to light render, confirming proper data color consistency. Red diamond marker and annotation text visible. Legend readable. Grid lines visible at proper opacity. All text legible in dark theme - PASS.

Score: 80/100

Category	Score	Max
Visual Quality	27	30
Design Excellence	10	20
Spec Compliance	13	15
Data Quality	15	15
Code Quality	9	10
Library Mastery	6	10
Total	80	100

Visual Quality (27/30)

• VQ-01: Text Legibility (8/8) - Font sizes explicitly set; all text readable in both themes
• VQ-02: No Overlap (6/6) - No overlapping elements; clear spacing
• VQ-03: Element Visibility (6/6) - Lines and points clearly visible and well-sized
• VQ-04: Color Accessibility (1/2) - Good contrast between lines, but colors don't use accessible palette
• VQ-05: Layout & Canvas (4/4) - Good proportions, nothing cut off
• VQ-06: Axis Labels & Title (2/2) - Descriptive labels with units (Cross-Entropy Loss)
• VQ-07: Palette Compliance (0/2) - CRITICAL: Uses #306998 (Python blue) and #FFD43B (yellow) instead of Okabe-Ito. First series must be #009E73 (brand green), second must be #D55E00 (vermillion).

Design Excellence (10/20)

• DE-01: Aesthetic Sophistication (4/8) - Generic styling with default colors, no custom design thought
• DE-02: Visual Refinement (2/6) - Library defaults with minimal customization
• DE-03: Data Storytelling (4/6) - Clear comparison of training vs validation; annotation helps identify optimal point

Spec Compliance (13/15)

• SC-01: Plot Type (5/5) - Correct line chart type
• SC-02: Required Features (4/4) - Training loss, validation loss, legend, optimal point annotation all present
• SC-03: Data Mapping (3/3) - X-axis epochs, Y-axis loss values correctly mapped
• SC-04: Title & Legend (1/3) - ERROR: Title says "pyplots.ai" instead of "anyplot.ai". Legend labels correct.

Data Quality (15/15)

• DQ-01: Feature Coverage (6/6) - Shows training loss, validation loss, and overfitting behavior
• DQ-02: Realistic Context (5/5) - Realistic neural network training curves
• DQ-03: Appropriate Scale (4/4) - Sensible epoch (1-50) and loss ranges

Code Quality (9/10)

• CQ-01: KISS Structure (3/3) - Simple linear structure: imports → data → chart → save
• CQ-02: Reproducibility (2/2) - for deterministic output
• CQ-03: Clean Imports (2/2) - Only necessary imports (altair, numpy, pandas)
• CQ-04: Code Elegance (2/2) - Appropriate complexity, no fake functionality
• CQ-05: Output & API (0/1) - CRITICAL: Saves as plot.png instead of plot-{THEME}.png. Doesn't read ANYPLOT_THEME environment variable. Should implement theme-adaptive colors per altair.md.

Library Mastery (6/10)

• LM-01: Idiomatic Usage (3/5) - Uses altair properly (Chart, mark_line, mark_point, encode), but missing theme-adaptive configure methods
• LM-02: Distinctive Features (3/5) - Uses layer composition (lines + points + markers + text), but generic for altair's capabilities

Score Caps Applied

• None

Strengths

• Clear, readable visualization of training vs validation loss curves in both light and dark renders
• Excellent use of annotation (red diamond + text) to highlight the optimal stopping point
• Good font sizing and spacing - all text is legible at full resolution
• Proper line and point visualization showing data density clearly
• Well-structured data generation simulating realistic neural network behavior

Weaknesses

• Mandatory: Color palette violates style guide - Uses Python blue (#306998) and yellow (#FFD43B) instead of required Okabe-Ito palette. First categorical series MUST be #009E73 (brand green), second MUST be #D55E00 (vermillion).
• Mandatory: No theme-adaptive chrome - Doesn't read ANYPLOT_THEME or apply theme-aware colors. Should set background, text colors, and grid colors based on theme per altair.md guide.
• Mandatory: Wrong output file format - Saves as plot.png instead of plot-light.png and plot-dark.png. Should generate both variants with theme-specific colors.
• Error: Title typo - Says "pyplots.ai" instead of "anyplot.ai"
• Missing library mastery - Doesn't use altair's theme-aware configuration methods (configure_view, configure_axis with theme-adaptive colors)

Issues Found

1. VQ-07 PALETTE FAILURE (0/2): Palette must be Okabe-Ito. First series: #009E73 (brand green), second: #D55E00 (vermillion). Currently uses Python default colors.

