Skip to content

rpings/Mimir

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

41 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Mimir

Drink from the well of knowledge in the AI era.

Mimir is an AI-powered knowledge base that automatically collects, processes, and archives information from ArXiv, GitHub, Hacker News, and Chinese AI news (量子位). Entries are organized into a Notion kanban board with AI-generated summaries, topic classification, and priority ratings.

Python 3.11+ License: MIT

Quick Start

git clone https://github.com/rpings/Mimir.git
cd Mimir
pip install -e .
cp .env.example .env   # edit with your keys
mimir setup            # create Notion databases
mimir collect          # collect and process entries

Required Environment Variables

Variable Description
LLM_API_KEY API key for LLM provider (e.g., DeepSeek)
NOTION_TOKEN Notion integration token
NOTION_ENTRIES_DB_ID Database ID for collected entries
NOTION_PARENT_PAGE_ID Parent page for mimir setup to create databases under

CLI Commands

Command Description
mimir setup Create Entries and Reports databases in Notion
mimir collect [--dry-run] Fetch sources, enrich with AI, write to Entries DB
mimir report [--week WNN] Generate weekly report, publish to GitHub Pages, write to Reports DB

collect

Downloads new content from all sources, deduplicates against existing Notion entries, runs AI enrichment (topic classification, summarization, priority scoring), and writes new entries to the Entries database.

Use --dry-run to preview what would be collected without writing anything.

setup

Creates the required Notion databases (Entries DB and Reports DB) with the correct schema under the specified parent page. Run once after initial configuration.

report

Generates a weekly HTML report summarizing trends, featured papers, notable projects, and important news. Reports are published to GitHub Pages and linked from the Reports database.

Configuration

See mimir.toml for all configuration options. Each option includes inline documentation and the corresponding environment variable override.

Architecture

Layer Technology
Sources ArXiv API, GitHub Trending, Hacker News, 量子位
AI Processing DeepSeek (3 prompts: paper, project, news)
Storage Notion (Entries DB + Reports DB)
Scheduling GitHub Actions (daily collect, weekly report)
Reports Jinja2 + HTML → GitHub Pages

See the design docs for detailed architecture and technical decisions.

License

MIT License - see LICENSE file.

About

Drink from the well of knowledge in the AI era

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors