LazyDigest:以本地 Gemma 驅動的零整理成本筆記摘要工具
LazyDigest: Zero-Effort Note Digests Powered by Local Gemma
LazyDigest 透過 Ollama 在本機使用 Google 的 Gemma 3 1B,把零散的 Markdown 筆記整理成每日摘要。工具每晚 03:00 處理筆記,為每則筆記生成分類、重點句及可選行動項目,且不會將資料傳送至外部 API。作者估計,每週約可節省 75 分鐘整理筆記的時間。
Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady Submission 🦥
This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady
What I Built
I built LazyDigest, a small background utility that takes the raw, unorganized Markdown notes I drop into a folder and turns them into a short daily digest: a category, a one-line takeaway and an optional action item for each note.
Instead of organizing folders, adding tags or re-reading messy scratchpads, LazyDigest runs quietly every night at 03:00 using our Faithful Servant, Google's Gemma 3 1B open-weight model, through Ollama. I dump files into a folder, go to sleep, and wake up to a summary.
Demo
(The video shows LazyDigest reading a folder of raw Markdown notes, writing a digest file offline, and moving the processed notes into the archive folder.)
Code
LazyDigest: Local Markdown Snippet Synthesizer
LazyDigest is an open-source, local-first background utility that reads the scattered Markdown notes you drop into a folder and turns them into one short daily digest — a category, a one-line takeaway, and an optional action item per note — using Google's Gemma model running locally through Ollama.
Built for the Hacktoberfest Open-Source AI Challenge 2026 (Week 3: Slow and Steady).
What It Does
-
Reads every
.mdfile indata/input_notes/. -
Analyzes each note with
gemma3:1bvia Ollama, without sending data to external APIs. -
Writes the results into
data/output_digests/digest_output.md, then moves processed notes intodata/processed/so they are never read twice.
Lazy Payoff
- Time saved: roughly 75 minutes a week of re-reading raw notes, tagging them, and copying out action items (our own estimate).
-
Set and forget: run
--daemonand it digests your notes every night at 03:00 while you sleep.
Setup &
…
How I Built It
-
Local inference:
ollama-pythonwithgemma3:1b. - No API keys: no cloud services or credentials.
- CPU friendly: temperature 0.1 for short, structured JSON on CPU-only laptops.
- Robust parsing: strips code fences, fills missing fields, and falls back to the note's first line if the model output is not valid JSON.
-
Archiving and scheduling: the
schedulelibrary runs it daily at 03:00 and moves processed files intodata/processed/so nothing is read twice. - Tests: the model is mocked, so the tests run without Ollama.
Why Does Open Innovation Matter?
- Privacy: personal scratchpads stay on my own machine.
- Offline: it works without an internet connection.
- Ownership: an open-weight model keeps a local tool working without API deprecations or surprise bills.
The Lazy Payoff
- Time saved: about 75 minutes a week (my estimate) that I used to spend re-reading notes and picking out action items.
- Effort saved: no more manual sorting or tagging; I just drop files in the folder.
Prize Categories
- Best Use of Gemma
來源:DEV Community · dev.to