Export the agent e2e suite as a release bundle for downstream repos

[ling-senpeng13]

What

Packages the agent e2e suite (e2e/) into a self-contained, version-stamped tarball — conductor-ai-e2e-python-<version>.tar.gz — attached to every GitHub release, so downstream repos (e.g. orkes-io/orkes-conductor) can pin the e2e suite to the exact python-sdk release they run against. Replaces the agentspan-sdk-e2e-python-* bundles formerly cut from agentspan-ai/agentspan. Mirrors conductor-oss/java-sdk#128 and conductor-oss/javascript-sdk#134.

• scripts/package-e2e-bundle.sh — stages e2e/ verbatim (suites import conductor.ai.agents by package) + requirements.txt pinning conductor-python[agents]==<version> from PyPI + run.sh (mirrors the agent-e2e.yml pytest invocation: -n 3 --dist=loadgroup, junit to results/junit-e2e.xml) + README.
• scripts/test-package-e2e-bundle.sh — static validator (file parity, py_compile, version pin, binary assets uncorrupted; no network).
• .github/workflows/release-agent-e2e-bundle.yml — packages, validates, sha256s, uploads on release: published. Handles this repo's no-v-prefix tags; pip's PEP 440 normalization makes ==2.0.0-rc2 resolve the PyPI artifact 2.0.0rc2.

Validation

• static validator green
• bundle venv-installs the published conductor-python[agents]==2.0.0-rc2 from PyPI; conductor.ai.agents imports
• full bundle run vs conductor-oss 3.32.0-rc.8 (real LLM calls, python 3.11): 133/134 executed tests pass (19 env-conditional skips, 1 xfail). The CLI-dependent tests need a current agentspan CLI (validated with the pinned 0.4.4, same as agent-e2e.yml — a stale local CLI fails them on retired endpoints).

Known issue surfaced (pre-existing, not introduced here)

test_suite11_langgraph.py::test_messages_state_detection fails deterministically with langgraph 1.2.9 (_input_is_messages is None). The [agents] extra floats langgraph >=1.1.3 and langchain forces >=1.2.5, so every fresh install gets a version where this fails — in-repo CI will hit it on its next cold pip resolve too. Also noted: the @credentials lifecycle tests hard-fail (FileNotFoundError) rather than skip when the CLI is absent, unlike the cli_skills suite which skips gracefully.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.