Skip to content

Features

How LociArc turns a link into knowledge you can reuse

This page describes each feature together with technical parameters taken from the running system.

01 / 08

Capture

Save where you are reading, without switching apps.

Specifications

  • System share sheet on iOS and Android
  • Manifest V3 browser extension for Chrome, Edge, Brave
  • Telegram bot: send a link to save it
  • Source detection: articles, YouTube, X, images
  • iOS · Share sheet
  • Android · Share
  • Chrome · Edge · Brave
  • Telegram bot
POST /vault/itemsLociArc API
02 / 08

Content processing

Content is extracted, summarised and classified in the background. The app shows the active stage.

Specifications

  • Stages: scraping, summarizing, embedding
  • 13 fixed topics, plus the reason for saving
  • Reading time estimate and cover image from og:image
  • Retry any failed item; unused AI units are refunded
  1. capture

    Capture

    Share from any app, save from the browser extension or send the link to the Telegram bot.

  2. scraping

    Extract

    Fetch the page and pull out the main text, title and cover image. Articles, YouTube, X and images are recognised.

  3. summarizing

    Summarise and classify

    Write a short summary, assign one of 13 topics, infer why it was saved and estimate reading time.

  4. embedding

    Index

    Convert the content into a 1,536-dimension vector stored in PostgreSQL with an HNSW index.

  5. succeeded

    Ready

    The item appears in your library, searchable by meaning and available as a source for questions.

iEach save uses 2 AI units. If processing stops part-way, unused units are refunded.

04 / 08

Cited Q&A

Answers are composed only from the relevant items in your library.

Specifications

  • Retrieves the 8 nearest items
  • Cosine similarity threshold of 0.4
  • Numbered citations link to the source item
  • Insufficient data: a clear reply and a refund
  1. 01 · embed(q)Vectorise the question
  2. 02 · top_k = 8Retrieve the 8 nearest items
  3. 03 · cos > 0.4Filter by similarity threshold
  4. 04 · answer + [n]Compose with citations

Illustrative example

Summarise what I have saved about sleep and productivity.
top_kcos
  • 10.82
  • 20.74
  • 30.61
  • 40.38
  • 50.33
  • 60.27
  • 70.22
  • 80.18

Your saved articles agree on one point: getting 7 to 9 hours of sleep has a clearer effect on focus than working longer hours12. Two other pieces suggest a fixed wake-up time and limiting screen light in the hour before bed3.

Sources

  1. 1Does sleeping less let you get more done?0.82
  2. 2Evening habits of deep workers0.74
  3. 3Circadian rhythm and blue light0.61
05 / 08

Knowledge Planet

An overview chart of your library by topic, updated as you save and read.

Specifications

  • Shelf height: items in the topic
  • Brightness: recency
  • Satellites: items saved this week
  • Tap a shelf to open the collection
Knowledge Planet screen in the LociArc app
06 / 08

Daily brief

A short digest each day so saved items do not sink out of sight.

Specifications

  • Scheduled in your time zone, in 15-minute steps
  • Leads with a topic cluster when 2 or more related items are unread
  • Delivered via Telegram, falling back to email
  • The digest paragraph uses 1 AI unit; the item list is always sent
07:00
Scheduled in your time zone, in 15-minute steps
07 / 08

Memory Box and time capsule

Two resurfacing mechanisms that bring older knowledge back at the right time.

Specifications

  • Freeze a note on an item, reopen it on the set date
  • Time capsule: items saved exactly one year ago
  • Delivered in your time zone
09/10/202509/10/2026

Δt = 365 d

08 / 08

Sharing and data

You decide which items are public, and you can take your data with you at any time.

Specifications

  • Share links use a random 16-character slug, 96 bits
  • Revoke a share link at any time
  • Export to JSON or Markdown
  • Delete your account inside the app

GET /share/k3Q9xT0bWm2LpZ7e

randomBytes(12) · base64url · 16 × 6 = 96 bit

Coming soon

Get notified when LociArc arrives on the App Store and Google Play

One email on release day. No marketing, and your address is never shared.

Your email is used only for the launch notice.