Using locally installed models on Mac M1/Use of alternative models like DEEPSEEK

[BroccoliFin]

Thank you for your work.
There are several questions.

• Is it possible to use a locally installed Hermes 3 Llama 3.1 8B model on Mac M 1 when creating an agent?
• If so, how and where to indicate the model used and the path to it, I did not find this information in the tutorial or it is not obvious?
• Is it possible to use models like DeepSeek v.2, v.3 by specifying the api key as in the case of OpenAI or are only those in the list available?
• If so, how can this be done? Is it enough to specify the following configuration in the .env.example file:
  DEEPSEEK_MODEL= , DEEPSEEK_API_KEY= and change accordingly in character.json "modelProvider": "deepseek" or is this more fundamental stuff that requires meddling with other files?

[github-actions]

Hello @BroccoliFin! Welcome to the ai16z community. Thank you for opening your first issue; we appreciate your contribution. You are now a ai16z contributor!

[AIFlowML]

Hello. @BroccoliFin

In your env you need to add this data

Ollama Configuration

OLLAMA_SERVER_URL=localhost:11434
OLLAMA_MODEL=model_name
USE_OLLAMA_EMBEDDING= # Set to TRUE for OLLAMA/1024, leave blank for local
OLLAMA_EMBEDDING_MODEL= # Default: mxbai-embed-large
SMALL_OLLAMA_MODEL= # Default: llama3.2
MEDIUM_OLLAMA_MODEL= # Default: hermes3
LARGE_OLLAMA_MODEL= # Default: hermes3:70b

Deepseek is coming.

[BroccoliFin]

Many thanks for the feedback and tips!!!

[BroccoliFin]

Another question - the documentation says that for local launch you need a GPU with a CUDA, but what about the combined memory on the Mac M1?