v1.0.0

Phase 1 POC - Closed-Loop Outcome Intelligence

Features

• Executive dashboard with KPI overview
• Workflow path analysis
• Agent version comparison
• Root cause insights engine
• Trace/feedback/outcome ingestion APIs

Tech Stack

• FastAPI + PostgreSQL backend
• DuckDB analytics engine
• React + TypeScript + Tailwind frontend
• Docker + Docker Compose
• GitHub Actions CI/CD

v0.3.0-beta

Beta release. Analytics service with DuckDB, outcome attribution engine, and 100K sample data generator.

v0.2.0-alpha

Alpha 2. PostgreSQL database models and basic API endpoints.

v0.1.0-alpha

Initial alpha release. Project scaffold and architecture design.

v2.0.0 — AgentLoop Experimentation & Recommendation Platform

Phase 2 Release — What should we change next?

New Components

• Experimentation Engine: full A/B testing lifecycle
• Statistical Attribution: 6 hypothesis tests + Bayesian comparison
• Cohort Analytics: user segmentation with drill-down
• Recommendation Engine: 5 recommendation types, P0-P3 priority
• Regression Detection: automatic incident creation
• Version Intelligence: lineage trees, version diff
• KPI Impact Engine: business translation layer
• SDK Layer: Python + TypeScript packages

Infrastructure

• Docker Compose startup maintained
• CI/CD pipeline extended
• 500K session synthetic data generator
• Full documentation

8 Questions Answered

Which agent version best? ✓ | Which prompt best? ✓ | Which experiment won? ✓ | Which cohorts benefit? ✓ | Which regressions caused changes? ✓ | What to do next? ✓ | What business impact? ✓ | Highest ROI recommendation? ✓

v1.5.0-beta — Recommendation Engine & Regression Detection

Beta Release

• Recommendation Engine: deterministic actionable insights (no LLM)
• Regression Detection: auto-incidents for prompt/tool/workflow/agent regressions
• Version Intelligence: lineage tracking, diff analysis, trend detection
• Incident management API
• Root cause hypothesis generation

v1.2.0-alpha — Cohort Analytics & SDK

Alpha 2

• Cohort Analytics: user segmentation (new/returning, enterprise/SMB, region, version)
• Python SDK: agentloop-sdk with trace/outcome/feedback/experiment tracking
• TypeScript SDK: @agentloop/sdk with full type safety
• Cohort heatmaps and comparison views

v1.1.0-alpha — Experimentation & Statistical Engine

Alpha Release

• Experimentation Engine: create, start, pause, stop, archive experiments
• Statistical Attribution: Chi-Square, Fisher Exact, Welch T-Test, Mann-Whitney U, Bayesian comparison
• Experiment comparison with confidence intervals and effect sizes