   • Fix: Replace scale with:

2. CQ-05 THEME RENDERING FAILURE (0/1): Must read ANYPLOT_THEME and apply theme-adaptive colors. Must save as plot-{THEME}.png.

   • Fix: Add at start:

   • Apply to chart via configure_view, configure_axis, configure_title, configure_legend

   • Save as: and

3. SC-04 TITLE TYPO: Line 91 says "pyplots.ai" but must be "anyplot.ai"

   • Fix: Change to:

AI Feedback for Next Attempt

Fix three critical failures: (1) Use Okabe-Ito palette (#009E73, #D55E00), (2) Implement full theme-adaptive rendering (read ANYPLOT_THEME, set background/text/grid colors, save as plot-{THEME}.png), (3) Fix title typo (anyplot.ai). See altair.md library guide for theme-adaptive pattern. Current implementation is technically solid but violates mandatory style requirements.

Verdict: REJECTED

This implementation has a score of 80/100 which would pass Attempt 3's ≥70 threshold, however it violates explicit mandatory requirements from the style guide and altair library rules:

• Okabe-Ito palette is non-negotiable (VQ-07 = 0)
• Theme-adaptive rendering is required (CQ-05 = 0, altair.md explicit requirement)
• Proper file naming convention must be followed (CQ-05)

These are not style preferences but fundamental architecture requirements. The implementation must be repaired to address all three critical failures.

[claude]

AI Review - Attempt 3/3

Image Description

Light render (plot-light.png): Warm off-white background (#FAF8F1 appearance). Title "line-loss-training · altair · anyplot.ai" is clearly visible at the top. X-axis labeled "Epoch" (1-50) with readable dark tick labels. Y-axis labeled "Cross-Entropy Loss" (0-2.8) with readable dark text. Two lines displayed: teal/green line showing training loss and orange line showing validation loss, both with visible point markers. Red diamond marker at epoch 25 indicating minimum validation loss with text annotation. Legend on right shows "Curve Type" with both series labeled. Grid lines subtle and visible. All text is readable with good contrast against light background - PASS.

Dark render (plot-dark.png): Dark near-black background (matching #1A1A17 specification). Title remains visible with light text. Axis labels and tick labels readable in light color against dark background. Training loss line appears as teal/green and validation loss as orange - colors are identical to light render, confirming proper data color consistency. Red diamond marker and annotation text visible. Legend readable. Grid lines visible at proper opacity. All text legible in dark theme - PASS.

Score: 80/100

Category	Score	Max
Visual Quality	27	30
Design Excellence	10	20
Spec Compliance	13	15
Data Quality	15	15
Code Quality	9	10
Library Mastery	6	10
Total	80	100

Visual Quality (27/30)

• VQ-01: Text Legibility (8/8) - Font sizes explicitly set; all text readable in both themes
• VQ-02: No Overlap (6/6) - No overlapping elements; clear spacing
• VQ-03: Element Visibility (6/6) - Lines and points clearly visible and well-sized
• VQ-04: Color Accessibility (1/2) - Good contrast between lines, but colors don't use accessible palette
• VQ-05: Layout & Canvas (4/4) - Good proportions, nothing cut off
• VQ-06: Axis Labels & Title (2/2) - Descriptive labels with units (Cross-Entropy Loss)
• VQ-07: Palette Compliance (0/2) - CRITICAL: Uses #306998 (Python blue) and #FFD43B (yellow) instead of Okabe-Ito. First series must be #009E73 (brand green), second must be #D55E00 (vermillion).

