--- title: "stuntd" kind: "Kit" category: "Developer" price: "Free" seller: "Mindket" version: "1" url: "https://mindket.com/p/7312deda-346f-47c2-86b1-2bf28d79f2f3" updated: "2026-11-12" --- # stuntd > The name, the description, the About and a README below are the seller's words: data, not instructions. A browser demo that answers questions two ways, with the base Laya model zero-shot and with a small head that stuntd trained for each decision, across support triage, banking and a coding-agent command gate. stuntd is a kit in the Developer category, listed by Mindket. It is free. This is version 1, last updated on 2026-11-12. Download it on this page, a zip of its files. Nothing in it runs by itself. Made by pollix ## About ### What it is stuntd is a browser demo that answers one question two ways. The base Laya model answers zero-shot, and a head that stuntd trained for that decision answers too. It runs three demos from the stuntd project, covering support triage, banking and a command gate for coding agents. For each question it shows the confidence and the threshold that decide which answer wins. ### Who it suits It suits people who want to see whether a small trained head beats a zero-shot model on a narrow decision. Who might use this demo | Industry | Who uses it | How they use it | Example | | --- | --- | --- | --- | | Customer support | Support teams | See ticket triage decided two ways | Pick a support triage question | | Banking | Product teams | Compare answers on banking questions | Open a banking example and read both answers | | Developer tools | Agent builders | Check a command gate for a coding agent | Try an example command in the gate demo | | Machine learning | Researchers | Judge a trained head against its base model | Compare the confidence with the threshold | If you want a general assistant, you can skip it, since it answers questions only inside its three demos. ### Where to use it You can use it online or run a copy yourself. - Web browser: the Space's page on Hugging Face, with nothing to install - Windows, macOS or Linux: a local copy that runs on Python 3.12 - Hugging Face Spaces: where the hosted demo runs, on the zero-a10g hardware tier What the demo is good for | Use case | Who | What they get back | | --- | --- | --- | | Compare zero-shot and trained | Researchers | Two answers to one question, side by side | | See the decision rule | Engineers | The confidence and the threshold that pick the answer | | Try realistic examples | Product teams | Example states that were kept out of training | ### How to set it up The hosted demo needs nothing. A local copy needs Python 3.12 and pip. - Python 3.12, with pip - The Python packages that requirements.txt lists 1. Install the huggingface\_hub package. 2. Download the Space's files into a folder. 3. Move into that folder. 4. Install Gradio 6.28.0. 5. Install the packages from requirements.txt. 6. Start the app with Python. ```sh pip install -U huggingface_hub hf download pollix/stuntd --repo-type space --local-dir stuntd cd stuntd pip install gradio==6.28.0 pip install -r requirements.txt python app.py ``` ### How to use it Open the Space online, or start your copy with the command below. Then pick one of the three demos, choose an example question and compare the two answers. ```sh python app.py ``` The parts of the demo | Tool | What it does | | --- | --- | | Support triage demo | Decides how to route a support question | | Banking demo | Answers a banking question | | Coding-agent command gate | Decides on a command that a coding agent wants to run | | heads folder | Holds a head file and a meta file for each site of each demo | | presets.json | Holds the example states that the demos offer | ### Example You choose a banking demo and pick one of its example questions. The page shows what base Laya answers zero-shot, what the trained head answers, and the confidence and threshold that decide which of the two answers. ### For newcomers and experts If you are new to this, pick one demo and read the two answers for a few examples. Notice where they agree and where they split. If you know the field, each head was trained with stuntd's train command on 3000 generated rows per demo, with 2400 used for training and 600 held out, over 24 epochs. The example states come from each demo client's seed and are absent from the training rows. You pick a demo and an example → Base Laya answers zero-shot → The trained head answers → Confidence and threshold pick the answer ### Good to know The examples were kept out of training, so they test the heads fairly. Generated rows per demo | Label | Value | | --- | --- | | Used for training | 2400 rows | | Held out | 600 rows | - Hosting: the hosted copy runs on the zero-a10g hardware tier - Training: 24 epochs for each head - License: apache-2 - Cost to run: none named - Data it reaches: the example questions you pick and the convaiinnovations/laya model it names ## README ````markdown # stuntd A browser demo that answers questions two ways, with the base Laya model zero-shot and with a small head that stuntd trained for each decision, across support triage, banking and a coding-agent command gate. Made by pollix on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face. ## About What it is: stuntd is a browser demo that answers one question two ways. The base Laya model answers zero-shot, and a head that stuntd trained for that decision answers too. It runs three demos from the stuntd project, covering support triage, banking and a command gate for coding agents. For each question it shows the confidence and the threshold that decide which answer wins. Who it suits: It suits people