test(examples): LLM-mocked tests for guides and python (#3601)

[Photon079]

Description

Part of #3601. Adds LLM-mocked tests for examples/guides/ (14 scripts) and examples/python/ (1 script), so every guide and python example runs end to end in CI with no API key, no network, and deterministic output.

Built on the mocking harness from #3958. This branch is based on that PR, so until it merges the diff also shows its harness files (cognee/tests/utils/example_mock_llm.py, cognee/tests/utils/example_runner.py, cognee/tests/examples/conftest.py). Please review only cognee/tests/examples/{guides,python}/. I will rebase to a clean diff once #3958 lands.

What's in this PR

• cognee/tests/examples/guides/test_guides.py: one test per script in examples/guides/ (14).
• cognee/tests/examples/python/test_python.py: examples/python/ (1 script; memory_provenance_demo, references_example, and schema_inventory_demo moved to examples/demos/ and are covered by the demos batch).

Each test loads the real on-disk example via import_example(...) and awaits its entrypoint under the shared isolated_example_env fixture (mocked LLM + embeddings, per-test tmp_path isolation), asserting it runs to completion with no exception. No example logic is copied into the tests.

How it works

• The harness (Feat/mocked examples test suite #3958) patches the single choke point LLMGateway.acreate_structured_output and uses cognee's built-in MOCK_EMBEDDING=true, so every structured LLM call gets a deterministic response and the real tokenizer is preserved.
• isolated_example_env points cognee's data/system roots at a per-test tmp_path and prunes before and after, so tests do not share graph/vector/relational state.
• A failing example (import error, signature drift, runtime exception) fails its test loudly.

Verification

Keyless run (no LLM_API_KEY, no network): 12 passed, 1 skipped, 2 xfailed.

env -u LLM_API_KEY -u OPENAI_API_KEY LOG_LEVEL=CRITICAL \
  python -m pytest cognee/tests/examples/guides/ cognee/tests/examples/python/ -p no:logging -q

Acceptance Criteria (progress toward #3601)

• Every script under examples/guides/ and examples/python/ has a corresponding mocked test.
• The tests run with no real API key and are deterministic.
• Tests fail loudly when an example breaks.
• Every script under all of examples/ (follow-up folder PRs).
• CI wiring (follow-up PR).

Scope & honest notes

• 1 skipped: guides/s3_storage talks to a real s3:// bucket (requires_aws).
• 2 xfailed, both harness follow-ups I will raise on Feat/mocked examples test suite #3958:
  • guides/custom_graph_model: the mocked KnowledgeGraph leaves the example's optional used_in: List[Field] = None field as None, and cognee re-validates the extracted graph against the full model, which requires a list. Fix belongs in build_mock_response: populate list-typed fields with [].
  • guides/consolidate_entity_descriptions: cognee's consolidate memify pipeline raises KeyError('id') on the minimal mocked graph. Needs a richer default mock graph.

Related: #3601 (issue), #3958 (harness).

Type of Change

• Other: test coverage

Pre-submission Checklist

• I have tested my changes thoroughly before submitting this PR
• This PR contains minimal changes necessary to address the issue/feature
• My code follows the project's coding standards and style guidelines
• I have added tests that prove my feature works
• I have added necessary documentation
• All new and existing tests pass
• I have searched existing PRs to ensure this change hasn't been submitted already
• I have linked any relevant issues in the description
• My commits have clear and descriptive messages

DCO Affirmation

I affirm that all code in every commit of this pull request conforms to the terms of the Topoteretes Developer Certificate of Origin.

[github-actions]

Hello @Photon079, thank you for submitting a PR! We will respond as soon as possible.