specialist

func (*Client) Describe ¶

func (c *Client) Describe(ctx context.Context, path, model, prompt string) (string, error)

Describe turns an image into text by sending it to a vision model and returning the model's textual description.

It is the building block for the future Phase 2 `read_image` tool, which will let any agent (including a text-only one) pull an image into the conversation as model-generated text on demand. It is deliberately NOT wired into the attachment routing path: that path Blocks images on text-only models and nudges the user to switch to a vision model, rather than silently substituting a lossy AI-generated description for the actual image. (Audio, by contrast, auto-falls back to faithful transcription — see attachments.Route.) So Describe is intentionally ahead of its consumer in this branch.

func (*Client) DescribeVideo ¶

func (c *Client) DescribeVideo(ctx context.Context, path, model, prompt string) (string, error)

DescribeVideo turns a video into text by sending it to a video-capable model (via a native video_url content part) and returning the model's textual description. Mirrors Describe; it is the building block for the read_video tool. Empty model ⇒ LocalAI's default chat model (which must support video input).

func (*Client) Transcribe ¶

func (c *Client) Transcribe(ctx context.Context, path, model string) (string, error)

Transcribe uploads the audio file to {baseURL}/audio/transcriptions and returns the text. An empty model lets LocalAI pick the first FLAG_TRANSCRIPT model.