1. Tier Your Models by Task Complexity
Use Claude Sonnet 4.6 or GPT-5.4 for high-level planning and slide narrative, while routing bulk CSV cleaning and log summarization to DeepSeek-V3.2 or Gemini 2.5 Flash to cut credit consumption by up to 70%.
Combine WorkBuddy's built-in TokenHub routing with your own API keys (Avatar → Settings → Models) to balance frontier reasoning quality and token cost.
Use Claude Sonnet 4.6 or GPT-5.4 for high-level planning and slide narrative, while routing bulk CSV cleaning and log summarization to DeepSeek-V3.2 or Gemini 2.5 Flash to cut credit consumption by up to 70%.
Select Gemini 2.5 Pro when cross-checking dozens of local PDFs or large codebases so the agent inspects full documents without chunk-boundary loss.
In Settings → Models, add any OpenAI-compatible API key (OpenAI, Anthropic, DeepSeek, Gemini, OpenRouter) or a local Ollama server to run tasks at zero WorkBuddy credit cost.
Always define the target folder and output filename (.pptx, .xlsx, .md) up front so the agent validates its output in a single pass without costly retries.