What AI providers does InitRunner support?
InitRunner supports a wide range of AI providers including OpenAI, Anthropic (Claude), Google (Gemini), Ollama (local models), Azure OpenAI, AWS Bedrock, Mistral, Groq, DeepSeek, OpenRouter, and any OpenAI-compatible endpoint. Providers are auto-detected from environment variables or can be set explicitly in YAML.
Can I import existing PydanticAI or LangChain agents into InitRunner?
Yes, InitRunner allows you to import existing PydanticAI or LangChain agents. You can run 'initrunner new --pydantic-ai agent.py' or 'initrunner new --langchain agent.py' to convert them to InitRunner's YAML format. This process automatically extracts model configurations, system prompts, and tool definitions.
Does InitRunner support multimodal input for agents?
Yes, InitRunner supports multimodal input including images, audio, video, and documents. You can use the '--attach' flag in the CLI to send files or URLs alongside text prompts, or configure attachments directly in your YAML. Vision, transcription, and document parsing are supported across the CLI, API server, and dashboard.
How does InitRunner handle agent memory and learning?
InitRunner provides three memory types: semantic, episodic, and procedural, allowing agents to remember facts, record outcomes, and learn policies. Session persistence is supported, and a consolidation pass rolls up episodic records into durable knowledge. Memory is on by default in ephemeral mode, with a '--resume' option to pick up where you left off.
Can InitRunner agents operate autonomously and with budget constraints?
Yes, agents can operate autonomously with the '--autopilot' flag, planning tasks, executing steps, and self-correcting. InitRunner also supports budget-aware agents through 'guardrails' in YAML, allowing you to set per-run token caps, session budgets, and daily/lifetime daemon budgets to prevent runaway loops and manage costs.