Skip to content

v0.4.0-beta: AI Autopilot, MCP Memory & Slack Agent Integration

Choose a tag to compare

View all tags
@DepthSight-Pro DepthSight-Pro released this 12 Jul 16:55
· 61 commits to main since this release
v0.4.0-beta
353a843

🚀 What's New in v0.4.0-beta

This major release introduces a full suite of autonomous AI capabilities, completely transforming how trading strategies are built, tested, and deployed on DepthSight.

🤖 1. AI Autopilot Agent

  • Autonomous Optimization Loop: A self-correcting cycle (Generate -> Backtest -> Learn -> Repeat) that builds and refines trading strategies automatically.
  • Multi-Agent Architecture: Built-in specialized roles (Generator, Critic, and Strategy Advisor) that work together to analyze results and suggest improvements.
  • Visual Chart Analysis: On the frontend, the agent can "see" chart screenshots to recognize patterns and build strategies based on visual context.

🧠 2. MCP (Model Context Protocol) Memory Server

  • Semantic Memory Retrieval: The AI now remembers past backtests, successful parameter combinations, and historical mistakes.
  • Dynamic Rules Context: Automatically feeds the AI with global trading guidelines and user-specific insights to prevent overfitting and over-filtering.
  • Continuous Learning: Every optimization loop adds new data to the vector database, making the agent smarter over time.

💬 3. Slack Agent Integration (/depthsight)

  • Full Platform Control via Slack: Run backtests, optimize strategies, and interact with the AI directly from your workspace.
  • Real-time Performance Cards: Beautifully rendered backtest reports with equity curves, KPIs, and PnL metrics delivered straight to your chat.
  • Interactive Workflows: Added interactive buttons after successful optimization to instantly Open in Editor or Deploy Live Bot.
  • Context Management: Added /depthsight clear command to easily reset the agent's context and prevent hallucination loops.

🛠️ Improvements & Bug Fixes

  • Fixed an issue with silent failures during the rendering of final Autopilot report cards.
  • Refactored backtest polling for long-running Celery tasks to ensure stable delivery of results.
  • Optimized the strictness of the AI Strategy Advisor to prevent generating strategies with overly restrictive confluence (the "2 trades only" issue).
Assets 2
Loading