- Restore pabot-based integration_tests with conservative parallelism
(<=2 processes by default) and support PABOT_PROCESSES/--processes
- Remove the temporary CI debug dump from noxfile.py
- Add robot/discovery_common.resource to silence non-fatal warning
- Document the change in implementation_plan.md
- Avoid deleting the workspace when OPENAI_API_KEY is unset
by guarding cleanup and initializing CONTEXT_DIR
- Prevents downstream Robot suites from losing .nox venv and
resource files after the basic scientific paper suite is skipped
Three fixes targeting CI integration_tests failures (18 failures on e8aa5ac):
1. Merge duplicate *** Settings *** blocks in 14 robot files into single
blocks. Multiple Settings sections are non-standard RF practice and
may cause resource import failures in certain Robot Framework versions
or CI environments.
2. Replace bare 'python' with ${PYTHON} variable in all Run Process
calls (14 files). Noxfile now passes --variable PYTHON:<venv-path>
to robot so tests use the venv interpreter regardless of PATH. This
fixes '/usr/local/bin/python: No module named cleveragents' on CI.
3. Add comprehensive CI debug output in noxfile.py: file existence
checks for .resource files, PATH/Python resolution, fixture dir
checks, and RF version. This will diagnose any remaining resource
import issues.
Also: remove hardcoded '/app/src' sys.path.insert in
system_prompt_template_rendering.robot (not portable to CI), and
add trailing newline to common.resource.
All 204 tests pass locally (4 excluded: 2 slow, 2 discovery).
- Restore session.env["PATH"] in integration_tests nox session to ensure
Run Process uses venv Python instead of system Python
- Convert bare Resource references to ${CURDIR}/ absolute paths across
30 robot files to fix CI resolution failures
- Add timeout=30s to all Run Process calls in rxpy_route_validation.robot
to prevent hanging tests
- Tag 2 rxpy tests as slow (require running actors unavailable on CI)
- Fix LangGraph test to use correct config file (LANGGRAPH_CONFIG)
- All 204 tests pass (4 excluded: 2 slow + 2 discovery)
- integration_tests: replace pabot with sequential robot execution to
eliminate FileNotFoundError caused by subprocess/FD exhaustion in
constrained CI containers (pabot spawns a robot subprocess per suite;
after ~24 suites the container cannot execve new processes)
- integration_tests: add resource debug output (open FD count, ulimit
values) to help diagnose future CI container issues
- integration_tests: remove _pabot_parallel_args (no longer needed);
slow_integration_tests session still available for parallel runs
- load_context_test: add env:TERM=dumb alongside NO_COLOR=1 to disable
all Rich terminal styling (NO_COLOR only disables color, not bold/
reset ANSI codes that may split substrings)
- load_context_test: add repr() debug logging around the --load-context
match to reveal any invisible characters on CI
- security_scan: create build/ directory before bandit writes its JSON
report (fails on fresh CI checkout where directory does not exist)
- integration_tests: cap pabot parallelism to 2 processes and explicitly
propagate venv bin/ to PATH, preventing FileNotFoundError for the
robot binary under CI resource constraints
- integration_tests: add env:NO_COLOR=1 to load_context_test.robot help
text assertions so Rich ANSI escape codes do not break substring
matching on CI
- integration_tests: tag initial_next_command_test as slow (requires
OPENAI_API_KEY for LLM agent invocation, unavailable on CI)
- Fix Rich Console line-wrapping breaking assertions in
context_unit_tests_steps.py: collapse newlines before checking for
filenames and overflow summaries (CI temp paths exceed 80 columns)
- Fix features.mocks import failure in database_integration.robot:
replace hardcoded sys.path '/app' with portable ${CURDIR}/..
- Fix --load-context help text assertions in load_context_test.robot:
merge stderr into stdout via stderr=STDOUT and remove duplicate test
- Add standalone dead_code nox session running vulture directly
- Rename security nox session to security_scan to match CI references
- Restore --exclude discovery to integration_tests nox session (lost
during merge conflict resolution)
- Rewrite .forgejo/workflows/ci.yml to route all jobs through nox sessions
- Fix coverage_report nox session: serial behave mode replaces broken parallel
mode (22% -> 97% accuracy), raise fail-under from 85% to 97%
- Pass posargs through format nox session for CI --check support
- Add 11 CI workflow validation scenarios (Behave) + Robot smoke test + ASV bench
- Add 108 new Behave scenarios covering 6 largest coverage gaps to reach 97%:
yaml_template_engine, actor/config, actor/registry, message_router,
context_analysis, context_service
- Update docs/development/ci-cd.md with nox-based CI docs and 97% threshold
- Restore implementation_plan.md verbose style, check off completed CI tasks
Verified: 1673 scenarios pass, 97% coverage, lint clean, typecheck clean
behave-parallel pulls behave>=1.2.6 which resolves to 1.3.3. Behave
1.3.3 no longer strips trailing colons from step text before tables,
breaking the architecture.feature step matcher.
Forgejo act_runner service container networking fails with 'network not
found'. Use docker:dind directly with --privileged and start dockerd
inline to avoid the services networking issue.
- Add DinD service container and docker:cli image so docker CLI is available
- Fix Dockerfile: remove missing uv.lock, fix build order (copy source before build)
- Remove helm job (k8s/ directory does not exist yet)
Forgejo shows the following error: ::error::@actions/artifact v2.0.0+, upload-artifact@v4+ and download-artifact@v4+ are not currently supported on GHES.
brent.edwards
CoreRasurae
freemo
