Skip to content

Show the status as a note and fold reference sections #17

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
rustfuture merged 1 commit into from
Oct 2, 2026
Merged

Show the status as a note and fold reference sections #17

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 14 additions & 1 deletion README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,7 +7,8 @@ Reflex Control decides whether an AI task can finish on its own, retry, or escal
[![CI](https://github.com/rustfuture/reflex-control/actions/workflows/ci.yml/badge.svg)](https://github.com/rustfuture/reflex-control/actions/workflows/ci.yml)
[![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)

**Status:** Research prototype (v0.4.0). Tested on synthetic fixtures; not production-validated.
> [!NOTE]
> **Status:** Research prototype (v0.4.0). Tested on synthetic fixtures; not production-validated.

- Runs fast local checks (exit codes, test results, git diffs) alongside semantic risk signals.
- Blocks high-risk actions like credential leaks or schema changes with a mandatory safety veto.
Expand All @@ -30,6 +31,8 @@ reflex demo verifier-gate

The demo prints local policy decisions without an API key or model download. Prebuilt Linux binaries require glibc; unsupported platforms can build from source. On Windows, use the [PowerShell installer](install.ps1).

<details><summary>Build from source</summary>

### Build from source

You need Git and Rust **1.88** or newer, including Cargo. Install the toolchain with [rustup](https://rustup.rs/); the first source build downloads Cargo dependencies.
Expand Down Expand Up @@ -79,6 +82,8 @@ cargo run --bin reflex -- run --provider jev \
--risk low
```

</details>

---

## How it works
Expand Down Expand Up @@ -113,6 +118,8 @@ The repository is organized as a Cargo workspace separating core abstractions fr

The repository includes standalone runnable examples under [`examples/`](examples/):

<details><summary>All examples</summary>

### Verifier Gate Example

Demonstrates both autonomous acceptance of safe tasks and hard safety veto rules against dangerous modifications (even when model confidence is high):
Expand Down Expand Up @@ -177,6 +184,8 @@ verification failing, budget exhausted -> escalate (reflex action: defer_to_fr
green tests, secret file touched -> escalate (reflex action: defer_to_frontier)
```

</details>

---

<a id="evaluation-benchmark--measured-results"></a>
Expand All @@ -192,6 +201,8 @@ Historical Candidate E aggregates, partition history, and metric definitions are

All fixture datasets and frozen policies reside in [`fixtures/`](fixtures/):

<details><summary>Dataset details</summary>

| Dataset File | Role | Sample Count |
| :--- | :--- | :---: |
| `fixtures/v2_eval_blind_test.json` | Curated synthetic held-out evaluation partition (legacy filename) | 100 tasks |
Expand All @@ -200,6 +211,8 @@ All fixture datasets and frozen policies reside in [`fixtures/`](fixtures/):
| `fixtures/v2_eval_dev.json` | Development & connectivity smoke testing split | 30 tasks |
| `fixtures/frozen_hybrid_config.json` | Frozen 0.3.0 policy parameters ($\tau_{\text{accept}} = 0.22, \theta_{\text{clean}} = 0.40, \text{risk}_{\text{frontier}} = 0.70$); fit with live Jev on the calibration partition | Config |

</details>

---

## Scope and Limitations
Expand Down
Loading