What problem does it solve?
When building AI applications with TanStack AI, it is hard to see what happens inside the pipeline — what requests go to providers, which chunks stream back, how tools and middleware behave, and where errors occur. This Skill provides a single debug option that turns on structured, category-tagged logging across every activity without writing custom instrumentation.
Core Features & Use Cases
- Unified debug option: Pass
debug: true | false | DebugConfig to chat(), summarize(), generateImage(), generateSpeech(), generateTranscription(), and generateVideo().
- Per-category toggles: Enable or silence eight categories — request, provider, output, middleware, tools, agentLoop, config, and errors — with unspecified flags defaulting to true.
- Custom logger integration: Pipe logs into pino, winston, or any logger implementing the
Logger interface via debug: { logger }.
- Use Case: While debugging a streaming chat agent, enable only the
provider and output categories to trace raw provider chunks and emitted output without noise from middleware or agent-loop logs.
Quick Start
Enable debug logging on a chat call by setting the debug option to true and ask the assistant to show the categorized log output.
