Shared Envs

[igardev]

In version v0.0.15, the concept of Env (environment) is introduced. This is a group of up to 4 models (completion, chat, embeddings, tools). Starting an env, starts it's models. If a model is configured as local, it will start a llama.ccp server. If the model uses external server, llama-vscode will just remember the configuration.
Each env contains information about it's models - name, endpoint, local start command, model for provider, is api key required. Some of the fields could be empty.
Envs could be exported from and imported into llama.vscode.

The goal of this this discussion is to share Envs (configurations) for different use cases.

Configurations with local gpt-oss 20B:
Local, full package - min, gpt-oss 20B ( more than 20GB VRAM HD 16 GB).json
Local, full package - medium, gpt-oss 20B (more than 32GB VRAM HD 20 GB).json
Local, full package - max, gpt-oss 20B (more than 48GB VRAM HD 30 GB).json

Here are the configurations, which are predefined in llama-vscode v0.0.14:

Only completions (less than 8GB VRAM).json
Only completions (less than 16GB VRAM).json
Only completions (more than 16GB VRAM).json
OpenAI gpt-oss 20B Agent LOCAL, chat ( more than 16GB VRAM).json
OpenAI gpt-oss 20B agent, chat - ( less than 8GB VRAM) (requires API key from OpenRouter).json
Silence.json
OpenAI gpt-0ss 20B chat only LOCAL (more than 16GB VRAM).json
Agent & chat (less than 16GB VRAM) (requires OpenRouter API key).json
Agent, comletions, chat - max (more than 32 GB VRAM).json
Agent, completions & chat - medium (less than 32GB VRAM)).json
Only chat & edit (CPU Only).json
Only chat & edit (less than 8GB VRAM).json
Only chat, chat with project context & edit (less than 16GB VRAM).json
Only chat, chat with project context & edit (more than 16 GB VRAM).json
Only completions - CPU.json