AI: Generate captions using the Ollama API

As an Ollama user, I would like to generate captions directly through the Ollama service API, so that I don't need to set up a dedicated PhotoPrism Vision API service if I don't need it otherwise.

Below is an example of the required vision.yml model configuration from our Developer Guide :

Models:
- Type: labels
  Default: true
- Type: nsfw
  Default: true
- Type: face
  Default: true
- Type: caption
  Name: gemma3
  Version: latest
  Prompt: Create a caption with exactly one sentence in the active voice that describes
    the main visual content. Begin with the main subject and clear action. Avoid text
    formatting, meta-language, and filler words.
  Resolution: 720
  Options:
    Temperature: 0.1
  Service:
    # Ollama API endpoint (adjust as needed):
    Uri: http://ollama:11434/api/generate
    FileScheme: base64
    RequestFormat: ollama
    ResponseFormat: ollama
Thresholds:
  Confidence: 10

Note that the Options parameters are optional and will have no effect when running Ollama through the dedicated Vision Service.

Documentation:

Related Discussions:

AI: Generate captions using the Ollama API #5011 #5123

The request and response handling should be refactored eventually for better maintainability. Other than that, these changes seem to work, so now both the PhotoPrism Vision API and the Ollama API request and response formats are supported. An updated preview build will be available for testing soon!

AI: Use defaults when returning classify label from response #5011 #5123

Index: Improve generation, caching and logging of labels #5011 #5123

Setup: Update compose.yaml examples and add ollama service #5123

Setup: Update compose.yaml config examples #5123