The Flask blueprints aren't used in production — engine.py with
Responder is what Fly actually runs.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Swagger UI Execute needs responses defined to display results.
Auto-detect search query params for /api/search routes.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Schema at /api/schema now introspects api.router.routes for paths
and docstrings instead of maintaining a static dict. Skips catch-all
and internal routes.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Schema served as JSON at /api/schema, docs rendered with Redoc at /api.
No more inline YAML — clean separation.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Inline OpenAPI spec avoids catch-all route conflict. Elements UI
with sidebar layout, no external schema fetch needed.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Avoids catch-all route intercepting /schema.yml.
Swagger UI at /api/docs, schema at /api/schema.yml.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
/api/icon/{article_path} didn't match paths with slashes like
essays/2026-03-25-foo. Changed to {article_path:path}.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Uses responder's built-in RateLimiter. Returns 429 with Retry-After
header when exceeded. Adds X-RateLimit-Limit and X-RateLimit-Remaining
headers to all responses.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
find_adjacent_posts result was never used in template (like
find_related_posts before it). Added _cached_render() to cache
parsed markdown in memory — content only changes on deploy so
there's no reason to re-parse on every request.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
find_related_posts was doing TF-IDF comparison against all 250+ essays
on every request (1 second each) but the result was never rendered in
the template. Essays now serve at 105 req/s instead of 19.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The catch-all route wasn't passing article_themes to the post template.
Now looks up detected themes from the themes cache for each essay.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Search:
- Build search index at startup instead of scanning 727 files per query
- Add /api/search/autocomplete for whole-site title matching
- Autocomplete now server-side, covers all content not just essays
- Remove client-side /api/blog preload from search page
AI section:
- Create 5 missing index.md files for navigational orphan directories
- Update "277 file" count to actual 352 across all references
- Tone down Lumina poetry index to match actual content quality
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Security:
- Add path traversal protection to serve_data_file and og_image routes
- Bound OG image cache to 256 entries to prevent memory growth
Performance:
- Pre-build legacy URL lookup tables at startup instead of rglob on every 404
- Extract _generate_pdf helper, eliminating duplicate PDF code
Templates:
- Add dark mode support to graph.html and error.html
Content:
- Remove third-person sidenotes from yoga-meditation.md, great-music.md,
and idea_amplification essay
- Fix sidenote formatting in yoga-meditation.md
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Recent writing now pulls 5 latest posts dynamically with icons
- Collapsed theme/beyond sections into flowing Threads paragraph
- Added .muted class for subtle text styling with dark mode support
- Removed redundant footer nav links
- Simplified subtitle styling
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Search endpoint now scans all markdown files in data/ directory
- Results show section labels (Essays, Software, Themes, etc.) instead
of hardcoded "Essay" type
- Updated placeholder text to reflect whole-site search
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Themes dropdown mirrors Browse structure with tree-item icons
- Populated from data/themes/ directory via /api/themes endpoint
- Search link added between Archive and Themes
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Warms all caches (blog, sidenotes, outlines, quotes, connections,
terms, themes) in background thread on startup.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Handle .md and .pdf extensions inside catch-all since path:path
is greedy. Add WeasyPrint availability check for PDF button.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* Initial Responder port — core content serving working
Homepage, markdown pages, directory listings, image galleries
all rendering. Flask template compatibility via RequestWrapper
and FakeConfig shims.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* Complete Responder port — all routes working
Ported all routes from Flask blueprints: archive, search, OG images,
RSS, sitemap, robots.txt, API endpoints, directory browser.
All 25+ routes returning 200.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* Update Dockerfile for Responder, warm caches on import
CMD now runs engine_responder:api via uvicorn.
Cache warming happens on module import for production.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* Rename engine_responder.py to engine.py
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* Update colophon for Responder
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* Fix essay pages, OG images, and PDF export
- Fix find_related_posts and find_adjacent_posts call signatures
- Fix OG image route to use path:path for nested paths
- Add PDF export route with WeasyPrint
- All test plan items passing
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Replace gunicorn/gevent with uvicorn for ASGI support
- Add asgiref to wrap Flask WSGI app for uvicorn compatibility
- Enable auto-reload in docker-compose for development
- Add current_path to all content render_template calls
- Simplify breadcrumb template to exclude current page
- Only show breadcrumbs for nested paths (> 1 segment)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Removed the engine_new.py file as part of restructuring.
- Implemented actual search functionality in api.py, allowing for searching through blog posts with relevance scoring and snippet extraction.
- Enhanced caching in cache.py with a clear_cache function and improved content cleaning for search indexing.
- Deleted outdated homepage-revised.html template and ensured search.html template displays search results with additional metadata such as date and matches found.
- Detect directory paths (ending with / or pointing to directories)
- Use generate_folder_icon() for directories instead of regular article icons
- Read index.md files for directory titles when available
- Fallback to directory name for folder icon generation
- Fixes AI personalities link and other directory links to show proper folder icons
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add themes directory to content cache for proper icon generation
- Remove colored dots from themes/index.md to enable icon display
- Fix JavaScript to handle links wrapped in strong tags
- Improve icon positioning with refined margins and alignment
- Prevent icons from loading on index/archive pages
- Fix navigation link to point to /themes instead of /themes/
- Optimize icon spacing and vertical alignment across devices
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Fix data URL display issue by using inline SVG instead of img src
- Add LRU cache to icon API for better performance
- Expand icon coverage to all internal links (not just essays)
- Add software, poetry, and talks directories to content cache
- Implement fallback logic to read files directly when not in blog cache
- Fix Requests icon to use full title "Requests: HTTP for Humans"
- Improve icon positioning with negative margins and vertical alignment
- Add graceful fallback icons for missing content
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Enhanced simple_extract_excerpt() function to filter markdown images
- Added support for linked images [](url)
- Improved line-by-line filtering to find meaningful content
- Reduced excerpt length to 150 characters for better display
- Removed prebuild cache system to simplify architecture
- Fixed markdown filtering in all excerpt generation functions
- Add cached markdown title extraction with @lru_cache(maxsize=1000)
- Replace expensive render_markdown_file() calls with fast regex title extraction
- Only read first 1000 chars instead of full file content for title extraction
- Essays directory and other large directories now load significantly faster
- First load: fast title extraction, subsequent loads: instant cache lookup
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add @lru_cache(maxsize=1000) to generate_unique_svg_icon() in svg_icon_generator.py
- Add @lru_cache(maxsize=500) to generate_folder_icon() in engine.py
- Import functools.lru_cache in both files
- Directory pages with many items (like /essays) now load much faster on repeat visits
- Icons generated once per unique title, then served from memory cache
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add reading progress bar for longer essays with smooth animation
- Implement search result snippets with highlighted query terms
- Tighten directory listing styles for better compactness
- Add back-to-parent navigation links in directory views
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Added a new SVG icon generator to create unique procedural icons for posts and directories.
- Implemented a grid layout for directory contents with hover effects and responsive adjustments.
- Updated archive and connections templates to display unique icons alongside article titles.
- Improved styling for archive posts, including flexbox layout for better alignment.
- Enhanced directory listing with icons and improved visual hierarchy.
- Refactored post template to include an icon next to the post title and added parent navigation links.
- Updated styles across templates for consistent icon sizes and responsive behavior.
- Use actual title from index.md content instead of directory name
- Falls back to directory name if no index.md or title exists
- Fixes issue where pages like /talks/#major-talks showed no title
- Now "Conference Talks" will properly display instead of generic "Talks"
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
kennethreitz