feat: configurable archive exclusion rules

[jonathanpopham]

Problem

The zip archiver has hardcoded skip directories (node_modules, vendor, dist, build, etc.) but no way for users to customize exclusions. Large repos may have project-specific directories (generated code, fixtures, datasets) that bloat the upload and slow analysis.

Proposal

• Support .supermodelignore file (gitignore syntax) at repo root
• Support --exclude flag on commands that upload: supermodel analyze --exclude 'fixtures/**'
• Support config-level exclusions in ~/.supermodel/config.yaml
• Respect .gitignore when not using git archive (the walkZip fallback)
• Show zip size in verbose output so users can see the impact

Current behavior

• git archive respects .gitignore (good)
• walkZip fallback only skips hardcoded dirs and dotfiles (insufficient)
• No user-configurable exclusions

[coderabbitai]

🔗 Related PRs

#25 - feat: wire blast-radius to /v1/analysis/impact endpoint [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!

[jonathanpopham]

Prior art from archived supermodeltools/Supermodel

Three-Layer Filtering

/apps/vscode_extension/src/extension/services/uploadService.ts — three passes:

1. Primary: VS Code workspace.findFiles() with include/exclude glob arrays
2. Secondary: .gitignore parsing via buildGitignoreMatcher() (lines 74-170) — reads ALL .gitignore files root-to-leaf, transforms patterns (anchors, negations, directory wildcards), returns ignore() matcher
3. Tertiary: additional exclude patterns via matchesGlobPattern() using minimatch

User-Configurable Globs

/apps/vscode_extension/src/extension/machines/backendProcessingMachine.ts (lines 267-304):

• supermodel.includeGlobs: ["**/*.{ts,tsx,js,jsx}", "**/*.py", "**/*.java", "**/*.rs", "**/*.go", ...]
• supermodel.excludeGlobs: ["**/node_modules/**", "**/dist/**", "**/.git/**", "**/*.md", "**/*.txt", "**/*.env"]
• supermodel.respectGitignore: boolean (default true)

Always Excluded

**/*.env files excluded regardless of config.

Batched Processing

BATCH_CONFIG = {
  DEFAULT_BATCH_SIZE: 50,
  LARGE_FILE_THRESHOLD: 5MB,
  MAX_BATCH_SIZE_BYTES: 20MB,
  BATCH_DELAY_MS: 10,
  PROGRESS_REPORT_INTERVAL: 100
}

Parallel limit of 5, GC triggered between batches via manageBatchMemory().

Key pattern for CLI

Support .supermodelignore (gitignore syntax) + --exclude flag + config-level patterns. The previous system's three-layer approach worked well — respect .gitignore first, then layer on tool-specific exclusions.