Skip to content

Releases

Enterprise Agent Reliability Lab v1.0.1

Choose a tag to compare

@qwsvd qwsvd released this 07 Sep 05:25
01a0809

Enterprise Agent Reliability Lab v1.0.1

This backward-compatible public delivery patch adds:

  • a recruiter-facing project landing page at the production root
  • verified standard FastAPI API docs, ReDoc, health, and OpenAPI surfaces
  • Render PORT compatibility shared by the runtime and Docker health check
  • synchronized package and OpenAPI version metadata
  • public delivery regression coverage across Python 3.11-3.14 and the production container

All existing Agent, MCP, Skills, reliability, tracing, evaluation, and benchmark behavior remains intact.

The repository does not claim real-LLM or official external benchmark results; its demos and local evals remain deterministic and offline.

Enterprise Agent Reliability Lab v1.0.0

Choose a tag to compare

@qwsvd qwsvd released this 06 Sep 19:04
4ffc78f

Enterprise Agent Reliability Lab v1.0.0

Production-oriented reference project for building and evaluating reliable enterprise AI agents.

Highlights

  • FastAPI enterprise after-sales backend with persistent SQLite state
  • LLM tool-calling Agent runtime
  • Official MCP SDK server/client integration
  • Repository Agent Skills with progressive loading
  • Provider-neutral reliability controls
  • Bounded retries, timeouts, loop protection, and side-effect safety
  • OpenTelemetry tracing
  • Deterministic end-to-end Agent evaluations
  • Regression gates with structured metrics
  • External benchmark integration boundary with τ³-bench adapter
  • Production Docker image
  • GitHub Actions CI across Python 3.11, 3.12, 3.13, and 3.14
  • Deterministic integrated demonstration workflow

Verification

  • 100 pytest tests passed
  • 10/10 deterministic evaluation cases passed
  • Regression gate: PASS
  • 18 reliability tests passed
  • 3/3 integrated demo scenarios passed
  • 60/60 demo evaluation checks passed
  • GitHub Actions CI passed
  • Production container validation passed
  • pip check passed
  • Compile/import validation passed
  • Credential and generated-artifact scans clean

Main components

  • app/agent/ Agent runtime and tool calling
  • app/mcp_integration/ MCP integration
  • app/skills/ Agent Skills
  • app/reliability/ reliability controls
  • app/tracing/ OpenTelemetry tracing
  • app/evals/ deterministic Agent evaluation framework
  • app/benchmarks/ external benchmark integration
  • app/demo.py integrated demonstration
  • Dockerfile production container
  • .github/workflows/ci.yml continuous integration
    This is the first stable project release.