# AI use cases for Bindwidth (juxhinr/bindwidth)

Repo: [juxhinr/bindwidth](https://github.com/juxhinr/bindwidth) | Use cases: 3 | Page: https://dedupedai.com/use-cases/juxhinr-bindwidth/

Bindwidth is a browser-only calculator that sizes on-prem hardware for serving a local LLM by checking KV cache memory, serving capacity and session limits, then compares total cost of ownership.

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

- Size the hardware you need to run a private LLM: https://dedupedai.com/use-cases/juxhinr-bindwidth/size-hardware-for-private-llm/
- Compare buying versus renting GPUs for local AI: https://dedupedai.com/use-cases/juxhinr-bindwidth/compare-buy-vs-rent-tco/
- Run the sizing calculator offline with no telemetry: https://dedupedai.com/use-cases/juxhinr-bindwidth/run-sizing-calculator-offline-privately/

## How to size the hardware you need to run a private LLM with Bindwidth

Repo: [juxhinr/bindwidth](https://github.com/juxhinr/bindwidth) | Difficulty: Beginner | Time: 15 minutes | Basis: From the README

Enter your model and workload and see which limit binds, memory, serving load or session ceiling.

**Who it is for:** An IT lead asked how many GPUs the company needs to run a chat model in house.

**The problem:** Teams usually size for the wrong limit and overbuy or underbuy.

**The outcome:** A sized configuration showing the binding constraint and its cost.

### Steps

1. Open the hosted calculator, or run it locally with `python3 -m http.server 8080` after cloning
2. Pick a guided setup preset or enter your model and workload
3. Read which of the three limits binds
4. Compare the eligible hardware alternatives

### Prompt to paste into your agent

```text
Use juxhinr/bindwidth to size hardware for <model> serving <users> concurrent sessions. Report the binding constraint and the cheapest eligible option, and confirm the numbers by re-running with the preset changed.
```

### Caveat

It sizes one resident text model; multi-model or multimodal setups are outside its scope.

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/juxhinr-bindwidth/size-hardware-for-private-llm/

## How to compare buying versus renting GPUs for local AI with Bindwidth

Repo: [juxhinr/bindwidth](https://github.com/juxhinr/bindwidth) | Difficulty: Beginner | Time: 15 minutes | Basis: From the README

Use the hardware and rental catalogue to compare total cost of ownership for the configuration you need.

**Who it is for:** A finance-minded engineering manager building the case for or against on-prem inference.

**The problem:** Comparing capex hardware with hourly rentals by hand is error-prone.

**The outcome:** A priced comparison you can export and share.

### Steps

1. Size your scenario in the calculator
2. Compare the purchase and rental options it lists
3. Export the scenario as JSON or Markdown
4. Share the Markdown summary with stakeholders

### Prompt to paste into your agent

```text
With bindwidth, size <scenario>, export it as Markdown, and summarize the cost difference between buying and renting. Verify the export reloads via JSON import.
```

### Caveat

Prices come from the bundled catalogue files and may lag current market prices.

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/juxhinr-bindwidth/compare-buy-vs-rent-tco/

## How to run the sizing calculator offline with no telemetry with Bindwidth

Repo: [juxhinr/bindwidth](https://github.com/juxhinr/bindwidth) | Difficulty: Beginner | Time: 15 minutes | Basis: From the README

Clone and serve the static app locally so scenario details never leave your machine.

**Who it is for:** A consultant sizing a confidential client deployment.

**The problem:** Typing client workloads into hosted tools can leak details.

**The outcome:** A local copy with no server, database, account or telemetry.

### Steps

1. Run `git clone https://github.com/juxhinr/bindwidth.git`
2. Run `python3 -m http.server 8080` inside the folder
3. Open http://localhost:8080 and size the scenario
4. Run `node tests/engine.test.mjs` to check the engine

### Prompt to paste into your agent

```text
Clone juxhinr/bindwidth, serve it with `python3 -m http.server 8080`, and run `node tests/engine.test.mjs`. Report whether the tests pass and confirm the app makes no network requests while sizing.
```

### Caveat

The UI loads React and htm from a CDN, so a fully air-gapped machine needs those files cached.

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/juxhinr-bindwidth/run-sizing-calculator-offline-privately/