Reading Inbox Synthesizer

An Obsidian plugin that turns your web-clippings backlog into reading memory: batch summaries, cross-article themes, and weekly digests.

The Problem

When Pocket shut down in 2025, a lot of people moved their reading into Obsidian with the Web Clipper — files they own, no server that can disappear. But the backlog moved with them. I kept seeing the same thing described on forums: the "to-read graveyard" just relocated to the vault.

You clip dozens of articles and never process them. A month later you have 40 unread clippings and no idea which ones still matter. This plugin reads the whole backlog and gives you one synthesis note: what each clipping says, which themes run across them, what you saved this week, and what has gone stale.

Features

• Reading inbox — Every clipping at a glance with its topics and estimated read time
• Themes — Topics shared across two or more clippings, with an AI-written consensus and the tension between sources
• This week — What you saved in the current week (Monday start)
• Needs attention — Clippings that have sat unread past your stale threshold, oldest first
• Summaries — A short summary and the key claims of every synced clipping

How It Works

1. Clip articles into Obsidian as you normally would (the official Web Clipper works out of the box)
2. Run Sync clippings from the command palette to extract summaries, claims, and topics
3. Run Generate reading report from the command palette or the ribbon icon
4. A Reading Synthesis.md note is written to your vault root and opened
5. Re-run either command any time — syncing is incremental, so unchanged clippings are never re-processed

Setup

1. Install the plugin from Obsidian Community Plugins
2. Go to Settings → Reading Inbox Synthesizer
3. Select your AI provider (Anthropic, OpenAI, OpenRouter, or custom)
4. Enter your API key
5. Set your clippings folder (default Clippings) or clipping tag (default clipping)
6. Run Sync clippings, then Generate reading report

Supported AI Providers

• Anthropic — Claude models (recommended: claude-sonnet-4-6)
• OpenAI — GPT models (recommended: gpt-4o-mini)
• OpenRouter — Access to many models including free options (recommended: meta-llama/llama-4-maverick)
• Custom — Any OpenAI-compatible endpoint

Free vs Pro

The free tier allows 10 total syncs. This is a one-time lifetime allowance — it does not reset monthly. A sync that fails does not consume one of your free uses: the use is reserved before the request and refunded if the sync does not complete.

Generate reading report is always free and unlimited. It makes no LLM call, so building a report from clippings you have already synced never consumes a sync — including after the free syncs are used up.

The free allowance is stored in this vault's local plugin data. Each vault has its own allowance. Removing the plugin's local data may reset the recorded usage. There is no account and no server-side record of your usage.

Pro is a one-time purchase that removes the sync limit — no subscription, and no account. Your license key is verified offline on your own machine; activation never requires network access. Pro covers the plugin only; AI provider usage is billed separately by the provider whose API key you configure. Upgrade at ibrh96.gumroad.com/l/yeulsi.

Privacy

• Your clippings never leave your device except to your chosen AI provider
• No servers, no databases, no telemetry
• Your API key is stored locally in Obsidian's data storage

License

The source code in this repository is licensed under the MIT License — see LICENSE.

EULA.md governs the official Pro purchase: activation, license keys, refunds, warranty, and support. It does not restrict the MIT-licensed source code.