Repository navigation
Onboarding self serve - #68
Conversation
✅ Maida verdict: pass4 blocking checks passed | 3/3 trials passed Behavior vs baselineNo baseline comparison was configured for this run. Next steps
Passing checks, report-only metrics, and trial evidencePassing checks
Trial evidence
Gated by Maida -- the local-first behavioral regression gate for AI agents. Action merge decisionEvaluated PR head: Mode: blocking. GitHub conclusion: success. Only a behavioral PASS with accepted configuration and successful check publication can authorize this commit. Configuration: unchanged. Policy always comes from the trusted PR base. See the named check for revision and configuration hashes. Next safe actionReview the observed coverage and confirm the required checks belong to this PR head. Continue normal code review and correctness/security checks; this behavioral PASS covers the observed executions only. Reproducibility InstructionsThe From this PR checkout, record a fresh local run and rerun the same assertion inputs: python -m pip install maida-ai==0.5.3
MAIDA_REPRO_DIR=$(mktemp -d)
git show 0df5653ec9195e4486a731d8c7574ef0cddf8386:.maida/policy.yaml > "$MAIDA_REPRO_DIR/policy.yaml"
maida run examples/ci_dummy_agent.py --policy "$MAIDA_REPRO_DIR/policy.yaml"Then run |
Keep check, PR, and workflow summaries actionable across behavioral verdicts, configuration review, missing evidence, setup errors, and publication failures. Avoid repeated infeasible trials and only offer baseline acceptance after a failed behavioral report.
Add a read-only GitHub verifier tying acceptance provenance, baseline-only commit, dispatch attempt, current-head report, required status, sticky comment, and strict protection settings together. Fail on missing or stale evidence and explicitly distinguish configured checks from actual merge-refusal verification.
…kthrough Lead with the shared pinned CLI demo and progressive practical guide, retain Python integration as the secondary path, and separate legacy v5 from coordinated blocking and acceptance features.
…ndencies Render the current core generator with consumer entrypoint and baseline paths, execute gate and acceptance installation against an offline local wheel, and isolate each job's interpreter. Retain an explicit compatibility path for the older reviewed CI generator.
…onfig Exercise the generated gate and acceptance capture with a tools-only pyproject plus a real requirements dependency. This reproduced the missing-install regression before the coordinated core fallback fix.
Accept immutable vMAJOR.MINOR.PATCHrcN tags through the existing tested and attested archive pipeline. Publish candidates as prereleases without replacing the latest stable release, and reject missing release classification before publication. Raise the tooling lock and workflow example to Python 3.12 while preserving retained dependency versions. Cover stable and candidate archives, malformed tags, release flags, and supported interpreter bounds.
a97b09e to
8cba703
Compare
Docs and tests for simplified onboarding