RSS Reader

AI-assisted RSS reader that generates daily markdown digests with article summaries, tags, and interestingness ratings.

Setup

npm install

Option A: Ollama (default — free, local)

# Install Ollama: https://ollama.com
ollama pull qwen2.5:7b          # summarization
ollama pull nomic-embed-text     # embeddings (novelty scoring & dedup)

No API key needed. The reader defaults to Ollama when no ANTHROPIC_API_KEY is set.

Option B: Anthropic API

export ANTHROPIC_API_KEY=sk-ant-...

Ollama is still used for embeddings regardless of LLM provider.

Usage

Manage feeds

Add feeds one by one:

npm run rss -- add https://example.com/feed.xml
npm run rss -- remove https://example.com/feed.xml
npm run rss -- list

Or bulk-import from a file (one URL per line, # comments and blank lines are skipped):

npm run sync-feeds                         # reads FEED.md in project root
npm run sync-feeds -- /path/to/feeds.txt   # custom file

Generate a digest

npm run digest

The digest pipeline:

1. Fetch all feeds and parse new articles (GUID-based dedup)
2. Extract full article content via Readability
3. Embed each article for novelty scoring and deduplication
4. Deduplicate — skip articles too similar to existing ones (cosine similarity ≥ 85%)
5. Summarize each article and generate tags via LLM
6. Output a markdown digest to ~/.rss-reader/digests/YYYY-MM-DD.md, sorted by novelty

Example output:

## [Article Title](https://example.com/article) ★★★★☆

**Source**: Blog Name | **Tags**: `ai`, `rust` | **Published**: 2026-03-25 | **Novelty**: 75%

This article introduces a novel approach to...

Publish to a blog

Point RSS_DIGESTS_DIR to your blog's content directory to generate blog-ready posts directly. Digests include Astro-compatible frontmatter with title, date, description, and series: "Daily Feed".

export RSS_DIGESTS_DIR=~/Dev/my-blog/src/content/blog
npm run digest
# → ~/Dev/my-blog/src/content/blog/daily-feed-2026-03-25.md

CRON (daily at 7am)

0 7 * * * cd /path/to/rss-reader && npm run digest >> /tmp/rss-reader.log 2>&1

Reset

To start fresh, delete the database:

rm ~/.rss-reader/rss-reader.db

Configuration

All via environment variables (or .env):

Variable	Default	Description
LLM_PROVIDER	auto	ollama or anthropic. Auto-detected if not set.
OLLAMA_MODEL	qwen2.5:7b	Ollama model for summaries
OLLAMA_BASE_URL	http://localhost:11434	Ollama server URL
ANTHROPIC_API_KEY	—	Anthropic API key (triggers auto-detect)
ANTHROPIC_MODEL	claude-sonnet-4-20250514	Anthropic model for summaries
EMBEDDING_MODEL	nomic-embed-text	Ollama model for embeddings
DEDUPE_THRESHOLD	0.85	Similarity above this → duplicate, skip
MAX_ARTICLE_AGE_DAYS	7	Skip articles older than N days
MAX_ARTICLES_PER_FEED	20	Max articles to process per feed
RSS_DATA_DIR	~/.rss-reader	Database and data directory
RSS_DIGESTS_DIR	~/.rss-reader/digests	Digest output directory

Architecture

src/
├── index.ts          CLI entry point (commander)
├── config.ts         Environment-based configuration
├── db.ts             SQLite schema & queries (better-sqlite3)
├── fetcher.ts        RSS parsing + full article extraction
├── summarizer.ts     Orchestrates LLM summarization
├── digest.ts         Markdown digest generation
├── embeddings.ts     Ollama embeddings, cosine similarity, novelty scoring
└── llm/
    ├── types.ts      LLMProvider interface (abstraction layer)
    ├── shared.ts     Prompt template + response parsing
    ├── index.ts      Provider factory
    ├── anthropic.ts  Anthropic/Claude implementation
    └── ollama.ts     Ollama implementation (local models)
scripts/
└── sync-feeds.sh     Bulk-import feeds from a text file

Data

Stored in ~/.rss-reader/:

• rss-reader.db — SQLite database (feeds, articles, tags, embeddings, digests)
• digests/ — Generated markdown files