Skip to content

Releases: partarstu/agentic-qa-framework

v2.1.0

14 May 22:46

Choose a tag to compare

What's New

New Agent Skills

  • PR Review — automated pull request review with configurable criteria
  • Run Unit Tests — skill for executing and reporting on unit test suites
  • Software Architect — skill for producing ADRs, decision matrices, and implementation plans

Retry Logic

  • Exponential backoff retry logic added for resilient agent operations
  • Increased embedding timeouts for improved stability

Turbo Mode

  • Turbo mode enabled in skills with restricted redirection operators for safety

Centralized Model Settings

  • Refactored agent creation to centralize model settings handling, reducing duplication

Improvements

  • Enhanced prompts for test case generation and acceptance criteria extraction
  • Dashboard improvements: scrolling fix and UX enhancements
  • Expanded logging for better observability
  • CI/CD: auxiliary service deployments are now conditional, significantly reducing build times
  • Added _TIMEZONE substitution support in cloudbuild.yaml

Bug Fixes

  • Fixed incidents missing attachments
  • Fixed agent wait behavior
  • Fixed DB connectivity issues
  • Fixed incorrect error message on dashboard login screen
  • Various stability, linting, and test assertion fixes

2.0.0

31 Jan 16:24
c4f24e7

Choose a tag to compare

These release notes cover the changes since v1.0.0, including the introduction of the Incident Reporter, Orchestrator Dashboard, and significant infrastructure improvements.

Suggested Version: v2.0.0 (due to breaking changes)

🚀 Key New Features

Incident Reporting & RAG System

A new suite of agents and services has been added to automate incident creation and management:

  • Incident Creation Agent: Automatically creates Jira bug reports for failed tests.
  • Duplicate Detection: Uses RAG (Retrieval-Augmented Generation) to check for existing similar issues before creating new ones.
  • Jira RAG Update Agent: Syncs Jira issues to the vector database for semantic search.
  • Embedding Service: A dedicated microservice using SentenceTransformer models for text embeddings.

Orchestrator Web UI

A comprehensive monitoring dashboard (orchestrator/ui) has been introduced:

  • Dashboard: Real-time visibility into agent status (AVAILABLE, BUSY, BROKEN), task history, and logs.
  • Agent Management: Broken agent classification (OFFLINE, TASK_STUCK) with auto-recovery mechanisms.
  • Security: JWT-based authentication for dashboard access.
  • Tech Stack: Built with React, TypeScript, and Vite.

Prompt Guard Service

  • New Service: services/prompt_guard_service offloads prompt injection detection models from the orchestrator and agents.
  • Performance: Improved architecture by separating heavy ML checks from the core logic.

Developer Skills

New agent skills have been added to the .agent/skills directory to assist developers:

  • adding-orchestrator-workflow
  • creating-new-agent
  • prepare-pr
  • writing-unit-tests

🏗 Infrastructure & CI/CD

  • CI Pipeline: Added .github/workflows/ci.yml for automated linting (ruff), testing (pytest), and security scanning (bandit, pip-audit).
  • Cloud Build: Optimized cloudbuild.yaml with conditional deployment logic for auxiliary services.
  • Vector DB: VectorDbService now strictly uses EMBEDDING_SERVICE_URL and communicates via gRPC (default port 6334) for improved performance.

⚠️ Breaking Changes

  • Mandatory Environment Variables:
    • PROMPT_GUARD_SERVICE_URL is now required if prompt injection checks are enabled.
    • EMBEDDING_SERVICE_URL is now mandatory; local embedding fallback has been removed.
  • Ports: QDRANT_GRPC_PORT defaults to 6334.

🛠 Refactoring & Improvements

  • Strict Artifact Parsing: Replaced ad-hoc parsing with strict Pydantic model validation (_get_model_from_artifacts) in the Orchestrator.
  • Resilience: Implemented retry logic with exponential backoff for VectorDbService.
  • Code Quality: Added ruff.toml for consistent linting and styling.
  • Dependencies: Cleaned up requirements.txt by moving heavy ML libraries to service-specific Dockerfiles.

🐛 Bug Fixes & Optimizations

  • Refined system prompts for better test case generation and coverage analysis.
  • Fixed logic in agent reservation and waiting intervals.
  • Corrected various minor issues in README.md and configuration files.

Primary Release

01 Nov 10:37
56ce5ac

Choose a tag to compare

Primary release with functional version of the framework.