Founders intro video — link from README + BLOG + env Space landing + model card

Notebook: strip malformed metadata.widgets so GitHub renders it again

Created using Colab

Clean root folder: 11 → 6 Python files, BLOG.md + README.md prominent

README: comprehensive overhaul for AI evaluation + judge first-impression

BLOG.md: hero-arc story version

Bump gradio_ui Space to gradio 5.49.0 + sdk_version 5.49.0

Add Gradio demo Space link to model card; pushed to HF Hub

Deploy Gradio demo Space + link from README + env Space landing

Move blog draft to BLOG.md in repo root for easy access

Add friendly landing page at / so judges don't see {'detail':'Not Found'}

README: surface all 5 charts in a 'TL;DR — story in 4 charts' section right under the headline numbers

Add journey/comparison/demo charts + push real model card to HF Hub

Draft mini-blog for HuggingFace publication

Add Gradio UI demo for video + standalone HF Space

Notebook eval cell: replace force-report stub with full v3 evidence-driven harness

Real GRPO results: Llama 3.2 3B avg 0.075 -> 0.482 (6.4x) on VAPT-Env

Add v3 hybrid eval with auto-submit on positive evidence

Add colab_eval_hybrid.py — hybrid eval script (scripted recon prefix + trained model + anti-collapse safety net) so users can wget it directly into Colab without paste-mangling

Notebook baseline: force generate_report so the grader's final_score is what we save

Notebook install: pin numpy/pandas/fsspec + auto-restart kernel after install

Notebook install: add openenv-core dep + drop --no-deps so security_audit_env's dependencies come along

Notebook: switch back to canonical openenv-core client (proper WebSocket session)

Rebuild notebook end-to-end: self-contained, no fragile externals

Notebook: force generate_report after vuln-budget + correct W&B columns

Notebook: auto-detect bf16/fp16 (T4 pre-Ampere doesn't support bf16)

Notebook: graceful fallback when Colab Secrets blocks WANDB_API_KEY

Real GRPO+TRL+Unsloth Colab notebook + pyproject subpackage fix

Polish round: curated multi-agent demo + Llama 3.2 3B baseline + video script

Phase 8: prompt rebalance — single-agent flow primary, delegation as optional path

Phases 4-7: nested inference prompt, delegation reward, Delegation Score grader, multi-agent README

Phase 3: server-side sub-agent infrastructure (spawn / budget / return)

Multi-agent foundations (Phases 1+2): dynamic attack surface + revealed_targets

new

Round 2 polish: defensible README claims, small-model-friendly inference prompt, brand consistency

Clean cruft files, fix README links after docs/ reorg, integrate Cursor's inference enhancements

new

Configure git-lfs to track *.png files

Merge branch 'main' of https://huggingface.co/spaces/Sayuj63/Vapt-env

Add root Dockerfile and update Space URLs for HF deployment

initial commit

first commit