streaming

Real-time streaming delivers LLM responses incrementally, character by character, as the model generates them. This Skill implements asynchronous streaming across OpenAI, Anthropic, Google Gemini, and Ollama, enabling per-chunk delivery so users see text appear naturally rather than waiting for full completion.

Yes. This Skill provides a unified streaming API that works consistently across OpenAI, Anthropic, Google Gemini, and Ollama, eliminating provider-specific complexity and letting you switch providers without rewriting your streaming logic.

Streaming automatically captures token usage and statistics from the final stream chunk, providing cost monitoring and optimization data without additional configuration. Usage metadata is extracted and integrated into the response.

Streaming improves perceived performance and user engagement by displaying output as it arrives, keeping interfaces responsive during long generations. Users see progress immediately rather than blank screens, making interactions feel more natural and immediate.

This Skill handles asynchronous streaming internally with delta-based chunks and robust error handling, so you don't need deep async expertise. The unified API abstracts provider differences and stream complexity behind a straightforward interface.

The Skill includes robust error handling built into its streaming implementation, managing interruptions and connection issues across providers. Error recovery is handled internally so your chat interface remains stable.