# AI use cases for LLM Visuals (DingoOz/llm-visuals)

Repo: [DingoOz/llm-visuals](https://github.com/DingoOz/llm-visuals) | Use cases: 4 | Page: https://dedupedai.com/use-cases/dingooz-llm-visuals/

A Rust terminal dashboard that reads a locally running LLM server and GPU drivers and shows live throughput, latency, GPU load, context fill and expert routing.

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

- Watch tokens per second and TTFT on a local server: https://dedupedai.com/use-cases/dingooz-llm-visuals/watch-local-llama-server-live/
- Preview every dashboard view with no GPU or model: https://dedupedai.com/use-cases/dingooz-llm-visuals/preview-dashboard-views-without-hardware/
- Record every request to a SQLite log for analysis: https://dedupedai.com/use-cases/dingooz-llm-visuals/log-llm-requests-to-sqlite/
- Build release binaries for six OS and CPU targets: https://dedupedai.com/use-cases/dingooz-llm-visuals/build-release-binaries-for-six-targets/

## How to watch tokens per second and TTFT on a local server with LLM Visuals

Repo: [DingoOz/llm-visuals](https://github.com/DingoOz/llm-visuals) | Difficulty: Intermediate | Time: 45 minutes | Basis: From the README

Attach the dashboard to a running llama-server and see tok/s, time to first token and GPU load update live.

**Who it is for:** A developer running llama.cpp locally who wants to know how fast a model answers and whether the GPU is the bottleneck.

**The problem:** A local server gives no live view of speed, latency or GPU use, so tuning a model setup is guesswork.

**The outcome:** A terminal dashboard that shows tokens per second, time to first token, GPU utilisation and memory for the server while a request runs.

### Steps

1. Install a Rust toolchain (1.75 or newer), then run `git clone https://github.com/DingoOz/llm-visuals`.
2. Run `cd llm-visuals` and then `cargo run --release -- --demo` to confirm the layout renders with synthetic data.
3. Start your model with metrics enabled, for example `llama-server --metrics`, using the flag the README shows as `llama-server.exe --metrics` on Windows. (inferred, check the README)
4. Run `cargo run --release` in the repo folder so the dashboard detects the running server.
5. Press `r` to rescan if the server is not listed, then send a prompt and watch tok/s, TTFT and the GPU panels update.

### Prompt to paste into your agent

```text
Using the repo DingoOz/llm-visuals, run `cargo run --release` against a llama-server listening on <server-port> with metrics enabled. Expected output: a dashboard listing that server with live tokens per second and time to first token while a request runs. Verify the model name and port shown match the server you started.
```

### Caveat

Throughput panels need a locally listening llama-server, and GPU panels depend on the vendor tool for your hardware, so a machine without these shows only part of the dashboard.

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/dingooz-llm-visuals/watch-local-llama-server-live/

## How to preview every dashboard view with no GPU or model with LLM Visuals

Repo: [DingoOz/llm-visuals](https://github.com/DingoOz/llm-visuals) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Run the built-in demo to see all views on synthetic data before setting up a real server.

**Who it is for:** A developer or reviewer who wants to see what the dashboard shows before installing GPU tools or loading a model.

**The problem:** Setting up servers and drivers just to see a dashboard layout is slow, and the layout may not suit the terminal they use.

**The outcome:** A rendered dashboard with synthetic server and GPU data, plus a check that the terminal shows truecolour correctly.

### Steps

1. Install a Rust toolchain (1.75 or newer) and run `git clone https://github.com/DingoOz/llm-visuals`.
2. Run `cd llm-visuals` and then `cargo run --release -- --demo`.
3. Step through the views with the keys listed under Keys in the README and press `q` to quit.
4. If colours look flat in a non-Windows-Terminal console, relaunch with the color option described in the README, for example `--color truecolor`.

### Prompt to paste into your agent

```text
Using the repo DingoOz/llm-visuals, run `cargo run --release -- --demo` in a terminal of at least 140 by 44 characters. Expected output: a dashboard showing a synthetic server and two synthetic GPUs. Verify that pressing `q` restores the terminal cleanly.
```

### Caveat

Demo mode uses synthetic data, so its numbers show the layout only and say nothing about a real model's speed.

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/dingooz-llm-visuals/preview-dashboard-views-without-hardware/

## How to record every request to a SQLite log for analysis with LLM Visuals

Repo: [DingoOz/llm-visuals](https://github.com/DingoOz/llm-visuals) | Difficulty: Intermediate | Time: 45 minutes | Basis: From the README

Start the dashboard with a log database so request rows are saved for later review.

**Who it is for:** A developer comparing model setups who wants saved request data rather than screenshots of a live screen.

**The problem:** Live dashboards disappear when closed, so runs with different models or settings cannot be compared afterwards.

**The outcome:** A SQLite file containing one row per logged request that can be opened with any SQLite tool.

### Steps

1. Install a Rust toolchain (1.75 or newer), clone the repo with `git clone https://github.com/DingoOz/llm-visuals`, and `cd llm-visuals`.
2. Start the dashboard with `cargo run --release -- --log-db llm.db`, using the `--log-db llm.db` option the README describes. (inferred, check the README)
3. Send several prompts to the running llama-server so rows are written.
4. Open `llm.db` with any SQLite client and check that rows exist for the requests you sent.

### Prompt to paste into your agent

```text
Using the repo DingoOz/llm-visuals, start the dashboard with `cargo run --release -- --log-db llm.db` while a llama-server is running. Expected output: an llm.db file in the repo folder with one row per request. Verify the row count rises after you send a prompt.
```

### Caveat

The README says rows are written to the database but does not describe the table layout, so check the schema yourself before writing queries.

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/dingooz-llm-visuals/log-llm-requests-to-sqlite/

## How to build release binaries for six OS and CPU targets with LLM Visuals

Repo: [DingoOz/llm-visuals](https://github.com/DingoOz/llm-visuals) | Difficulty: Advanced | Time: 90 minutes | Basis: From the README

Cross-compile archives for Linux, macOS and Windows on x86-64 and ARM64 from one machine.

**Who it is for:** A maintainer who ships a Rust CLI and needs binaries for every platform without a separate build machine for each.

**The problem:** Producing and testing binaries for three operating systems and two CPU families by hand is slow and error-prone.

**The outcome:** Archives for each supported target written to a dist folder on the local machine.

### Steps

1. Install a Rust toolchain (1.75 or newer) and a C compiler, since SQLite is compiled from bundled C source.
2. Run `git clone https://github.com/DingoOz/llm-visuals` and `cd llm-visuals`.
3. Run `./scripts/cross-build.sh` to build every target the host can link.
4. Check the archives written to `dist/` and compare the set against the six-target table in the README.

### Prompt to paste into your agent

```text
Using the repo DingoOz/llm-visuals, run `./scripts/cross-build.sh` from the repo root. Expected output: archives for each target this host can link, written to dist/. Verify the count of archives matches the number of targets the script reports as buildable.
```

### Caveat

Some targets need toolchains the host may lack, so the script builds only what this host can link, and the README says Windows and macOS targets are built on their own runners in CI.

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/dingooz-llm-visuals/build-release-binaries-for-six-targets/