Automated README summary
Automatically summarized from the README on 18 Oct 2026; not a run by jevbooks. Run instructions and author-reported outcomes still need verification.
What it is
A practical demo showing how to add a fast safety check before an AI agent acts. It screens messages for a simplified banking assistant before they reach its Gemini model or tools.
How it uses Jev
A custom LangChain middleware runs a focused Jev classification in the agent's before_agent hook. It evaluates the user's message against explicit banking policies. Jev's unsafe score is converted into a safety score; the request passes only if it meets the 0.65 threshold, otherwise the agent is stopped with a refusal.
Primitives:score
Technique worth stealing
Use a dedicated fast classifier as an input gate before the main agent model and tools run.
Try it
Clone, run uv sync, add API keys to .env, then run uv run --with jupyterlab jupyter lab main.ipynb.
Model judgments from README excerpts
Project signals use repository metadata, an existing summary and the first 700 README characters, not the full project. “Not found” means not detected in that material; it does not establish absence. These are the original judgments with recorded models and dates, not source audits or run results.
Models & evaluation dates
- Inclusion gate
jev-1.13.0- Project signals
jev-1.13.0- Category
jev-1.13.0- Pattern tags
jev-1.13.0
Evidence
These signals use repository metadata, an existing summary and at most 700 cleaned README characters. ≥ 0.60 means reported in that excerpt; ≤ 0.40 means not found there. The middle band is inconclusive. None is a source-code audit or a test result. The runnable-tool question excludes research experiments; the recommendation question asks for a learning shortlist, not a quality verdict.
- Jev is core to the projectFound in material0.82
- Shows a System One patternnot found in the excerpt0.23
- Describes handling uncertaintynot found in the excerpt0.08
- Reports measured numbersnot found in the excerpt0.06
- Is a runnable toolFound in material0.72
- Top-10 learning shortlist signalunclear0.43
- Is a replica of the modelnot found in the excerpt0.04
- Solves a common problem (0–2)score on a 0–2 scale1.53 / 2
- About JevFound in material0.97
no-judgment band (0.40–0.60)
no-judgment band (0.40–0.60)
The Jev model and date for each judgment stage are listed above. Card written by DeepSeek V4.1 Flash from the README on 18 Oct 2026.