§ tool · point 53 courier

A customizable browser scout.

Customizable Selenium-based query program. Targets specific segments of a web navigation path, not just the links. Grafts cleanly onto Collector's engine to know what to look for. Reads, scrolls, navigates. Browser-Credentialed (or Sandboxed). Visible (or Headless). Specific (or Generalized). Interactive (or Automatic).

§ the pivotal approach · scoped natural language paths

Prompt-Tuned RSS Curation.

Most of the time, you query and get a list to go through or trust a provider behind a curtain. Courier starts the show and reveals an easy to tune query program that listens to your goal, not just the search. Pre-prompt with next-stage analysis. Wait for page loads or navigate away from unrelated sites to get closer to the content you need. Take control of your own research processes with the help of Courier.

* > courier query "Personal data takeout and archive" --target "I want complete offline archives of my photos, mail, and documents from the platforms I use, kept in open formats I can still read in twenty years" --analysis-prompt "A practical takeout plan: what to export from each platform, which open formats to keep, and how to verify the archives are complete" --breadth 8 --depth 2 --chat

Config: /home/user/.config/point53/courier
Profile: sandboxed (ephemeral)
Pause enabled: 15s per page (e to extend to 150s)
Keep-open: browser will remain open after pipeline completes
RSS browser: firefox (visible)
Summarization browser: firefox (visible)
Run directory: /home/user/.local/share/point53/courier/runs/2026-06-07T18-22-57_Personal-data-takeout-and-archive-I-want-complete-offline-archives-of-my-photos-mail-and-documents-from-the-platforms-I-use-kept-in-open-f
Query #60 recorded in /home/user/.local/share/point53/courier/courier.db
Cleared ephemeral state for fresh query
Querying configured search provider…
Paused 15s — interact with the browser.
    [any key] continue | [e] extend to 150s
    Pause expired, continuing...
Links maintained for evaluation.
Generating Custom RSS with nemotron-3-nano:30b
Customized RSS Converter Executed.
RSS XML parsed from AI model output!
Feed written to /home/user/.local/share/point53/courier/runs/2026-06-07T18-22-57_Personal-data-takeout-and-archive-I-want-complete-offline-archives-of-my-photos-mail-and-documents-from-the-platforms-I-use-kept-in-open-f/feed-1.xml
Processing 8 articles via Collector engine
pdf enabled but pdfplumber is not installed: No module named 'pdfplumber'
Install the PDF extra: uv tool install 'p53-collector[pdf]' --index https://dist.point53.ai/simple/
Paused 15s — interact with the browser.
    [any key] continue | [e] extend to 150s
    Pause expired, continuing...

  ⋮  2 more pages · paused for review between each  ⋮

Could not get page https://blog.dataportability.example/takeout-guide/
TRY AGAIN, ERROR: site unreachable (dnsNotFound) — skipping

  ⋮  remaining pages fetched · 2 sites skipped (unreachable)  ⋮

ARTICLE DB UPDATE DONE
Links cleared for population from latest RSS feed.
https://takeout.google.com/
Paused 15s — interact with the browser.
    [any key] continue | [e] extend to 150s
    Pause expired, continuing...
https://privacy.apple.com/
https://github.com/immich-app/immich

  ⋮  13 sources visited · paused between each · 2 skipped (unreachable)  ⋮

Generating Custom RSS with nemotron-3-nano:30b
Feed written to /home/user/.local/share/point53/courier/runs/2026-06-07T18-22-57_Personal-data-takeout-and-archive-I-want-complete-offline-archives-of-my-photos-mail-and-documents-from-the-platforms-I-use-kept-in-open-f/feed-2.xml
Processing 8 articles via Collector engine

  ⋮  7 pages · paused for review between each  ⋮

ARTICLE DB UPDATE DONE
Running distill on ephemeral storage
DISTILL DONE (13 articles)
Transcript written to /home/user/.local/share/point53/courier/runs/2026-06-07T18-22-57_Personal-data-takeout-and-archive-I-want-complete-offline-archives-of-my-photos-mail-and-documents-from-the-platforms-I-use-kept-in-open-f/transcript.md
Sources written to /home/user/.local/share/point53/courier/runs/2026-06-07T18-22-57_Personal-data-takeout-and-archive-I-want-complete-offline-archives-of-my-photos-mail-and-documents-from-the-platforms-I-use-kept-in-open-f/sources.jsonl (13 articles)
Articles archived to courier.db under Query #60
Summarizing with gemma4:31b
Based on the gathered guides and documentation, here is a practical takeout plan for building offline archives you actually control.

### **The Shape of the Plan**
Every major platform now ships a data-export path — the differences are in completeness and format. The advice that recurs across the sources: export first, normalize to open formats second, verify third, and only delete anything at the source once all three are done.

### **1. Photos**
*   **Export:** Request the full archive at original quality (https://takeout.google.com/). Expect it to arrive split across several archives, with dates and locations in JSON sidecar files rather than inside the images.
*   **Normalize:** Merge the sidecar metadata back into the image files as EXIF so each photo stays self-describing — a JPEG with embedded EXIF needs no companion app to be read in twenty years.
*   **Going forward (optional):** A self-hosted Immich library (https://github.com/immich-app/immich) can ingest the takeout directly and gives phones somewhere to sync that you own.

### **2. Mail and Documents**
*   **Mail:** Prefer mbox or EML — both are plain text, and every mail client of the last thirty years can open them.
*   **Documents:** PDF for finished things, ODF or plain Markdown for living ones. Skip proprietary formats on the way out.

  ⋮  contacts & calendars to vCard/iCS · per-platform export paths (https://privacy.apple.com/ among them) — condensed for display  ⋮

Post-analysis using gemma4:31b

  ⋮  verification checklist · file counts & checksums per export · a yearly open-and-read drill · refresh cadence per platform — condensed for display  ⋮

Report written to /home/user/.local/share/point53/courier/runs/2026-06-07T18-22-57_Personal-data-takeout-and-archive-I-want-complete-offline-archives-of-my-photos-mail-and-documents-from-the-platforms-I-use-kept-in-open-f/report.md
Manifest written to /home/user/.local/share/point53/courier/runs/2026-06-07T18-22-57_Personal-data-takeout-and-archive-I-want-complete-offline-archives-of-my-photos-mail-and-documents-from-the-platforms-I-use-kept-in-open-f/manifest.json
Chat option selected, your analysis is already saved (but be sure to save these chat messages elsewhere)...
    Ctrl+D to continue to next stage.
> 
§ keep in mind

Before you install.

  • browser Firefox or Chrome installed as non-default web browser is needed. Webdriver is downloaded automatically on first run.
  • configuration We recommend looking at the configuration (default per-OS locations, like ~/.config/point53/courier) and adjusting where queries are dropped according to your preference. Hint: Check the URL bar when querying wherever you fancy.
  • captchas Do them yourself. We will never try to bypass human checks. Scraping fails silently in headless mode.
  • ai providers Ollama by default; 10 GB of VRAM or unified memory is all you need. Every query and summary runs on-device. Prefer the cloud? Add [anthropic] during install, set ANTHROPIC_API_KEY, and edit via config before first run.
design note · 01

Selenium underneath.

Real browser. Real DOM. Real network. The honest tax for not lying to sites that don't want bots.

design note · 02

History is local.

You have resolvable links and structured analysis of them. Want to revisit a query or analysis result? Open up the shared markdown.

design note · 03

CAPTCHAs escalate.

We do not solve them. We do not pretend. The session pauses; you take the wheel or skip the countdown.