Create a new, empty Counter object. And if given, count elements from an input iterable. Or, initialize the count from another mapping

Fallback: assign feeds to their search target subcategory without Claude.

Build the categorization prompt with hybrid approach. If user has existing categories, instructs AI to reuse them only when accurate. Ot

Merge duplicate keys from _SUBREDDITS_RAW into a single dict. The raw list alternates between (category, subcategory) tuples and [name] lists.

Merge duplicate keys from _PODCASTS_RAW into a single dict.

Merge duplicate keys from _NEWSLETTERS_RAW into a single dict.

Merge duplicate keys from _CHANNELS_RAW into a single dict.

Build Feedly search queries from taxonomy. Feedly search works best with short queries (1-2 words).

Build a human-readable summary and preview from tool result. Returns tuple: (summary_string, preview_list_or_none)

Call Claude API with tools using streaming for real-time response delivery. Args: user_id: User ID for tool execution messag

Use Claude to categorize discovered feeds into target subcategories.

Categorize a batch of feeds using Claude.

Categorize a batch of feeds using Claude.

Categorize a batch of feeds using Claude.

Categorize a batch of feeds using Claude.

Categorize a batch of feeds using Claude.

Use Claude to categorize feeds that don't have categories.

Use Claude to categorize feeds that don't have categories from Feedly search.

Use Claude to categorize feeds that don't have categories.

Use Claude to categorize feeds that don't have categories.

Check Redis cache for limit status, fall back to MongoDB.

apps/newsletters/models.py: Convert timestamp to datetime. If timestamp is empty or invalid, use current date.

Clean up old prometheus files from dead processes

Inner clustering logic, separated so SoftTimeLimitExceeded is caught cleanly.

Compute composite trending scores for PopularFeed entries on-the-fly. Combines three signals (normalized to 0.0-1.0): - Subscription vel

Count regex classifiers by scope. Very few regex docs exist so all counts are fast.

Generate error message for archive/pro users hitting daily limit.

Decode a Google News redirect URL to the actual article URL. Uses Google's batchexecute API with signature/timestamp from the article page.

Detect newsletter platform from feed URL.

Phase 1: Search iTunes by genre with multiple keywords per genre.

Phase 2: Search by taxonomy keywords to fill niche categories.

Search by taxonomy keywords to fill niche categories.

Discover Ghost publications via the Ghost Explore sitemap. 1. Fetch https://explore.ghost.org/sitemap.xml for all /p/{slug} URLs 2. F

Discover Medium publications by scraping TopPub.xyz tag pages. TopPub.xyz is a leaderboard of ~20,000 Medium publications organized by tags.

Paginate through /subreddits/popular to get top subreddits.

Query the Feed table for well-read RSS feeds and output JSON candidates.

Browse Substack categories to discover top newsletters.

Ask Claude to suggest real RSS feed URLs, then verify each one via HTTP. For each underpopulated subcategory, Claude suggests feeds it knows

Get expiry timestamp for keys.

Apply stored XPaths to HTML and extract story dicts.

Fetch and parse OPML files from GitHub repositories.

Fetch and parse GitHub OPML collections for feed discovery.

Fetch page HTML with proxy fallbacks for forbidden feeds.

Download an image and return it as a base64-encoded string with its media type. VLMs accept images in two ways: 1. By URL ({"type": "url",

Fetch playlist snippet from the YouTube API.

Fetch a story page using ScrapingBee residential proxy.

Identify subcategories with fewer than min_feeds feeds.

Convert category feeds response into flat list of feed entries.

convert a datetime into an RFC 822 formatted date Input date must be in GMT.

Format time remaining until daily reset at midnight in user's timezone.

Generate a short title from the first query.

Phase 2: Generate feeds for one category with its subcategories.

Generate effective Feedly search queries for each subcategory. Uses Claude to generate short, keyword-based queries that work well with

Phase 1: Generate taxonomy structure (categories + subcategories, no feeds).

Get Reddit OAuth2 app-only access token.

Get full content of a specific archive.