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

# API

> API reference for Gemini: every `gemini.api.*` operation with input and output types.

Every `gemini.api.*` operation is listed below with parameter shapes and return types from the plugin Zod schemas.

<Info>
  **New to Corsair?** See [API access](/concepts/api), [authentication](/concepts/auth), and [error handling](/concepts/error-handling).
</Info>

## Content

### countTokens

`content.countTokens`

Count tokens in text using Gemini tokenization, for cost estimation and input limit checks

**Risk:** `read`

```ts theme={null}
await corsair.gemini.api.content.countTokens({});
```

**Input**

| Name       | Type       | Required | Description                                                                 |
| ---------- | ---------- | -------- | --------------------------------------------------------------------------- |
| `model`    | `string`   | Yes      | Gemini model to count tokens against, e.g. gemini-2.5-flash, gemini-2.5-pro |
| `contents` | `object[]` | Yes      | The content to count tokens for                                             |

<AccordionGroup>
  <Accordion title="contents full type">
    ```ts theme={null}
    {
      role?: user | model,
      parts: {
        text?: string,
        inlineData?: {
          mimeType: string,
          data: string
        }
      }[]
    }[]
    ```
  </Accordion>
</AccordionGroup>

**Output**

| Name                      | Type     | Required | Description |
| ------------------------- | -------- | -------- | ----------- |
| `totalTokens`             | `number` | Yes      | —           |
| `cachedContentTokenCount` | `number` | No       | —           |

***

### embedContent

`content.embedContent`

Generate a numerical vector embedding for text, for semantic search and similarity comparison

**Risk:** `read`

```ts theme={null}
await corsair.gemini.api.content.embedContent({});
```

**Input**

| Name       | Type     | Required | Description                                                                                                           |
| ---------- | -------- | -------- | --------------------------------------------------------------------------------------------------------------------- |
| `model`    | `string` | Yes      | Gemini embedding model, e.g. gemini-embedding-001                                                                     |
| `content`  | `object` | Yes      | The content to embed                                                                                                  |
| `taskType` | `string` | No       | Optional task type hint, e.g. SEMANTIC\_SIMILARITY, RETRIEVAL\_QUERY, RETRIEVAL\_DOCUMENT, CLASSIFICATION, CLUSTERING |
| `title`    | `string` | No       | Optional title, used with RETRIEVAL\_DOCUMENT taskType                                                                |

<AccordionGroup>
  <Accordion title="content full type">
    ```ts theme={null}
    {
      role?: user | model,
      parts: {
        text?: string,
        inlineData?: {
          mimeType: string,
          data: string
        }
      }[]
    }
    ```
  </Accordion>
</AccordionGroup>

**Output**

| Name        | Type     | Required | Description |
| ----------- | -------- | -------- | ----------- |
| `embedding` | `object` | Yes      | —           |

<AccordionGroup>
  <Accordion title="embedding full type">
    ```ts theme={null}
    {
      values: number[]
    }
    ```
  </Accordion>
</AccordionGroup>

***

### generateContent

`content.generateContent`

Generate text or speech audio from a prompt using a Gemini Flash/Pro model

**Risk:** `read`

```ts theme={null}
await corsair.gemini.api.content.generateContent({});
```

**Input**

| Name                | Type       | Required | Description                                                          |
| ------------------- | ---------- | -------- | -------------------------------------------------------------------- |
| `model`             | `string`   | Yes      | Gemini model to generate with, e.g. gemini-2.5-flash, gemini-2.5-pro |
| `contents`          | `object[]` | Yes      | Conversation contents to generate a response for                     |
| `generationConfig`  | `object`   | No       | —                                                                    |
| `safetySettings`    | `object[]` | No       | —                                                                    |
| `systemInstruction` | `object`   | No       | —                                                                    |

<AccordionGroup>
  <Accordion title="contents full type">
    ```ts theme={null}
    {
      role?: user | model,
      parts: {
        text?: string,
        inlineData?: {
          mimeType: string,
          data: string
        }
      }[]
    }[]
    ```
  </Accordion>

  <Accordion title="generationConfig full type">
    ```ts theme={null}
    {
      temperature?: number,
      topP?: number,
      topK?: number,
      candidateCount?: number,
      maxOutputTokens?: number,
      stopSequences?: string[],
      responseModalities?: TEXT | IMAGE | AUDIO[]
    }
    ```
  </Accordion>

  <Accordion title="safetySettings full type">
    ```ts theme={null}
    {
      category: string,
      threshold: string
    }[]
    ```
  </Accordion>

