---

license: apache-2.0

base_model: Qwen/Qwen2.5-Coder-3B-Instruct

library_name: gguf

pipeline_tag: text-generation

language:

- en

tags:

- gguf

- exe-ai-terminal

- extended-workflow

- guardian

- tool-use

- llama.cpp

---

<div align="center">

<picture>

<source media="(prefers-color-scheme: dark)" srcset="https://huggingface.co/exeterminal/Exe-Guard-Dynamic-GGUF/resolve/main/exe-chevron-dark.png">

<img src="https://huggingface.co/exeterminal/Exe-Guard-Dynamic-GGUF/resolve/main/exe-chevron-light.png" alt="Exe AI" width="72">

</picture>

<h1>Exe AI</h1>

<a href="https://exe-hq.net"><b>exe-hq.net</b></a>

</div>

---

<div align="center">

<img src="exe-guard-hero.png" alt="Exe Guard Dynamic — Extended Workflow guardian for the Exe AI Terminal" width="100%">

</div>

Exe Guard Dynamic

A tiny guardian model for the Extended Workflow feature of the Exe AI Terminal

Website to the Main AI Agent Harness: https://exe-hq.net

When a tool step fails, it reads the report of that failure and writes the **one

instruction** the user should send next to fix it — a corrected command, the real

file, a path inside the shared folder. It suggests; it never acts.

!Per-case results — base 42% vs. trained 100% on 24 tool-repair cases

What it does

The guardian watches tool calls and speaks up only when a step failed in a way a

person would want a suggestion for. It is built for exactly six kinds of failure:

| # | Failure | The fix it should name |

|---|---|---|

| 1 | Command typo (npm run buld) | the correctly spelled command |

| 2 | Wrong Python environment (a .venv/ exists) | .venv/bin/python3 … (never a global pip install, never source activate) |

| 3 | edit_file old-text not found | read the file first, then edit with the exact text |

| 4 | Binary / unreadable file | the readable file (e.g. the .log), or run_command for archives |

| 5 | Path rejected (outside the shared folders) | a path inside the released folder |

| 6 | Run stopped after a timeout | re-run in the background |

It answers in one imperative English sentence, no greeting, no explanation.

Intended use

Drop-in as the small background model behind the Exe AI Terminal's Extended Workflow.

It is a specialist: it turns a failed-step report into a single corrective instruction.

Out of scope: general chat, code generation, vision, or any use outside the

failed-step-repair task. It is not a general assistant.

Files

All builds carry an importance matrix (imatrix) computed from the model's own task

data, and were tested on 24 held-out repair cases at temperature 0.1. "Test" is the

number of those 24 cases solved correctly — a task metric, not perplexity.

| File | Type | Bits | Size | Test (of 24) |

|---|---|---|---|---|

| Exe-Guard-Dynamic-Q8_0.gguf | K/legacy | 8 | 3.06 GB | 24 / 24 |

| Exe-Guard-Dynamic-Q6_K.gguf | K-quant | 6.5 | 2.36 GB | 24 / 24 |

| Exe-Guard-Dynamic-Q5_K_M.gguf | K-quant | 5.5 | 2.07 GB | 24 / 24 |

| Exe-Guard-Dynamic-Q4_K_M.gguf | K-quant | 4.8 | 1.80 GB | 24 / 24 |

| Exe-Guard-Dynamic-Q4_K_S.gguf | K-quant | 4.5 | 1.71 GB | 24 / 24 |

| Exe-Guard-Dynamic-IQ4_XS.gguf | I-quant · recommended | 4.25 | 1.62 GB | 24 / 24 |

| Exe-Guard-Dynamic-Q3_K_L.gguf | K-quant | 4.0 | 1.59 GB | 19 / 24 |

| Exe-Guard-Dynamic-Q3_K_M.gguf | K-quant | 3.9 | 1.48 GB | 20 / 24 |

| Exe-Guard-Dynamic-IQ3_M.gguf | I-quant | 3.66 | 1.39 GB | 24 / 24 |

| Exe-Guard-Dynamic-IQ3_S.gguf | I-quant | 3.44 | 1.36 GB | 24 / 24 |

| Exe-Guard-Dynamic-Q2_K.gguf | K-quant | 3.0 | 1.19 GB | 24 / 24 |

| Exe-Guard-Dynamic-IQ2_M.gguf | I-quant | 2.7 | 1.06 GB | 23 / 24 |

| Exe-Guard-Dynamic-IQ2_S.gguf | I-quant | 2.5 | 0.99 GB | 22 / 24 |

| Exe-Guard-Dynamic-IQ2_XS.gguf | I-quant | 2.06 | 0.96 GB | 22 / 24 |

| Exe-Guard-Dynamic-IQ1_M.gguf | I-quant · experimental | 1.75 | 0.79 GB | 9 / 24 |

| Exe-Guard-Dynamic-IQ1_S.gguf | I-quant · experimental | 1.56 | 0.74 GB | 10 / 24 |

| Exe-Guard-Dynamic-f16.gguf | full precision | 16 | 5.75 GB | 24 / 24 |

IQ4_XS is the recommended build. It is smaller than Q4_K_M and solves the

same 24 of 24 — every build here carries an importance matrix, and the I-quants use

it to spend their bits where the model actually needs them.

The same pattern holds further down: at low bit-widths the I-quants (IQ3, IQ2) beat

the K-quants of similar size (Q3_K drops to 19–20/24). The 1-bit builds are

included for the curious but are not recommended.

Prompt and sampling

Two messages only — a fixed system instruction and the failure report — with

temperature 0.1, max_tokens 200, thinking off, context 4096. No conversation

history.

Base model and license

• Base: Qwen/Qwen2.5-Coder-3B-Instruct
• License: Apache-2.0 (base and this derivative). You may use, modify, rebrand and

redistribute; the origin of the base model must be named — it is, here.

Training

A thin LoRA adapter on top of the base, trained locally (Apple Silicon, MLX) on

synthetic examples of failed tool steps and their one-sentence corrections, built to

match the exact report format the terminal produces. The adapter was fused into the

base and then quantized.

Evaluation

On 24 held-out repair cases at temperature 0.1, the untrained base solves **10 / 24

(42%) and Exe Guard Dynamic solves 24 / 24 (100%)**. The base fails completely on

the two cases that need this training — wrong Python environment and wrong edit target

(0%) — which the trained model fixes entirely.

Transparency

This is a fine-tuned derivative of an openly licensed base model, released with its

provenance, intended use, limits and evaluation stated above, in line with

transparency expectations for shared models (incl. the EU AI Act).