v0.8.0: OpenAI-Compatible Provider + Agent Invoke

[idapixl]

cortex-engine v0.8.0

Two features that change how agents use cortex:

1. OpenAI-Compatible LLM Provider

One adapter, entire ecosystem. Set an env var, pick a model:

Provider	Env Var	Example Model
DeepSeek	DEEPSEEK_API_KEY	deepseek-chat
Hugging Face	HF_TOKEN	mistralai/Mistral-Nemo-Instruct-2407
OpenRouter	OPENROUTER_API_KEY	deepseek/deepseek-chat
OpenAI	OPENAI_API_KEY	gpt-4o
LM Studio	(none needed)	Whatever you loaded

Auto-detects the provider from your env vars. No URL configuration needed for known providers.

2. agent_invoke Tool

Dispatch tasks to a cortex-backed agent. It queries cortex for existing knowledge, completes the task with your cheap LLM, and stores findings back. Knowledge compounds across sessions.

Before: Spawn an expensive subagent that starts cold every time.
After: Call agent_invoke — it knows what you researched last week.

What This Enables

• Background research at $0.01/task instead of $1+
• Knowledge that accumulates across sessions
• Any LLM as your agent backend — local or cloud
• The foundation for autonomous cognitive processes

Try it: npm install cortex-engine@0.8.0

Release notes | Wiki