The assistant
Ask, Build, Learn, choosing who answers and what is sent.
- The assistant: What the assistant does and what it never does.
- Ask: Questions, answers and grey drafts.
- Build: The assistant builds on the canvas step by step.
- Learn: Lessons you build yourself.
- Grey drafts: Suggestions on the canvas and accepting them.
- Pausing and stopping: The step limit, Continue, Stop and undoing a build.
- Choosing who answers: Cloud models and models on your own computer.
- Ollama: Using a model served by Ollama.
- LM Studio: Using a model served by LM Studio.
- llama.cpp server: Using a model served by llama-server.
- vLLM and other compatible servers: Any server that speaks the OpenAI-compatible API.
- OpenAI: Getting a key, cost and settings.
- Anthropic: Getting a key, cost and settings.
- Google Gemini: Getting a key, cost and settings.
- OpenAI-compatible services: A gateway or regional cloud service at its own address.
- Which model to take: Suggested models, what they cost and how much memory they need.
- Keys and where they are kept: Pasting, remembering and forgetting a key.
- The context window: Why a local model forgets and how to give it more room.
- Web search and pages: When the assistant reads the internet and how to turn it off.
- What is sent to the model: Exactly what a question carries.
- The conversation log: Saving what was sent and answered, to have an answer looked into.
- Failed answers: Every error message of the assistant and what to do.
- Working without a model: What works when no assistant is chosen.
- Limits of the assistant: What the assistant cannot do yet.