Ollama: dynamic model list from /api/tags not wired to model dropdown

[schauerr-lcc]

Summary

When connecting to a local Ollama instance, the "Test connection" button
successfully queries /api/tags and reports the correct number of installed
models. However, the model dropdown remains static and doesn't populate from
the discovered models. Any model not in the hardcoded list returns a 404 on
actual API calls.

Steps to reproduce

1. Set provider to Ollama, enter a valid base URL (e.g. http://localhost:11434)
2. Click "Test connection" — reports correct installed model count
3. Open the model dropdown — shows only hardcoded options, not installed models
4. Select a hardcoded model that isn't installed locally → 404 on generate call

Expected behavior

After a successful connection test, the model dropdown should be populated
from the /api/tags response so users can select from their actually
installed models.

Suggested implementation

The /api/tags response returns a models array with name fields. After
a successful connection test, these could replace or augment the static
dropdown options for the Ollama provider. The data is already in hand at
connection-test time — it just needs to be fed to the select element.

Environment

• App running from file:// (static HTML, no server)
• Ollama running locally on Windows
• Tested with Ollama 2 models installed; one of these--llama3.1:8b--is in the
  hardcoded list, but still doesn't match as the app omits the 8b in its request

Notes

Great idea for an app! This is going to be incredibly useful to me and my team.