feat(substack): upgrade to SDK 2.0.0 with unit + integration tests

[Tram-Nguyen87]

Summary

Migrates the Substack integration to autohive-integrations-sdk~=2.0.0.

Substack's public API requires no authentication, so this integration has no credentials, no auth config, and no error-signalling action returns — it relies on context.fetch raising HTTPError/RateLimitError on non-2xx (the SDK v2 behaviour).

Changes

Source (substack.py)

• FetchResponse unwrap — access the body via .data at all 5 context.fetch call sites: get_publication_posts, get_post, search_publications, search_posts, get_post_comments. The two dict-guarded sites assign body = response.data first so isinstance(...) checks the body, not the FetchResponse.
• No ActionError conversion — the integration has no ActionResult(data={"error": ...}) patterns; the 1.x source had no try/except either, so the no-error-handling stance is unchanged and consistent. Hardening is out of pure-migration scope.

config.json

• Version 1.0.0 → 2.0.0.
• Comment item fields made nullable (body, date, author_name, author_id, like_count, id). Comments are returned as raw API passthrough (unlike posts/publications, which _drop_none cleans). The live API returns body: null on deleted comments — without nullable types, SDK v2 output validation recurses into the array items and raises VALIDATION_ERROR on a successful 200. Confirmed against the live API, not speculative.

requirements.txt

• Pin autohive-integrations-sdk~=2.0.0.

Tests

• Removed the legacy unittest suite (tests/context.py, tests/test_substack.py).
• Unit (test_substack_unit.py) — 38 tests: every action, FetchResponse mocks, _drop_none null-field handling, validation errors, and a regression test (test_null_comment_fields_pass_output_validation) reproducing the null-body case the schema fix addresses.
• Integration (test_substack_integration.py) — 14 tests, live, no auth: every action against a real public publication (default https://www.astralcodexten.com, overridable via SUBSTACK_TEST_PUBLICATION_URL), plus a bad-slug test asserting HTTPError propagates so the real_fetch raise path is actually exercised. real_fetch returns FetchResponse and raises HTTPError/RateLimitError on non-ok per repo convention.

Verification (run on the final commit, efbbf02)

Check	Result
validate_integration.py	✅ 0 errors, 0 warnings
check_code.py (lint, format, bandit, pip-audit, config-sync, fetch patterns)	✅ CODE CHECK PASSED
Unit tests	✅ 38 passed
Integration tests (live)	✅ 14 passed, 1 skipped

Self-review

Reviewed against the upgrading-sdk-v2, writing-unit-tests, and writing-integration-tests skills, plus an independent fresh-context review of the branch diff. The comment-nullability bug and the missing error-path test coverage were both found and fixed pre-emptively during that review rather than left for a reviewer to flag. The 1 skip is expected (get_post_comments skips when a post has no numeric id).

[github-actions]

🔍 Integration Validation Results

Commit: f89226f5e16b693b3b1db206c599f527edca5811 · docs(substack): add test env vars to .env.example
Updated: 2026-06-22T21:37:34Z

Changed directories: substack

Check	Result
Structure	✅ Passed
Code	✅ Passed
Tests	✅ Passed
README	✅ Passed
Version	✅ Passed

✅ Structure Check output

Validating 1 integration(s)...

============================================================
Integration: substack
============================================================
✅ All checks passed!

============================================================
SUMMARY
============================================================
Integrations validated: 1
Total errors: 0
Total warnings: 0

✅ All validations passed!

✅ Code Check output

----------------------------------------
Checking: substack
----------------------------------------

📦 Installing dependencies...

🐍 Checking Python syntax...
   ✅ Syntax OK

📥 Checking imports...
   ✅ Imports OK

📄 Checking JSON files...
   ✅ JSON files OK

🔍 Linting with ruff...
   ✅ Lint OK

🎨 Checking formatting with ruff...
   ✅ Formatting OK

🔒 Scanning for security issues with bandit...
   ✅ Security OK

🛡️ Checking dependencies for vulnerabilities with pip-audit...
   ✅ Dependencies OK

🔗 Checking config-code sync...
   ✅ Config-code sync OK

🔄 Checking fetch patterns...
   ✅ Fetch patterns OK

========================================
✅ CODE CHECK PASSED
========================================

✅ Tests Check output

Integration   Tests  Coverage        Status
-------------------------------------------
substack     38/38      100%      ✅ Passed
-------------------------------------------
Total        38/38            ✅ All passed

✅ Tests passed: substack

✅ README Check output

========================================
✅ README CHECK PASSED
========================================

✅ Version Check output

✅ substack: 1.0.0 → 2.0.0 (major bump)

========================================
✅ VERSION CHECK PASSED
========================================

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: efbbf029bf

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[Tram-Nguyen87]

confirmed and fixed in 4c8360f.

tests/conftest.py was prepending the substack/ directory to sys.path, which makes substack.py importable as a top-level module and shadows the substack package. Reproduced in a clean interpreter: with substack/ on sys.path[0], import substack resolves to substack/substack.py (a module, not a package), so the package-style imports (from substack import substack, from substack.substack import _normalise_url) fail at collection.

Fix: removed the sys.path manipulation and rely on repo-root package resolution (the root conftest.py puts the repo root on sys.path, and --import-mode=importlib does not re-insert test dirs). This matches the circle integration, which uses the same __init__ re-export + package imports with no path hack. The conftest is kept (the validator requires tests/conftest.py) but now only documents the import strategy.

Verified after the fix: unit 38 passed, integration 14 passed / 1 skipped (live), validate_integration.py 0 errors, check_code.py passed.

[TheRealAgentK]

Requesting one small should-fix before merge:

substack/tests/test_substack_integration.py reads SUBSTACK_TEST_PUBLICATION_URL and SUBSTACK_TEST_POST_SLUG, but this PR does not add blank template entries for them in the root .env.example. Even though both are optional, the integration-test checklist asks that every env var read by integration tests is documented there.

Suggested addition:

# -- Substack --
# Optional: publication URL used by Substack live integration tests.
# SUBSTACK_TEST_PUBLICATION_URL=
# Optional: post slug used by Substack get_post live integration tests.
# SUBSTACK_TEST_POST_SLUG=

Everything else in the SDK v2 migration, tests, and validation looked good.

[TheRealAgentK]

LGTM now