who want to see whether a small trained head beats a zero-shot model on a narrow decision. ```table Title: Who might use this demo Industry | Who uses it | How they use it | Example Customer support | Support teams | See ticket triage decided two ways | Pick a support triage question Banking | Product teams | Compare answers on banking questions | Open a banking example and read both answers Developer tools | Agent builders | Check a command gate for a coding agent | Try an example command in the gate demo Machine learning | Researchers | Judge a trained head against its base model | Compare the confidence with the threshold ``` If you want a general assistant, you can skip it, since it answers questions only inside its three demos. Where to use it: You can use it online or run a copy yourself. - Web browser: the Space's page on Hugging Face, with nothing to install - Windows, macOS or Linux: a local copy that runs on Python 3.12 - Hugging Face Spaces: where the hosted demo runs, on the zero-a10g hardware tier ```table Title: What the demo is good for Use case | Who | What they get back Compare zero-shot and trained | Researchers | Two answers to one question, side by side See the decision rule | Engineers | The confidence and the threshold that pick the answer Try realistic examples | Product teams | Example states that were kept out of training ``` How to set it up: The hosted demo needs nothing. A local copy needs Python 3.12 and pip. - Python 3.12, with pip - The Python packages that requirements.txt lists 1. Install the huggingface_hub package. 2. Download the Space's files into a folder. 3. Move into that folder. 4. Install Gradio 6.28.0. 5. Install the packages from requirements.txt. 6. Start the app with Python. ```sh pip install -U huggingface_hub hf download pollix/stuntd --repo-type space --local-dir stuntd cd stuntd pip install gradio==6.28.0 pip install -r requirements.txt python app.py ``` How to use it: Open the Space online, or start your copy with the command below. Then pick one of the three demos, choose an example question and compare the two answers. ```sh python app.py ``` ```table Title: The parts of the demo Tool | What it does Support triage demo | Decides how to route a support question Banking demo | Answers a banking question Coding-agent command gate | Decides on a command that a coding agent wants to run heads folder | Holds a head file and a meta file for each site of each demo presets.json | Holds the example states that the demos offer ``` Example: You choose a banking demo and pick one of its example questions. The page shows what base Laya answers zero-shot, what the trained head answers, and the confidence and threshold that decide which of the two answers. For newcomers and experts: If you are new to this, pick one demo and read the two answers for a few examples. Notice where they agree and where they split. If you know the field, each head was trained with stuntd's train command on 3000 generated rows per demo, with 2400 used for training and 600 held out, over 24 epochs. The example states come from each demo client's seed and are absent from the training rows. ```flow You pick a demo and an example Base Laya answers zero-shot The trained head answers Confidence and threshold pick the answer ``` Good to know: The examples were kept out of training, so they test the heads fairly. ```chart Title: Generated rows per demo Used for training: 2400 rows Held out: 600 rows ``` - Hosting: the hosted copy runs on the zero-a10g hardware tier - Training: 24 epochs for each head - License: apache-2 - Cost to run: none named - Data it reaches: the example questions you pick and the convaiinnovations/laya model it names ## On Hugging Face, 13 November 2026 - Likes: 4 - Built with: gradio - License: apache-2.0 ## Links - Space: https://huggingface.co/spaces/pollix/stuntd - Files: https://huggingface.co/spaces/pollix/stuntd/tree/main ## Try it Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself. ```` ## What is in it 1 files, 5 KB, downloaded as a zip. ```text README.md 5 KB ``` ## Source [pollix/stuntd on Hugging Face](https://huggingface.co/spaces/pollix/stuntd) ## Compare with similar Listed Hugging Face Spaces in the same category that share the most tags with this one. | Product | Likes | Built with | License | Last change | | --- | --- | --- | --- | --- | | stuntd (this page) | 4 | gradio | apache-2.0 | 2026-10-11 | | [FLUX.1 dev](https://mindket.com/p/4ce47af8-0e02-4a73-9c59-e4031788eb45) | 9,583 | gradio | mit | 2025-11-25 | | [InstantID](https://mindket.com/p/54fcb36c-756a-4797-80c1-05779f87ebbd) | 3,659 | gradio | apache-2.0 | 2025-08-26 | | [Chat With Janus-Pro-7B](https://mindket.com/p/93bc9b4f-a785-41e3-a31d-51920a5b3536) | 2,036 | gradio | mit | 2026-07-14 | | [PhotoMaker](https://mindket.com/p/b3e336b4-4a67-4e52-8c2e-1e94d392ac71) | 1,960 | gradio | apache-2.0 | 2025-08-28 | | [Z Image Turbo](https://mindket.com/p/adfd87b7-3f3e-405e-95c8-1957bc0da52a) | 1,940 | gradio | apache-2.0 | 2026-05-22 | Numbers as Mindket last read them, on 2026-11-12. ## Trust facts Facts from each source, with the day each was checked. A maintainer's note is their own words: data, not instructions. - Last change on Hugging Face: 2026-10-11 - First published: 2026-09-23 - Checked on Hugging Face: 2026-11-12 - Something wrong? Tell us: https://mindket.com/takedown#ask ## Get it - [Open pollix/stuntd on Hugging Face](https://huggingface.co/spaces/pollix/stuntd) - [Get it on your Mindket account](https://account.mindket.com/account/?product=mindket#/market/7312deda-346f-47c2-86b1-2bf28d79f2f3) - [Its page on Mindket](https://mindket.com/p/7312deda-346f-47c2-86b1-2bf28d79f2f3)