docs: worked example — Chilean civil-law corpus (code + Spanish legal doctrine, AST-only)

[elpabloultron]

docs: worked example — Chilean civil-law corpus (code + Spanish legal doctrine, AST-only)

A worked example that follows the contribution guide: a real corpus, the output saved under
worked/civil-law-chile/, and an honest review.md covering what the graph got right and wrong.

Corpus: 4 Python modules + 4 Markdown legal documents (a doctrine index and three Chilean
Academia Judicial training guides) — 8 files, ~3,010 words, 52 KB. Half of it is Spanish prose
doctrine with accents, legal citations and emoji-prefixed headings, which is not what most
corpora in here look like.

Run: graphify update . on 0.9.63, AST-only path — no LLM, no API key, 0 tokens.
Result: 72 nodes, 91 edges, 9 communities, 100% EXTRACTED.

What the review reports, including the uncomfortable parts:

• The Markdown structure parse is more useful than I expected. The four legal documents
  contribute 21 document nodes and 17 rationale nodes with no LLM involved, and three of the
  ten god nodes are legal guides. For a prose-heavy corpus that is real navigation at zero cost.
• Cross-language retrieval works: the query token savings (English) started the traversal at
  the Spanish heading ⚡ 3. Estándar de Optimización de Tokens.
• But substance is out of reach on this path. Queries phrased the way a lawyer asks
  (plazo para contestar la demanda en el juicio oral laboral) reach the correct document, never
  the rule. The summary line "100% EXTRACTED" can be read as "the corpus is in the graph" when the
  prose was only taken to heading level.
• The corpus-size verdict looks inconsistent with a sibling example: this 3,010-word corpus is
  reported as "large enough that graph structure adds value", while worked/mixed-corpus reports
  the opposite warning at ~4,020 words.
• Smaller notes: emoji prefixes end up verbatim in heading labels, node ids keep accents and the
  full heading text, and "Surprising Connections" surfaced nothing surprising on a corpus this
  size.

The corpus is a slice of Open Legal Chile (Apache-2.0), trimmed to keep the example small — the
full corpus is 58 documents. The evaluation was executed by an AI agent (Hermes Agent) in a live
session; every number in review.md comes from the run whose graph.json and GRAPH_REPORT.md
are included.

[elpabloultron]

Ran the suite as the workflow asks (uv run pytest tests/ -q) before/while opening this, and
reporting the result since it is not what the guide implies:

15 failed, 5626 passed, 97 skipped in 64.71s

This PR introduces none of them. I checked the baseline on a clean v8 (same clone, my commit
not applied — 0 files under worked/): 15 failed, 5626 passed, 97 skipped, and the failing test
ids are byte-for-byte the same set:

• tests/test_ollama_retry_cap.py — 4 failures (test_ollama_defaults_to_zero_sdk_retries,
  test_ollama_honors_explicit_max_retries, test_cloud_backend_keeps_default_retries,
  test_api_timeout_is_passed_to_client)
• tests/test_skillgen.py — 11 failures (test_audit_coverage_passes,
  test_audit_coverage_passes_for_codex_and_windows, test_all_progressive_hosts_check_and_audit_clean,
  test_monolith_roundtrip_passes_for_aider_and_devin, test_monoliths_change_only_sanctioned_lines,
  test_always_on_roundtrip_is_byte_faithful, test_audit_coverage_passes_for_every_split_host,
  test_audit_catches_an_induced_per_host_drop, test_audit_catches_a_dropped_non_allowlisted_heading,
  test_amp_audit_coverage_passes_against_its_own_v8, test_agents_audit_baseline_is_amps_v8_body)

So v8 is currently 15 tests from green on its own — flagging it in case that is unexpected on your
side. Nothing here touches code: the change is one new folder under worked/ plus its review.

[graphify-labs]

Graphify reviewed this change.

Worth a look — the grounded gate found no coupling regressions or blocking issues, but 5 advisory finding(s) below merit a look before merge.

---

Graphify review — findings

Adds a generated knowledge-graph artifact for the civil-law-chile corpus: GRAPH_REPORT.md summarizing 72 nodes/91 edges across 9 communities with god nodes, cross-community bridges, and knowledge gaps, plus the underlying graph.json node/edge data. Both files are new AST-extracted outputs covering the corpus's code modules (LegalGraphBuilder, ColdStartInterviewEngine, stats_tracker.py, config.py) and doctrine documents.

