v0.9.0
LatestMarch 18, 2026
Chaos injection & failure simulation
- NewF4 Chaos Engine — inject latency spikes, model timeouts, and rate-limit errors into any pipeline node
- NewFailure cascade visualization shows how a single node failure propagates downstream
- NewPre-built chaos profiles: "Black Friday traffic", "Model provider outage", "Embedding drift"
- ImprovedSimulation runs now complete 40% faster with optimized parallel node execution
- FixedFixed edge case where re-ranker nodes with custom scoring functions would silently drop results
v0.8.0
February 24, 2026
Export engine & team collaboration
- NewF5 Export Engine — generate PDF, CSV, and JSON reports from any simulation run
- NewTeam workspaces with shared pipelines, simulation history, and role-based access
- NewShareable simulation links with read-only access for stakeholders
- ImprovedDashboard redesign with faster load times and improved metric density
- FixedResolved memory leak in long-running simulations exceeding 10,000 requests
v0.7.0
January 30, 2026
Cost optimization engine
- NewF3 Cost Lens — real-time token-level cost tracking across every node in your pipeline
- NewCost forecasting: project spend at 10x, 50x, and 100x current traffic
- NewAutomatic detection of redundant LLM calls with suggested consolidation patterns
- ImprovedImproved latency profiling accuracy to sub-millisecond resolution
- ImprovedAdded support for Anthropic Claude 3, Mistral, and Cohere models in node configuration
v0.6.0
January 6, 2026
Latency profiling & bottleneck detection
- NewF2 Latency Profiler — flame-chart style visualization of per-node execution time
- NewAutomatic bottleneck detection highlights the slowest node in every simulation
- NewP50 / P95 / P99 latency breakdowns with historical comparison
- ImprovedPipeline canvas now supports zoom, pan, and node grouping
- FixedFixed YAML parser failing on multi-line prompt templates
v0.5.0
December 10, 2025
Initial beta — pipeline builder & simulation
- NewF1 Pipeline Builder — drag-and-drop canvas for constructing AI system architectures
- NewNode types: LLM, Embedding, Vector Search, Re-ranker, Custom Function
- NewBasic simulation engine with configurable traffic patterns and request volumes
- NewCLI tool with init, simulate, and export commands
- NewFree tier with 3 pipelines, 500 simulation runs per month