techread

A local CLI tool that fetches daily technical blogs/writings, ranks them for a busy reader, and (optionally) summarizes them using a local LM Studio model.

Quickstart

1) Install (recommended via pipx)

pipx install -e .

Or using a venv:

python -m venv .venv
source .venv/bin/activate
pip install -e .

2) Add sources

techread sources add https://www.allthingsdistributed.com/atom.xml --name "All Things Distributed" --weight 1.3 --tags "distributed,sre"
techread sources add https://netflixtechblog.com/feed --name "Netflix TechBlog" --weight 1.1 --tags "systems,platform"

3) Fetch and view a digest

techread fetch
techread digest --today --top 10

4) Summarize a post (requires LM Studio running locally)

Start LM Studio and make sure a supported model is available, then:

techread summarize 42 --mode takeaways

Config

techread looks for a config file at:

• macOS/Linux: ~/.config/techread/config.toml
• Windows: %APPDATA%\techread\config.toml

Example config.toml:

db_path = "~/.local/share/techread/techread.db"
cache_dir = "~/.local/share/techread/cache"
llm_model = "mistral-small-3.2"
default_top_n = 10
topics = ["distributed systems", "spark", "kafka", "data platform", "reliability", "llm"]
min_word_count = 500

Commands (high level)

• techread fetch
• techread rank --today
• techread digest --today
• techread summarize <id>
• techread open <id>
• techread mark <id> --read|--saved|--skip|--unread
• techread sources ...

Source hygiene:

• techread sources purge [--dry-run] removes posts with word_count < min_word_count.

Notes

• RSS/Atom is preferred. techread fetches linked articles, extracts readable text via trafilatura, and stores it locally in SQLite.
• Summaries are cached by content hash so repeated runs are fast.

MIT License.