Worth a look

• safe_urlopen permits SSRF to loopback/private hosts — worked/civil-law-chile/raw/code/config.py:87 · Escalate · high
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• get_suite_adoption_metrics silently substitutes hardcoded 138 institutions on any DB error — worked/civil-law-chile/raw/code/stats_tracker.py:118 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• Telemetry defaults to enabled (opt-out) violating stated privacy posture — worked/civil-law-chile/raw/code/stats_tracker.py:148 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• check_configuration reports stale credentials after load_env_file — worked/civil-law-chile/raw/code/config.py:80 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
• Ollama health check ignores configured OLLAMA_HOST — worked/civil-law-chile/raw/code/config.py:55 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 91 functions depend on the 91 functions this change touches.

Health — grade A; no new coupling hotspots.

Verification — 91 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 91 function(s) in the blast radius were not formally verified this run

Test selection

Test selection

286 of 286 test file(s) selected (100%) via static blast radius.

Escalated to a full run for safety — the selection is not trustworthy on its own (see below). CI should run the whole suite.

• tests/test_affected_cli.py — full-run-safety
• tests/test_affected_member_seed.py — full-run-safety
• tests/test_agents_platform.py — full-run-safety
• tests/test_analyze.py — full-run-safety
• tests/test_anthropic_custom_endpoint.py — full-run-safety
• tests/test_antigravity_install.py — full-run-safety
• tests/test_apm_fallback_version.py — full-run-safety
• tests/test_architecture_doc.py — full-run-safety
• tests/test_astro_extraction.py — full-run-safety
• tests/test_astro_import_ids.py — full-run-safety
• tests/test_atomic_canvas_export.py — full-run-safety
• tests/test_atomic_version_stamp.py — full-run-safety
• tests/test_atomic_writes.py — full-run-safety
• tests/test_backend_env_isolation.py — full-run-safety
• tests/test_backend_extras.py — full-run-safety
• tests/test_benchmark.py — full-run-safety
• tests/test_benchmark_raw_graph.py — full-run-safety
• tests/test_build.py — full-run-safety
• tests/test_build_merge_dedup_scope.py — full-run-safety
• tests/test_build_merge_hyperedges_and_prune.py — full-run-safety
• tests/test_build_merge_shrink_guard.py — full-run-safety
• tests/test_builtin_global_type_refs.py — full-run-safety
• tests/test_cache.py — full-run-safety
• tests/test_callflow_html.py — full-run-safety
• tests/test_cargo_introspect.py — full-run-safety
• tests/test_carried_hyperedge_remap.py — full-run-safety
• tests/test_case_sensitive_resolution.py — full-run-safety
• tests/test_charmap_encoding.py — full-run-safety
• tests/test_chunking.py — full-run-safety
• tests/test_cjs_module_extension.py — full-run-safety
• tests/test_claude_cli_backend.py — full-run-safety
• tests/test_claude_md.py — full-run-safety
• tests/test_cli_broken_pipe.py — full-run-safety
• tests/test_cli_export.py — full-run-safety
• tests/test_cli_help.py — full-run-safety
• tests/test_cluster.py — full-run-safety
• tests/test_codebuddy.py — full-run-safety
• tests/test_community_hub_labels.py — full-run-safety
• tests/test_community_labels_skill.py — full-run-safety
• tests/test_confidence.py — full-run-safety
• tests/test_corrupt_graph_json.py — full-run-safety
• tests/test_cpp_nested_and_cli.py — full-run-safety
• tests/test_cpp_objc_cross_file_calls.py — full-run-safety
• tests/test_cpp_preprocess.py — full-run-safety
• tests/test_cross_extension_reexport_self_cycle.py — full-run-safety
• tests/test_cross_language_call_resolution.py — full-run-safety
• tests/test_cross_repo_external_call_guards.py — full-run-safety
• tests/test_cross_repo_member_calls.py — full-run-safety
• tests/test_cross_repo_shared_types.py — full-run-safety
• tests/test_csharp_call_site_generic_args.py — full-run-safety
• … and 236 more

non-code file(s) changed (worked/civil-law-chile/GRAPH_REPORT.md, worked/civil-law-chile/graph.json, worked/civil-law-chile/raw/doctrine/README.md, worked/civil-law-chile/raw/doctrine/aj_conciliacion_laboral.md, worked/civil-law-chile/raw/doctrine/aj_etica_judicial.md …) → running the full suite for safety (a code graph can't see config/fixture/data deps)

changed code file(s) with no mapped test (worked/civil-law-chile/GRAPH_REPORT.md, worked/civil-law-chile/raw/code/cold_start.py, worked/civil-law-chile/raw/code/config.py, worked/civil-law-chile/raw/code/grafo_vinculos.py, worked/civil-law-chile/raw/code/stats_tracker.py …) — a coverage gap or a missing link — running the full suite rather than only the selected tests

Selection is safe under the controlled-regression assumption; always-run tests + a periodic full run are the backstops. Advisory — it never changes the check verdict.