🧪 Experimental Preview: AI Bot Generation & Snapshot Flow

We are thrilled to introduce an early look at the Insights snapshot flow, providing a new mechanism for creating customized trading and analysis bots. Please note that AI Bot Generation is currently an experimental preview feature.

As we continue to test and refine this capability, early adopters can explore a robust new system guided directly by sophisticated prompt templates and a newly implemented placeholder catalog. The update introduces the ability to selectively customize default system and user prompts while maintaining deep BYOK (Bring Your Own Key) model filtering. Under the hood, the backend has been upgraded to strictly validate bot snapshot prompt placeholders and support fully paginated AI config prompts, laying a strong foundation for future releases.

💬 Insights Chat Interface Revamped

A significant engineering effort went into refining the Insights Chat UI, eliminating visual bugs such as "tearing" gaps and volatile scroll jumps. The interface now features a smooth, native scroll behavior that calculates offsets directly from structural bounding boxes instead of arbitrary timeouts.

Whether you are jumping into a historical session or firing off a new prompt, the introduction of a dynamic runway spacer guarantees that messages comfortably reach the viewport top and anchor accurately without layout shifting. To complement these UI shifts, insights messages utilizing compact snapshot cards will intelligently reflow, drastically improving readability.

⚙️ System Hardening & Performance

Critical improvements were made to system settings and performance configurations via optimizations to the RAG (Retrieval-Augmented Generation) analysis service. The insights snapshot killswitch was aligned to guarantee configuration parity across the system, while the resolution scope for Web Search Chat API keys was reinforced to prevent unauthorized leakage and guarantee securely isolated context resolution.

🛡️ Expanded Testing & Reliability

To secure these sweeping architectural changes and prepare the experimental bot features for future general availability, an extensive suite of cross-stack tests has been merged. Robust testing pipelines now enforce rules for AI snapshot naming, RAG configuration contracts, prompt template context processing, and BYOK models on the backend. Meanwhile, the frontend has received dedicated testing suites evaluating feature reliability and UI resilience.