Decisions

1. LLM: default to aistudio.google.com, as it has generous free usage
2. Gmail API for this util
3. For Utility 2, step 5, put these summaries in D:\FSS\KB\Business\_WorkingOn\Feed\Medium with date filenames YYYY-MM-DD
4. I have put much work into defining my business identity and strategy. While still a work in progress, these KB folders should provide a good starting point to define what I am working on, and what is important. Review ALL notes in each of d:\FSS\KB\Business\01_Identity\ and d:\FSS\KB\Business\02_Strategy\. Also review D:\FSS\KB\Business\04_AI\Optimizing AI Benefits.md defining the importance of AI, and D:\FSS\KB\Business\02_Strategy\Progress Acceleration.md . We need these configuration details to define what is important and over time improve the ranking and content of the summaries, perhaps with a template. An AI skill might be the proper tool, but if possible I would like this to be a utility that is automatically executed .

---

Background, Question

Everyday, I receive these medium daily digests (See attached screenshot, at C:\tmp\ScreenShots\comet_InbUkSZwWr.png). They include a dozen or so articles curated for my interests. But some take 10 to 20 minutes or more to read. Is there a way to automatically summarize each of these articles, and rank them according to my interests and how well they help my business mission? The ideal output would be an Obsidian markdown file that has a section for each article, including the URL to the full article.

Task

Based on the Response from Perplexity below, carefully detail a plan to automate this task, and I will review and approve it before you implement it. It is probably best to create two utilities. The lower level utility does the summarization and creating a section of a markdown note for a single source. The source initially could be an e-mail link, but this should be generalized to handle different types of input including PDFs, word documents, Google documents, YouTube videos, audio files etcetera. This specific implementation should handle processing of my medium digest articles I receive daily.

---

Response from Perplexity

Yes, you can automate this end‑to‑end: grab each article URL from the Medium digest, summarize and score it with an LLM, then emit a daily Obsidian‑ready markdown note.[mail.google]​

Overall approach

You’ll want a small script (Python is ideal) that runs on a schedule (e.g., via Windows Task Scheduler or cron) and does this workflow:[mail.google]​

1. Fetch your Medium Daily Digest from Gmail (using Gmail API, filtered by subject like “Medium Daily Digest”).[mail.google]​

2. Parse the email HTML to extract article titles and URLs.[mail.google]​

3. For each URL, fetch the article content (via Medium’s HTML; optional: use an extractor like Newspaper3k or Readability).[mail.google]​

4. Call an LLM API (OpenAI, Anthropic, etc.) with a custom prompt that:

   • Summarizes in 3–6 bullet points.

   • Extracts key takeaways tied to your business mission.

   • Produces a relevance score (e.g., 1–10) plus tags.

5. Sort the results by score.

6. Generate a markdown file in your Obsidian vault, one section per article, including URL.[mail.google]​

Ranking by “business mission”

You’ll encode your mission and interests directly into the LLM prompt, e.g.:

• Briefly describe your business mission and top priorities (lead gen, product design, AI workflow optimization, etc.).

• Ask the model to return:

  • business_relevance_score (1–10).

  • why_it_matters_for_mission (2–3 sentences).

  • tags (e.g., ["AI-agents","productivity","marketing"]).

You can then sort by that numeric score in code before writing to markdown.

Example markdown structure for Obsidian

Your script can emit something like:

text

# Medium Daily Digest – 2026-02-20 ## 9.2 – OpenClaw: The Open-Source AI Agent That Grew 190K GitHub Stars in 14 Days - URL: https://medium.com/... - Read time: 11 min - Tags: ai-agents, product-strategy, dev-tools ### Summary - Point 1 … - Point 2 … - Point 3 … ### Why it matters for my mission Short explanation focused on your business goals. --- ## 8.5 – The “Ego Death” Secret: Why Anthropic Is 1,000x More Efficient Than Google - URL: https://medium.com/... …

Obsidian will treat this like any other note; you can add daily‑note links or tags easily.

Concrete implementation options

You have a few good paths given your stack:

• Python script + Gmail API + LLM API + filesystem write:

  • Libraries: google-api-python-client (Gmail), beautifulsoup4 or selectolax (HTML), requests for HTTP, chosen LLM SDK.

  • Run via cron/Task Scheduler and point the output folder at your Obsidian vault.

• SvelteKit/web app variant (optional):

  • Build a small internal tool where you paste the digest HTML (or forward to a custom address), then the backend does the same pipeline and writes/returns markdown for you to drop into Obsidian.

What I can do next for you

If you’d like, I can:

• Draft a Python script skeleton that:

  • Accepts the raw digest email body (pasted or via API).

  • Extracts the links.

  • Calls an LLM (you choose provider).

  • Outputs the exact markdown format you want.

Tell me:

1. Which LLM/API you prefer (OpenAI, Anthropic, etc.), and

2. Whether you’re okay authenticating to Gmail via the Gmail API, or prefer a “paste the digest email source into a script” workflow.