Design Excellence (10/20)

• DE-01: Aesthetic Sophistication (4/8) - Generic styling with default colors, no custom design thought
• DE-02: Visual Refinement (2/6) - Library defaults with minimal customization
• DE-03: Data Storytelling (4/6) - Clear comparison of training vs validation; annotation helps identify optimal point

Spec Compliance (13/15)

• SC-01: Plot Type (5/5) - Correct line chart type
• SC-02: Required Features (4/4) - Training loss, validation loss, legend, optimal point annotation all present
• SC-03: Data Mapping (3/3) - X-axis epochs, Y-axis loss values correctly mapped
• SC-04: Title & Legend (1/3) - ERROR: Title says "pyplots.ai" instead of "anyplot.ai". Legend labels correct.

Data Quality (15/15)

• DQ-01: Feature Coverage (6/6) - Shows training loss, validation loss, and overfitting behavior
• DQ-02: Realistic Context (5/5) - Realistic neural network training curves
• DQ-03: Appropriate Scale (4/4) - Sensible epoch (1-50) and loss ranges

Code Quality (9/10)

• CQ-01: KISS Structure (3/3) - Simple linear structure: imports → data → chart → save
• CQ-02: Reproducibility (2/2) - np.random.seed(42) for deterministic output
• CQ-03: Clean Imports (2/2) - Only necessary imports (altair, numpy, pandas)
• CQ-04: Code Elegance (2/2) - Appropriate complexity, no fake functionality
• CQ-05: Output & API (0/1) - CRITICAL: Saves as plot.png instead of plot-{THEME}.png. Doesn't read ANYPLOT_THEME. Should implement theme-adaptive colors.

Library Mastery (6/10)

• LM-01: Idiomatic Usage (3/5) - Uses altair correctly but missing theme-adaptive configure methods
• LM-02: Distinctive Features (3/5) - Uses layer composition but generic for altair

Score Caps Applied

• None

Strengths

• Clear, readable visualization of training vs validation loss in both light and dark renders
• Excellent annotation (red diamond + text) highlighting optimal stopping point
• Good font sizing - all text legible at full resolution
• Proper line and point visualization
• Well-structured data generation with realistic overfitting behavior

Weaknesses

• Mandatory: Non-compliant color palette - Uses #306998 (Python blue) and #FFD43B (yellow) instead of Okabe-Ito. First series must be #009E73 (green), second must be #D55E00 (vermillion).
• Mandatory: No theme-adaptive rendering - Doesn't read ANYPLOT_THEME or apply theme colors. Must implement background, text, and grid colors based on theme.
• Mandatory: Wrong file format - Saves as plot.png instead of plot-{THEME}.png format required by altair.md
• Error: Title typo - Says "pyplots.ai" instead of "anyplot.ai"
• Missing altair mastery - Doesn't use configure_view, configure_axis with theme-aware colors per altair.md

Issues Found

1. VQ-07 PALETTE FAILURE: Use Okabe-Ito palette. First series: #009E73, second: #D55E00.
2. CQ-05 THEME FAILURE: Read ANYPLOT_THEME, apply theme-adaptive colors, save as plot-{THEME}.png per altair.md pattern.
3. SC-04 TITLE TYPO: Change "pyplots.ai" to "anyplot.ai"

AI Feedback for Next Attempt

Score of 80 is above the Attempt 3 threshold (≥70), however this implementation violates mandatory requirements and should be REJECTED:

1. Palette (VQ-07): Non-negotiable. Must use Okabe-Ito with #009E73 as first series.
2. Theme rendering (CQ-05): Required per altair.md. Must read ANYPLOT_THEME, apply theme colors to all chrome elements, save as plot-{THEME}.png (not plot.png).
3. Title format (SC-04): Must be "anyplot.ai" not "pyplots.ai"

These are architectural requirements, not style preferences. All three must be fixed before approval.

Verdict: REJECTED