WARNING: NOT AN INVESTMENT TOOL. This is a technical demonstration of multi-agent AI systems. All output is AI-generated and must not be used as financial advice.
Release History

Changelog

What's shipped in ProspectAI — the multi-agent brain behind this system (Market Analyst → Technical Analyst → Fundamental Analyst → Draft Strategist → Critic → Final Strategist) and the web frontend that drives it. Newest first.

ProspectAI Module

v1.9.0

  • enhance LONG-BUY trade setup logic to ensure stop_loss and entry zones are anchored to current_price
  • harden PositionRecommendation schema: auto-correct above-zone trade_setup invariant violations before Pydantic validation
  • remove scaled_entry_setups from draft output in ProspectAIFlow
  • remove SCALED-ENTRY action and related logic from PositionRecommendation schema
  • implement reasoning-based action selection in Draft Strategist, removing SCALED-ENTRY and enhancing Critic validation
ProspectAI Module

v1.8.1

  • catch LONG-BUY outside entry zone, degenerate zones, and bucket accounting errors
  • add risk aversion profile selector to portfolio allocation
  • implement continuous scrolling marquee for recent-wins strip with hover pause and seamless looping
  • enhance token tracking in ExecutionTracker and AnthropicCachingCompletion for improved cache metrics
ProspectAI Module

v1.8.0

  • update tests to use AnthropicCachingCompletion and improve structure
  • replace CachingLLM with make_caching_llm factory for improved prompt caching
  • implement CachingLLM for prompt caching and enhance execution tracking metrics
  • implement long-buy wins enrichment with trigger price and recommendation label
ProspectAI Module

v1.7.1

  • Maintenance and stability improvements
ProspectAI Module

v1.7.0

  • add risk_profile parameter (conservative / aggressive) flowing through the entire pipeline with per-profile allocation caps, stop-loss multipliers, and R/R ratios
  • PortfolioAllocatorTool applies hardcoded bounds per profile; Draft Strategist, Critic, and Final Strategist receive qualitative profile guidance
  • InvestorStrategicOutput and CriticOutput schemas carry risk_profile field
  • Backend /api/analyze exposes risk_profile query param; analytics tracks usage by profile
  • Web UI adds Conservative / Aggressive toggle before pipeline trigger