> ## Documentation Index
> Fetch the complete documentation index at: https://docs.bizyair.ai/llms.txt
> Use this file to discover all available pages before exploring further.

#  Concepts

> A glossary of BizyAir's core concepts to help you quickly understand the platform

## Glossary

| Term                 | Description                                                                                                                 |
| -------------------- | --------------------------------------------------------------------------------------------------------------------------- |
| Token                | The smallest unit used for LLM / VLM billing and context metering                                                           |
| Context window       | The maximum number of tokens a model can process in a single inference; exceeding the limit requires truncation or chunking |
| Inference parameters | `temperature`, `top_p`, `max_tokens`, `seed`, etc., controlling the randomness, length, and reproducibility of the output   |
| Positive prompt      | Text describing the target image / content                                                                                  |
| Negative prompt      | Describes elements to avoid (supported by some models)                                                                      |
| Balance              | The platform's unified billing balance (USD), shared between the web and plugin clients                                     |
| API Key              | A 48-character account-level credential used for REST API authentication                                                    |
| request\_id          | The unique identifier returned after each task submission, used to query results                                            |
| Outputs              | The task result set, containing fields such as `object_url` / `output_ext` / `audit_status`                                 |

## Model Types and Modalities

On BizyAir, models are classified by modality: text (LLM), vision (VLM), image, video, and audio.

By invocation method, they are divided into four types: "Online Playground", "API Integration", "ComfyUI Node", and "AI App".

Each model's support modality, input/output format, and billing method are noted on its [Node Pricing](/pricing/node-price) detail page.

## Prompt Engineering and Output Control

* **Prompt**: The positive prompt describes the target content; the negative prompt describes elements to avoid.
* **Aspect control**: Control output dimensions via `width` / `height` / `aspect`.
* **Sampling quality**: Control sampling steps and guidance strength via `steps` / `guidance`.
* **Reproducibility**: Fix `seed` to lock the result; set to `-1` for random.
* **Safety limits**: Overly long text triggers a `Prompt Validation Error`; sensitive descriptions trigger a safety intercept.

<Tip>
  Video generation time scales with duration — validate the effect with a short clip first before scaling up.
</Tip>

## Account System and Content Lifecycle

* **Single identity system**: The web and API sides share the same account and balance.
* **API Key**: An API Key is an account-level credential. Keep it safe; if leaked, reset it under "User Center → API Keys".
* **Content retention**: Content is retained for 15 days by default — please download it in time.