[SyedFahad7]

What does this PR do?

Fixes #4885.

Stacked on PR A (fix/4881-keep-weaker-evidence, #4881 / also_seen). This branch is cut from that branch and targets v8, so until PR A merges the diff also shows PR A's commit 3304de16. This PR's own change is the single commit 6e6e68a2. Review or merge PR A first.

graphify path and the MCP shortest_path tool label each hop separately. A path with one INFERRED or AMBIGUOUS hop among EXTRACTED hops therefore reads as a list of independent tags, and nothing marks the path as a whole as uncertain. Both now end with one summary line.

Repro from the issue (Report.cs: IReport.Build(), PdfReport/CsvReport implementing it, Exporter.Run() calling _report.Build(); graphify update .):

$ graphify path report_exporter_run report_pdfreport_build
Shortest path (2 hops):
  .Run() --calls [INFERRED]--> .Build() --dispatches_to [INFERRED]--> .Build()
Path confidence: INFERRED (weakest hop 2: dispatches_to, score 0.75)

The MCP shortest_path result ends with the same line.

How the summary is computed (_path_confidence_line in graphify/serve.py, also called by the CLI path in graphify/cli.py):

• A hop's strength is the weakest evidence it carries. That includes all of its parallel edges and, from PR A, its also_seen entries.
• Hops are compared by _CONFIDENCE_RANK (graphify/build.py) first, then by the lowest confidence_score. A missing score never counts as weaker than a present one. On a tie, the first such hop is named.
• , score X is printed only when the weakest evidence has a score.
• A hop with no recognised confidence tag makes the result Path confidence: unknown (hop N: <relation> has no confidence tag). The other hops cannot vouch for it.
• The shortest_path tool description and the README sentence about graphify path mention the new line.

Type of change

• Bug fix
• New feature
• Documentation
• Tests or CI
• Refactor
• Security fix

Verification & Invariants

Invariant: a path is never reported as more certain than its least certain hop (CONTRIBUTING, "Keep inference visibly separate from extracted fact"). The existing hop segments are byte-identical. The only output change is one extra line after them, so anything matching Shortest path (N hops): or hop substrings is unaffected. Nothing is persisted, so no graph or cache state changes.

• Read the CONTRIBUTING.md guide.
• Reproduced the issue and identified the invariant.
• Made the smallest fix necessary.
• Added a regression test (if bug fix) or isolated boundary test.
• Kept the PR description synchronized with the final implementation.
• Documented any limitations / unsupported cases explicitly.

Limitations:

• The summary covers only the single shortest path that path returns. Comparing several routes to the same target is not attempted.
• [Feature]: print confidence_score in path and explain #4882 (PR fix(serve): print confidence_score in path, explain and neighbors output #4888, scores printed in hop tags) edits the same confs = sorted(...) hop-tag line in serve.py and cli.py. Whichever lands second has a trivial one-line merge per file. This PR's own additions (hop_datas, the summary line, the helper) are outside those lines.

How was this tested?

tests/test_path_confidence.py (7 tests) covers:

• the exact repro line through the CLI and through the MCP _shortest_path_text;
• tier taking precedence over score;
• the lowest score breaking a tier tie, and a missing score not counting as weaker;
• all-EXTRACTED paths naming the first hop;
• parallel edges and also_seen counting toward their hop;
• a hop with no tag making the path unknown.

On PR A's commit, without this change, the module fails (_path_confidence_line does not exist), and the repro output has no summary line.

uv run pytest tests/test_path_confidence.py tests/test_edge_also_seen.py tests/test_path_cli.py tests/test_serve.py -q   # 215 passed
uv run pytest tests -q                               # 8983 passed, 17 skipped, 1 xfailed
uv run ruff check .                                  # All checks passed
uv run pyright                                       # 667 errors, identical to the upstream v8 baseline (no new ones)
graphify update . && graphify path report_exporter_run report_pdfreport_build   # output above

Graphify-specific checklist

• I updated generated skill artifacts (uv run python -m tools.skillgen --bless) when changing their source fragments. (No fragments touched.)
• I confirmed that AST/structural extraction remains deterministic (no ambient state dependencies like ENV variables).
• I reviewed changes for security implications (no unsafe interpolation into shell/Python).
• I confirmed no API keys or local-only graph data are included.
• (If applicable) I disclosed AI authorship in my commit messages.

  

Stacked on #4891 — review/merge that first; this PR includes its commit.

Update (follow-up commit 659e7db)

• Commits in this PR: 6e6e68a2 (this change) and 659e7db9 (follow-up), on top of fix(build): keep a collapsed edge's weaker evidence in also_seen (#4881) #4891.
• EXTRACTED hops no longer print a score (and their stored score is ignored when picking the weakest hop); NaN/infinite scores are treated as missing. This matches fix(serve): print confidence_score in path, explain and neighbors output #4888's tag format.
• Tests: 10 in tests/test_path_confidence.py (was 7); targeted run 218 passed; full suite 8986 passed, 17 skipped, 1 xfailed.

[github-actions]

Thanks for the pull request, @SyedFahad7. A maintainer will review it soon.

Want to talk it through while it is in review? Come join us on our Discord server. For longer-form discussion there is also GitHub Discussions.

A couple of things that speed up review: make sure the test suite passes on Python 3.10 and 3.13, and that the change keeps extraction deterministic.

[graphify-labs]

Graphify reviewed this change.

No issues found by static checks — no coupling regressions or blocking issues in the code graph. That is not the same as safe to merge: see what was not checked.

Not checked: tests were not run; formal verification proved 0 of 4 changed function(s) (4 not verified).

