AI Chat and Summary Blocks

The AI Chat feature provides an interactive interface for querying sentiment data. Powered by a fine-tuned language model, it accepts natural language inputs—e.g., “What’s the sentiment on $SOL today?”—and returns concise, context-aware responses based on distilled insights from all the gathered batches. Due to high context, we process chats also according to USD value priority to keep them more relevant and more recently updated.

Both features rely on a shared inference endpoint, running on a subset of the GPU cluster with a latency target of 500ms per query. The Summary Block updates hourly, while the AI Chat responds in real-time, making them ideal for both quick checks and deep dives.