  <Accordion title="systemInstruction full type">
    ```ts theme={null}
    {
      role?: user | model,
      parts: {
        text?: string,
        inlineData?: {
          mimeType: string,
          data: string
        }
      }[]
    }
    ```
  </Accordion>
</AccordionGroup>

**Output**

| Name            | Type       | Required | Description                                                                |
| --------------- | ---------- | -------- | -------------------------------------------------------------------------- |
| `candidates`    | `object[]` | No       | —                                                                          |
| `usageMetadata` | `object`   | No       | —                                                                          |
| `text`          | `string`   | No       | Convenience field: first candidate text with markdown code fences stripped |

<AccordionGroup>
  <Accordion title="candidates full type">
    ```ts theme={null}
    {
      content?: {
        role?: user | model,
        parts: {
          text?: string,
          inlineData?: {
            mimeType: string,
            data: string
          }
        }[]
      },
      finishReason?: string,
      index?: number,
      safetyRatings?: any[]
    }[]
    ```
  </Accordion>

  <Accordion title="usageMetadata full type">
    ```ts theme={null}
    {
      promptTokenCount?: number,
      candidatesTokenCount?: number,
      totalTokenCount?: number,
      cachedContentTokenCount?: number
    }
    ```
  </Accordion>
</AccordionGroup>

***

## Images

### generateImage

`images.generateImage`

Generate a raster image (JPG/PNG/WebP) from a prompt using a Nano Banana image model

**Risk:** `write`

```ts theme={null}
await corsair.gemini.api.images.generateImage({});
```

**Input**

| Name               | Type       | Required | Description                                                                                                             |
| ------------------ | ---------- | -------- | ----------------------------------------------------------------------------------------------------------------------- |
| `model`            | `string`   | Yes      | Nano Banana image model, e.g. gemini-2.5-flash-image, gemini-3-pro-image-preview, gemini-2.0-flash-exp-image-generation |
| `prompt`           | `string`   | Yes      | Text prompt describing the image to generate                                                                            |
| `referenceImages`  | `object[]` | No       | Optional reference/input images for image editing or composition                                                        |
| `generationConfig` | `object`   | No       | —                                                                                                                       |

<AccordionGroup>
  <Accordion title="referenceImages full type">
    ```ts theme={null}
    {
      mimeType: string,
      data: string
    }[]
    ```
  </Accordion>

  <Accordion title="generationConfig full type">
    ```ts theme={null}
    {
      temperature?: number,
      topP?: number,
      topK?: number,
      candidateCount?: number,
      maxOutputTokens?: number,
      stopSequences?: string[],
      responseModalities?: TEXT | IMAGE | AUDIO[]
    }
    ```
  </Accordion>
</AccordionGroup>

**Output**

| Name            | Type       | Required | Description |
| --------------- | ---------- | -------- | ----------- |
| `images`        | `object[]` | Yes      | —           |
| `candidates`    | `object[]` | No       | —           |
| `usageMetadata` | `object`   | No       | —           |

<AccordionGroup>
  <Accordion title="images full type">
    ```ts theme={null}
    {
      mimeType: string,
      contentBase64: string
    }[]
    ```
  </Accordion>

  <Accordion title="candidates full type">
    ```ts theme={null}
    {
      content?: {
        role?: user | model,
        parts: {
          text?: string,
          inlineData?: {
            mimeType: string,
            data: string
          }
        }[]
      },
      finishReason?: string,
      index?: number,
      safetyRatings?: any[]
    }[]
    ```
  </Accordion>

  <Accordion title="usageMetadata full type">
    ```ts theme={null}
    {
      promptTokenCount?: number,
      candidatesTokenCount?: number,
      totalTokenCount?: number,
      cachedContentTokenCount?: number
    }
    ```
  </Accordion>
</AccordionGroup>

***

## Models

### listModels

`models.listModels`

List available Gemini and Veo models and their capabilities/limits

**Risk:** `read`

```ts theme={null}
await corsair.gemini.api.models.listModels({});
```

**Input**

| Name        | Type     | Required | Description |
| ----------- | -------- | -------- | ----------- |
| `pageSize`  | `number` | No       | —           |
| `pageToken` | `string` | No       | —           |

**Output**

| Name            | Type       | Required | Description |
| --------------- | ---------- | -------- | ----------- |
| `models`        | `object[]` | Yes      | —           |
| `nextPageToken` | `string`   | No       | —           |

