Defaults and settings

Adding a new model in the config.json

The config.json containing the models specifications is located in ~/.cache/deadend/config.json. The file is supposed to handle the text generation models but also the text embedding models. When defining a model, the overall schema is the following :

"<provider>:<model_name>": {
  "provider": "<provider>", // corresponds to the provider, for ex. openai, anthropic...
  "model_name": "<model_name>", // corresponds to the model name, gpt-5
  "api_key": "<api_key>", // could be optional if the corresponding ENV var is set, however it's better to just add it here
  "base_url": "<base_url>", // In the case of specific gateways or providers (or using ollama), you might need to set the base URL
  "type_model": null, // this field is to be set only if the model is an embedding model
  "vec_dim": null   // Also in case of an embedding model, we can set up the vector space dimension, and is defaulted to 1536
}

For example the config.json could be the following :

{
  "provider": {
    "anthropic:claude-sonnet-4-5": {
      "provider": "anthropic",
      "model_name": "claude-sonnet-4-5",
      "api_key": "<api_key>",
      "base_url": null,
      "type_model": null,
      "vec_dim": null
    },
    "openrouter:qwen/qwen3-embedding-8b": {
      "provider": "openrouter",
      "model_name": "qwen/qwen3-embedding-8b",
      "api_key": "<api_key>",
      "base_url": "https://openrouter.ai/api/v1/embeddings",
      "type_model": "embeddings",
      "vec_dim": 4096
    },
  }
}