Archyve should update its list of ModelConfigs based on what's available in its ModelServers. This would avoid the user having to ensure that a model is configured properly in both Archyve and Ollama.
For models that were available in Ollama and are not any longer, there should probably be a boolean like available?, so Archyve knows not to query it, but can retain the ModelConfig if Conversations refer to it.
Eventually, it would be great if Archyve could trigger an ollama pull via the API, if configured to do so. Then, any models it needed (like the default embedding and generation models) it could have Ollama fetch.
Not sure what do about OpenAI ModelServers. Same behaviour would be nice, if the API supports it.
Archyve should update its list of ModelConfigs based on what's available in its ModelServers. This would avoid the user having to ensure that a model is configured properly in both Archyve and Ollama.
For models that were available in Ollama and are not any longer, there should probably be a boolean like
available?
, so Archyve knows not to query it, but can retain the ModelConfig if Conversations refer to it.Eventually, it would be great if Archyve could trigger an
ollama pull
via the API, if configured to do so. Then, any models it needed (like the default embedding and generation models) it could have Ollama fetch.Not sure what do about OpenAI ModelServers. Same behaviour would be nice, if the API supports it.