# AI use cases for Start Lux Decision (StartLuxLabs/StartLux-Decision)

Repo: [StartLuxLabs/StartLux-Decision](https://github.com/StartLuxLabs/StartLux-Decision) | Use cases: 3 | Page: https://dedupedai.com/use-cases/startluxlabs-startlux-decision/

StartLux-Decision is an open inference engine and HTTP server for a family of decision models that return a calibrated probability for every option of typed questions.

This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL.

- Route support tickets to teams with typed decision questions: https://dedupedai.com/use-cases/startluxlabs-startlux-decision/triage-support-tickets-locally/
- Test decision requests on the hosted 27B model without a GPU: https://dedupedai.com/use-cases/startluxlabs-startlux-decision/try-decisions-without-a-gpu/
- Send decisions with no fitting option to a person: https://dedupedai.com/use-cases/startluxlabs-startlux-decision/flag-unsure-decisions-for-review/

## How to route support tickets to teams with typed decision questions with Start Lux Decision

Repo: [StartLuxLabs/StartLux-Decision](https://github.com/StartLuxLabs/StartLux-Decision) | Difficulty: Intermediate | Time: 90 minutes | Basis: From the README

Run a local decision server and get a team, urgency and severity answer for each ticket.

**Who it is for:** Support or operations lead with a queue of incoming tickets and GPUs they can dedicate to a decision server.

**The problem:** Tickets arrive as free text and someone has to decide which team handles each one, how urgent it is and how severe the impact is.

**The outcome:** Each ticket gets a probability for every team option, a yes or no urgency probability and an expected severity level, returned from one local endpoint.

### Steps

1. Run `git clone https://github.com/StartLuxLabs/StartLux-Decision.git` and then `cd StartLux-Decision`.
2. Run `pip install -e '.[vllm]'` in the same Python environment as vLLM.
3. Run `hf download startlux-models/StartLux-Decision-v1.1-4B-AutoThink --local-dir models/StartLux-Decision-v1.1-4B-AutoThink`.
4. Run `startlux-decide serve --model models/StartLux-Decision-v1.1-4B-AutoThink --launch-vllm --gpus 0,1,2,3 --port 8090`, changing the GPU list to match your machine.
5. Send each ticket as a POST to /v1/systemone on port 8090 using the curl request from the Quick start, with the ticket text in `state` and your own `questions`.

### Prompt to paste into your agent

```text
Using the StartLux-Decision repo, start the decision server with the 4B AutoThink model and send the ticket text <ticket_text> to POST /v1/systemone with a choice question named team, a noul question named urgent and a score question named severity. Return the answers JSON. Verify that the team answer is one of the listed criteria keys.
```

### Caveat

The README's local setup assumes vLLM with GPUs and the 4B model download, so the first run depends on that hardware and network access.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/startluxlabs-startlux-decision/triage-support-tickets-locally/

## How to test decision requests on the hosted 27B model without a GPU with Start Lux Decision

Repo: [StartLuxLabs/StartLux-Decision](https://github.com/StartLuxLabs/StartLux-Decision) | Difficulty: Intermediate | Time: 15 minutes | Basis: From the README

Send decision requests to the hosted 27B AutoThink model with a Bamboo API key.

**Who it is for:** Developer or analyst who wants to evaluate the decision format before setting up local GPUs.

**The problem:** Setting up vLLM and downloading a model takes time and hardware that a first test may not justify.

**The outcome:** A working decision request and response from the hosted 27B AutoThink model, using the same request format as the local server.

### Steps

1. Sign in at bamboonode.dev and create a Bamboo API key. (inferred, check the README)
2. Send a POST request to the decisions endpoint, /v1/decisions, with your key and a JSON body containing `state` and `questions`, following the format shown in the Quick start. (inferred, check the README)
3. Read the `answers` object and each question's probabilities in the response.
4. Adjust your `questions` and repeat until the answers match what you expect for your own examples.

### Prompt to paste into your agent

```text
Using the StartLux-Decision README, build a POST request to /v1/decisions on bamboonode.dev with the state <state_text> and one choice question with the criteria <criteria_list>. Return the response answers and the model name. Verify that the model name in the response is the 27B AutoThink model.
```

### Caveat

The README says the hosted service is free without usage limits, but the key steps and endpoint details in this list are inferred and should be checked on bamboonode.dev.

Risk, credentials: The hosted call needs a Bamboo API key, so keep it in an environment variable and do not paste it into prompts or shared logs.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/startluxlabs-startlux-decision/try-decisions-without-a-gpu/

## How to send decisions with no fitting option to a person with Start Lux Decision

Repo: [StartLuxLabs/StartLux-Decision](https://github.com/StartLuxLabs/StartLux-Decision) | Difficulty: Intermediate | Time: 60 minutes | Basis: From the README

Use none_probability to route questions where no listed option seems right.

**Who it is for:** Operations reviewer who handles decisions with fixed option lists and needs to catch cases where none of the options fit.

**The problem:** A choice question can return a confident-looking answer even when none of the listed options is correct.

**The outcome:** Questions whose none_probability is 0.2 or higher go to a human reviewer, and the rest are handled automatically.

### Steps

1. Start a decision server as described in the triage use case, or use the hosted model.
2. Add `"none_check": true` and `"think_mode": "auto"` to the request body, and keep `"gate": 0.6` unless you need more reasoning. (inferred, check the README)
3. Read `none_probability` on each choice answer.
4. Route any answer with `none_probability` of 0.2 or more to review, and handle the rest automatically.

### Prompt to paste into your agent

```text
Using the StartLux-Decision repo, send <question_text> with the listed <criteria> to POST /v1/systemone with none_check set to true. Return the choice, its confidence and none_probability. Verify that any answer with none_probability of 0.2 or more is marked for human review.
```

### Caveat

The README gives 0.2 as a suggested threshold from its own evaluation, so you should check it against your own cases before relying on it.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/startluxlabs-startlux-decision/flag-unsure-decisions-for-review/