Skip to content

Releases: pablocaeg/sloptotal

Release list

v1.1.0 — Site check, uploads, startup fixes

v1.1.0 — Site check, uploads, startup fixes Latest
Latest

Choose a tag to compare

View all tags
@pablocaeg pablocaeg released this 28 Sep 10:15
v1.1.0
3163058

Green CI, correct scores from the first request, and two new features.

New

  • Site check: find out whether a website was built with Lovable, v0, Bolt, Base44, Replit or Same. It uses only markers the builders leave in deployed sites, each verified on live deployments, and it shows the evidence rather than a made-up percentage. Available as a home-page tab, a card on URL reports, and POST /api/scan/site.
  • Upload .pdf, .docx, .txt or .md on the Text tab, or through POST /api/extract.
  • scripts/smoke_test.py: an end-to-end check of every route and all 23 engines.

Fixed

  • Scans that arrived while models were loading could come back wrong: four classifiers scored 0.0, or GPT-2 loaded with a random head and pinned Binoculars at 1.0.
  • On a fresh Docker volume, the first scan timed out while models were still downloading. It now waits for them.
  • URL scans could reach private, loopback or cloud-metadata addresses.
  • Report colours disagreed with the verdict text.

Measured

  • Re-measured on a fresh RAID sample (180 texts) with upgraded dependencies: AUC 0.979, 1 of 40 human texts called "Likely AI", 0 of 26 literary passages flagged.
  • Seven newer open detectors benchmarked; Gradient (DeBERTa-v3-large) is next in line as an engine. See FINDINGS.md.

Run it

docker run -p 8000:8000 -v sloptotal-models:/app/models ghcr.io/pablocaeg/sloptotal:1.1.0

Full changes: CHANGELOG · #1

Assets 2
Loading

v1.0.1 — Setup fixes & docs

v1.0.1 — Setup fixes & docs

Choose a tag to compare

View all tags
@pablocaeg pablocaeg released this 03 Jul 09:15
v1.0.1
7ec2b86

Summary

Fixes neural engine loading on modern HuggingFace models, improves first-run setup docs, and adds a Qwen/Gemma engine roadmap.

What's fixed

  • Model loading — bump transformers>=4.46 and tokenizers>=0.21 (older tokenizers broke ReMoDetect, Fakespot, TMR, Desklib, SuperAnnotate)
  • Missing dependency — add beautifulsoup4 for URL scraping
  • Stale cache — auto-purge and skip cached reports that contain engine load failures

Setup (new users)

git clone https://github.com/pablocaeg/sloptotal.git
cd sloptotal
python3.11 -m venv venv && source venv/bin/activate
pip install --upgrade pip && pip install -r requirements.txt
./start.sh

Requirements: Python 3.10+, 4 GB RAM minimum, ~2 GB disk for models. No GPU required.

High-RAM CPU servers (64 GB): see README for performance profile tuning.

Links

Loading