Formal verification. PR-changed functions: 0/4 verified (0 proven, 0 may-equivalent, 0 distinguished) · 4 not verified (4 vacuous).

Not verified on this run: build\_from\_json (vacuous: never exercised), dispatch\_command (vacuous: never exercised), \_build\_server (vacuous: never exercised), \_shortest\_path\_text (vacuous: never exercised).

---

Graphify review — findings

Adds an overall confidence line to graphify path and the shortest_path MCP tool, reporting the path's weakest hop (e.g. Path confidence: INFERRED (weakest hop 2: calls, score 0.7)). A hop with no recognised confidence tag makes the whole path report as unknown. When two edges with the same relation between the same nodes collapse, the lower-confidence copy is no longer thrown away: the winner keeps the best one and records the others in a new also_seen list (confidence, score, source file and location, weakest first). Re-feeding a written graph.json through the build is idempotent, and graphify path now shows the tags of every copy on a hop, such as [AMBIGUOUS/INFERRED]. README and docs/how-it-works.md describe the new line and the also_seen edge field.

No blocking issues surfaced. 3 lower-confidence candidates did not survive cross-model review.

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 2342 functions depend on the 427 functions this change touches.

Health — this change adds coupling hotspots:

• worse: build_from_json() — 283 callers, 21 callees
• worse: dispatch_command() — 5 callers, 137 callees
• worse: main() — 117 callers, 3 callees
• worse: _shortest_path_text() — 17 callers, 5 callees

Verification — 2342 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: 2128 function(s) in the blast radius were not formally verified this run

Test selection

Test selection

471 of 471 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/pi.test.ts — full-run-safety
• tests/test_affected_cli.py — impact, full-run-safety
• tests/test_affected_defines.py — full-run-safety
• tests/test_affected_member_seed.py — full-run-safety
• tests/test_affected_overrides.py — full-run-safety
• tests/test_affected_path_file_node.py — impact, full-run-safety
• tests/test_agents_platform.py — impact, full-run-safety
• tests/test_alias_folding.py — impact, full-run-safety
• tests/test_analyze.py — impact, 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 — impact, full-run-safety
• tests/test_atomic_version_stamp.py — full-run-safety
• tests/test_atomic_writes.py — impact, full-run-safety
• tests/test_backend_env_isolation.py — full-run-safety
• tests/test_backend_extras.py — full-run-safety
• tests/test_backend_opt_in.py — impact, full-run-safety
• tests/test_benchmark.py — impact, full-run-safety
• tests/test_benchmark_raw_graph.py — impact, full-run-safety
• tests/test_benchmark_real_query.py — impact, full-run-safety
• tests/test_binary_media_words.py — full-run-safety
• tests/test_blade_extractor.py — full-run-safety
• tests/test_build.py — impact, full-run-safety
• tests/test_build_located_semantic_identity.py — impact, full-run-safety
• tests/test_build_merge_dedup_scope.py — impact, full-run-safety
• tests/test_build_merge_hyperedges_and_prune.py — impact, full-run-safety
• tests/test_build_merge_shrink_guard.py — impact, full-run-safety
• tests/test_builtin_global_type_refs.py — full-run-safety
• tests/test_c_include_resolution.py — impact, full-run-safety
• tests/test_cache.py — full-run-safety
• tests/test_cache_long_paths.py — full-run-safety
• tests/test_cache_stale_import_target.py — full-run-safety
• tests/test_callflow_html.py — full-run-safety
• tests/test_cargo_introspect.py — full-run-safety
• tests/test_cargo_missing_manifest.py — full-run-safety
• tests/test_carried_hyperedge_remap.py — impact, full-run-safety
• tests/test_case_sensitive_resolution.py — full-run-safety
• tests/test_charmap_encoding.py — impact, full-run-safety
• tests/test_chunking.py — impact, full-run-safety
• tests/test_cjs_module_extension.py — full-run-safety
• tests/test_claude_cli_backend.py — impact, 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 — impact, full-run-safety
• tests/test_cli_help.py — full-run-safety
• tests/test_cloud_cta.py — impact, full-run-safety
• … and 421 more

non-code file(s) changed (README.md, docs/how-it-works.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 (README.md, docs/how-it-works.md) — 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.

Formal verification

Could not verify: Could not verify build\_from\_json.

The verifier did not have enough to check build\_from\_json, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: no capturable inputs from the test suite; property tier: not verifiable: all 11 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous

Could not verify: Could not verify dispatch\_command.

The verifier did not have enough to check dispatch\_command, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: no capturable inputs from the test suite; property tier: not verifiable: all 40 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly SystemExit — names the real obstacle, not a sampling gap)

Could not verify: Could not verify \_build\_server.

The verifier did not have enough to check \_build\_server, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: no capturable inputs from the test suite; property tier: not verifiable: all 40 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly ValueError — names the real obstacle, not a sampling gap)

Could not verify: Could not verify \_shortest\_path\_text.

The verifier did not have enough to check \_shortest\_path\_text, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: no capturable inputs from the test suite; property tier: not verifiable: all 35 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly KeyError — names the real obstacle, not a sampling gap)

· 1 grounded finding(s) anchored inline below; 3 more finding(s) on lines outside this diff (see the check run).

[graphify-labs]

⚠️ Health regression — _shortest_path_text()

17 callers depend on it (afferent coupling).

Grounded coupling-delta finding (deterministic), not an LLM guess.

[safishamsi]

Landed in v0.9.98, now on PyPI (pip install graphifyy==0.9.98). Landed via cherry-pick (your CI matrix was gated). path reports the overall confidence as the weakest hop. Thanks for the fix, @SyedFahad7.