<AccordionGroup>
  <Accordion title="models full type">
    ```ts theme={null}
    {
      name: string,
      baseModelId?: string,
      version?: string,
      displayName?: string,
      description?: string,
      inputTokenLimit?: number,
      outputTokenLimit?: number,
      supportedGenerationMethods?: string[],
      temperature?: number,
      topP?: number,
      topK?: number
    }[]
    ```
  </Accordion>
</AccordionGroup>

***

## Videos

### generateVideos

`videos.generateVideos`

Generate a text-to-video clip using a Veo model; returns an operation name for status tracking

**Risk:** `write`

```ts theme={null}
await corsair.gemini.api.videos.generateVideos({});
```

**Input**

| Name         | Type     | Required | Description                                  |
| ------------ | -------- | -------- | -------------------------------------------- |
| `model`      | `string` | Yes      | Veo video model, e.g. veo-2.0-generate-001   |
| `prompt`     | `string` | Yes      | Text prompt describing the video to generate |
| `image`      | `object` | No       | —                                            |
| `parameters` | `object` | No       | —                                            |

<AccordionGroup>
  <Accordion title="image full type">
    ```ts theme={null}
    {
      bytesBase64Encoded: string,
      mimeType: string
    }
    ```
  </Accordion>

  <Accordion title="parameters full type">
    ```ts theme={null}
    {
      aspectRatio?: string,
      personGeneration?: string,
      numberOfVideos?: number,
      durationSeconds?: number,
      negativePrompt?: string
    }
    ```
  </Accordion>
</AccordionGroup>

**Output**

| Name            | Type      | Required | Description                                                        |
| --------------- | --------- | -------- | ------------------------------------------------------------------ |
| `operationName` | `string`  | Yes      | Pass to GEMINI\_GET\_VIDEOS\_OPERATION or GEMINI\_WAIT\_FOR\_VIDEO |
| `done`          | `boolean` | No       | —                                                                  |

***

### getVideosOperation

`videos.getVideosOperation`

(Deprecated — use videos.waitForVideo) Check the status of a Veo video generation operation

**Risk:** `read`

```ts theme={null}
await corsair.gemini.api.videos.getVideosOperation({});
```

**Input**

| Name            | Type     | Required | Description                                             |
| --------------- | -------- | -------- | ------------------------------------------------------- |
| `operationName` | `string` | Yes      | The operation name returned by GEMINI\_GENERATE\_VIDEOS |

**Output**

| Name       | Type      | Required | Description |
| ---------- | --------- | -------- | ----------- |
| `name`     | `string`  | Yes      | —           |
| `done`     | `boolean` | No       | —           |
| `metadata` | `any`     | No       | —           |
| `error`    | `object`  | No       | —           |
| `response` | `object`  | No       | —           |

<AccordionGroup>
  <Accordion title="error full type">
    ```ts theme={null}
    {
      code?: number,
      message?: string
    }
    ```
  </Accordion>

  <Accordion title="response full type">
    ```ts theme={null}
    {
      generateVideoResponse?: {
        generatedSamples?: {
          video?: {
            uri?: string
          }
        }[]
      }
    }
    ```
  </Accordion>
</AccordionGroup>

***

### waitForVideo

`videos.waitForVideo`

Poll a Veo video generation operation until it completes and return the generated video

**Risk:** `write`

```ts theme={null}
await corsair.gemini.api.videos.waitForVideo({});
```

**Input**

| Name             | Type     | Required | Description                                             |
| ---------------- | -------- | -------- | ------------------------------------------------------- |
| `operationName`  | `string` | Yes      | The operation name returned by GEMINI\_GENERATE\_VIDEOS |
| `pollIntervalMs` | `number` | Yes      | Delay between status checks                             |
| `timeoutMs`      | `number` | Yes      | Give up and return done=false after this long           |

**Output**

| Name            | Type      | Required | Description |
| --------------- | --------- | -------- | ----------- |
| `operationName` | `string`  | Yes      | —           |
| `done`          | `boolean` | Yes      | —           |
| `data`          | `object`  | No       | —           |
| `error`         | `object`  | No       | —           |

<AccordionGroup>
  <Accordion title="data full type">
    ```ts theme={null}
    {
      video_file?: {
        mimeType: string,
        contentBase64: string
      }
    }
    ```
  </Accordion>

  <Accordion title="error full type">
    ```ts theme={null}
    {
      code?: number,
      message?: string
    }
    ```
  </Accordion>
</AccordionGroup>

***