LLM agent

Build a conversational or agentic LLM over your own functions using Runware-hosted models. The model reasons, decides when to call the tools you define, reads what they return, and answers with that data in hand. Runware exposes an OpenAI-compatible endpoint, so this is the standard chat-completions + tool-calling format, not the imageInference task shape the image skills use.

Inputs to collect

• The job: plain chat/reasoning, or tool-calling over the user's own functions.
• The tools (if agentic): each function's name, what it does, its arguments, and how the user runs it. The model never touches the systems directly. It only sees the schemas you hand it and the JSON they return.
• The system prompt: the agent's role and how eagerly it should act ("gather context before acting" vs "act immediately").
• Constraints: latency/cost target, context-window needs, whether reasoning depth matters.

Models

Address by AIR. Confirm each is live and inspect its schema via runware-models + runware-run before calling. Never hardcode a stale pick.