aic-visualizer

KO | EN

gitlite — search

#anthropic #php #html #kubernetes #tensorflow #swift #csharp #golang #claude #json

(★ 9)

Visualize GHC ai credit spend based on conversation shape

File Explorer

ZIP Download

- screen1.png
- screen2.png
- app.rs
- main.rs
- model.rs
- parser.rs
- storage.rs
- ui.rs
.gitignore
Cargo.lock
Cargo.toml
LICENSE
README.md

# Use via CDN

jsDelivr

jsDelivr serves any public GitHub repository as a CDN with zero setup. Pick a version and a file to get a ready-to-paste link and snippet.

Version

File

Link

Example

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

[![Built With Ratatui](https://img.shields.io/badge/Built_With_Ratatui-000?logo=ratatui&logoColor=fff)](https://ratatui.rs/)

A terminal UI for simulating LLM conversations and visualizing their token
usage and cost. Describe each turn in plain language, pick a pricing model, and
watch the per-turn and cumulative cost build up as context accumulates.

![screen1.png](https://raw.githubusercontent.com/epic-64/aic-visualizer/main/docs/screen1.png)
![screen2.png](https://raw.githubusercontent.com/epic-64/aic-visualizer/main/docs/screen2.png)

## Requirements

- [Rust](https://rustup.rs/) (stable, edition 2024, Rust 1.85+)

## Running

```sh
cargo run
```

For an optimized build:

```sh
cargo run --release
```

Run the tests with:

```sh
cargo test
```

## Using the app

The app walks through a few screens:

1. **Model select**: pick a built-in model (Opus, Sonnet, Haiku, GPT-5, …) or
   press `c` to define a custom one with your own prices.
2. **Start**: begin a blank conversation, load a built-in example, or reopen a
   previously saved conversation.
3. **Chat**: type a description of each turn and press Enter to bill it.

### Describing a turn

Turns are free-form; the parser picks out the numbers and what they refer to.
Numbers may use `k`/`m` suffixes. Examples:

```
2k instructions, 2k skills, 10k tool defs, 1k prompt, 1k output
500 in, 500 out
500 in, 500 out, 500 thinking, repeat 10, marker: after review
```

Keywords: `prompt`/`input`/`tools`/`instructions` → input, `out`/`response`/
`completion`/`answer` → output, `think`/`reason` → thinking (billed like output
but not carried into context), anything with `cach` → an explicit cached-token
override, and `repeat N` / `N times` to apply the same turn multiple times.

### Markers

Markers are positional annotations dropped between turns. Either send one on its
own line:

```
marker
marker: reviewed here
```

…or tuck one onto the end of a turn line; it is stripped from the stored turn
and a marker is placed right after it:

```
300 prompt, 400 out, marker: after review
```

### Other chat commands

- `clear`: wipe the current conversation.

## Keyboard shortcuts

**Global**

- `Ctrl-C`: quit

**Model select**

- `↑`/`↓` (or `k`/`j`): move · `Enter`: choose · `c`: custom model · `q`: quit

**Start picker**

- `↑`/`↓`: move · `Enter`: open · `d`/`Delete`: delete a saved conversation ·
  `Esc`: back

**Chat**

- `Enter`: submit turn · `↑`/`↓`: recall previous input lines · `Esc`: back to
  model select
- `Ctrl-S`: save conversation · `Ctrl-T`: new tab · `Ctrl-W`: close tab ·
  `Tab`/`Shift-Tab`: switch tabs
- Mouse wheel: scroll the turn history

## Saved conversations

Conversations are saved as plain-text files under
`~/.token-visualizer/conversations/`. Each file carries its model and prices in
`# key: value` headers followed by one turn (or marker) per line, so a saved
conversation restores itself completely on reload.

## License

Released under the [MIT License](LICENSE).

aic-visualizer

Owner Repos

File Explorer

# Use via CDN

Was this content helpful?

Recent Feedback

aic-visualizer

Owner Repos

File Explorer

# Use via CDN

Was this content helpful?

Recent Feedback

Download README