Import Flow
Follow the steps in order.
Steps 3, 5 and 7 are optional.
Step 6 must be re-run after each import to update AI Search.
1
Import Articles
Upload an export from WhatsApp, Medium or CSV/JSON (Pocket, Raindrop, Instapaper…)
120 articles
💬
WhatsApp
.zip or .txt — WhatsApp export
Ⓜ
Medium
medium-export.zip — export archive
📋
CSV / JSON
Import from third-party apps
💡 Pocket: Export · Raindrop: Settings → Backup CSV · Chrome: bookmark manager → Esporta
2
Fetch Titles and Content
Visits each URL and downloads title + readable text (requires internet connection)
Completed
120
Total
120
Fetched
✓
Completed
Fetching can take several hours with many articles (~2–5 sec/URL).
You can close the page — the process continues in the background.
3
Fix Wrong Titles
optional
Delete unrecoverable articles or reset those with wrong titles
Optional
A — No title
All articles have a title.
B — Title = site name
No wrong titles detected.
4
AI Analysis
Ollama generates summaries, key points and tags for each article with text (local AI engine)
Ollama offline
Ollama is not running.
Open the Ollama app or run
ollama serve in the terminal.
100
AI Candidates
100
Analysed
—
To analyse
💡 About 10–30 seconds per article (depends on hardware and model). The process continues in the background if you close the page.
5
Clean Categories
optional
Find duplicates in categories, subcategories and tags — merge them with one click
Optional
→
Examples: "Machine Learning" = "machine-learning" — "A.I." = "AI" — "Artifical" ≈ "Artificial" (typo). Instant analysis, no AI required.
6
AI Search — Indexing
Generates semantic vectors for each article — enables concept search in AI Search
Completed
120
Totale
120
Indexed
✓
Completed
💡 About 0.5 sec per article with nomic-embed-text (local, no cloud). The process continues in the background if you close the page. Re-run after each import to index new articles.
7
Data Cleanup
Fixes artifact titles, UUIDs and normalises tags — optional, after bulk imports
Already clean
Artifact titles
Resets titles that are random technical codes (e.g. "Fxjxvs8 0ga") — will be re-fetched by the enricher
UUID titles (automatic technical IDs)
Fixes titles that are technical IDs from video links (e.g. "8387fe884422") — recovers the real title from the URL
Normalise tags
Converts underscores to hyphens, removes synonyms and tags that are too short or generic
No artifacts found — run after the next import
💡 Fast and safe operation — does not delete articles, only modifies problematic titles and tags. Recommended after every bulk WhatsApp import.