feat(relationship_detector): #387 POTENTIALLY_STALE writer via aelf doctor --detect-stale - #444
Merged
Conversation
Reviewer's GuideImplements the producer-side POTENTIALLY_STALE edge writer and wires it into Sequence diagram for aelf doctor --detect-stale executionsequenceDiagram
actor User
participant CLI as aelf_doctor_CLI
participant Doctor as _cmd_doctor
participant DetectStale as _cmd_doctor_detect_stale
participant RelConfig as RelationshipDetectorConfig
participant Store as MemoryStore
participant Writer as write_potentially_stale_edges
User->>CLI: aelf doctor --detect-stale [flags]
CLI->>Doctor: _cmd_doctor(args, out)
Doctor->>Doctor: check args.relationships
Doctor->>Doctor: check args.detect_stale
Doctor->>DetectStale: _cmd_doctor_detect_stale(args, out)
DetectStale->>RelConfig: load_relationship_detector_config()
DetectStale->>RelConfig: apply jaccard/confidence/max_pairs overrides
DetectStale->>Store: _open_store()
activate Store
DetectStale->>Writer: write_potentially_stale_edges(store, jaccard_min, residual_overlap_min, confidence_min, max_candidate_pairs)
activate Writer
Writer->>Writer: relationships_audit(store, thresholds)
Writer->>Writer: filter LABEL_CONTRADICTS pairs
Writer->>Writer: select pairs with score < confidence_min
loop for each sub_confidence pair
Writer->>Store: get_belief(a_id), get_belief(b_id)
Writer->>Writer: choose src,dst by created_at then id
Writer->>Store: get_edge(src, dst, EDGE_POTENTIALLY_STALE)
alt edge exists
Writer->>Writer: n_edges_skipped_existing++
else edge missing
Writer->>Store: insert_edge(Edge(src,dst,POTENTIALLY_STALE))
Writer->>Writer: n_edges_written++
end
end
Writer-->>DetectStale: PotentiallyStaleWriteReport
deactivate Writer
DetectStale->>Store: close()
deactivate Store
DetectStale->>DetectStale: format_write_report(report)
DetectStale->>CLI: print report to out
DetectStale-->>Doctor: exit code 0 or 1
Doctor-->>User: process exit
Class diagram for PotentiallyStaleWriteReport and related writer utilitiesclassDiagram
class PotentiallyStaleWriteReport {
+int n_pairs_audited
+int n_sub_confidence
+int n_edges_written
+int n_edges_skipped_existing
+int n_edges_skipped_self_pair
}
class RelationshipDetectorConfig {
+float jaccard_min
+float residual_overlap_min
+float confidence_min
+int max_candidate_pairs
}
class MemoryStore {
+get_belief(belief_id)
+get_edge(src_id, dst_id, edge_type)
+insert_edge(edge)
+close()
}
class Edge {
+str src
+str dst
+str type
+float weight
}
class RelationshipDetectorModule {
+write_potentially_stale_edges(store, jaccard_min, residual_overlap_min, confidence_min, max_candidate_pairs) PotentiallyStaleWriteReport
+format_write_report(report) str
+relationships_audit(store, jaccard_min, residual_overlap_min, confidence_min, max_candidate_pairs)
}
RelationshipDetectorModule ..> PotentiallyStaleWriteReport : returns
RelationshipDetectorModule ..> MemoryStore : uses
RelationshipDetectorModule ..> Edge : inserts
RelationshipDetectorModule ..> RelationshipDetectorConfig : thresholds
MemoryStore "*" o-- Edge : contains_edges
File-Level Changes
Assessment against linked issues
Possibly linked issues
Tips and commandsInteracting with Sourcery
Customizing Your ExperienceAccess your dashboard to:
Getting Help
|
|
Warning Rate limit exceeded
To keep reviews running without waiting, you can enable usage-based add-on for your organization. This allows additional reviews beyond the hourly cap. Account admins can enable it under billing. ⌛ How to resolve this issue?After the wait time has elapsed, a review can be triggered using the We recommend that you space out your commits to avoid hitting the rate limit. 🚦 How do rate limits work?CodeRabbit enforces hourly rate limits for each developer per organization. Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout. Please see our FAQ for further information. ℹ️ Review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (4)
📝 WalkthroughWalkthroughThis PR implements the ChangesPOTENTIALLY_STALE Edge Writer Implementation
Sequence DiagramsequenceDiagram
actor User
participant CLI as aelf doctor CLI
participant Store as MemoryStore
participant Detector as relationship_detector
participant Graph as Edge Graph
User->>CLI: aelf doctor --detect-stale
CLI->>Store: load store (with --relationships-* overrides)
CLI->>Detector: write_potentially_stale_edges(store, ...)
Detector->>Detector: relationships_audit() → contradicts pairs
Detector->>Detector: filter pairs where score < confidence_min
Detector->>Detector: sort by created_at (newer → src)
loop Each contradicting pair
Detector->>Store: check if POTENTIALLY_STALE edge exists
alt Edge not idempotent
Detector->>Store: insert (src, dst, POTENTIALLY_STALE, weight=1.0)
Detector->>Graph: edge written
else Already exists
Detector->>Detector: increment skipped count
end
end
Detector->>CLI: return PotentiallyStaleWriteReport
CLI->>User: format_write_report() → output (edges_written, skipped, etc.)
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
Comment on lines
+20
to
+23
| from aelfrice.relationship_detector import ( | ||
| LABEL_CONTRADICTS, | ||
| write_potentially_stale_edges, | ||
| ) |
![sourcery-ai[bot]](https://avatars.githubusercontent.com/in/48477?s=60&v=4){kind=small}
There was a problem hiding this comment.
Hey - I've found 1 issue, and left some high level feedback:
- In
_cmd_doctor,--relationshipsis checked before--detect-stale, so if a user passes both flags only the relationships path runs; consider either enforcing mutual exclusivity at argparse level or documenting the precedence explicitly to avoid confusion. - In
_cmd_doctor_detect_stale, the CLI overrides forrelationships_jaccard,relationships_confidence, andrelationships_max_pairsare cast directly to float/int without error handling; wrapping these conversions in a small validator that surfaces a cleanValueError(similar to the threshold errors you already handle) would provide a friendlier CLI failure mode for bad inputs.
Prompt for AI Agents
Please address the comments from this code review:
## Overall Comments
- In `_cmd_doctor`, `--relationships` is checked before `--detect-stale`, so if a user passes both flags only the relationships path runs; consider either enforcing mutual exclusivity at argparse level or documenting the precedence explicitly to avoid confusion.
- In `_cmd_doctor_detect_stale`, the CLI overrides for `relationships_jaccard`, `relationships_confidence`, and `relationships_max_pairs` are cast directly to float/int without error handling; wrapping these conversions in a small validator that surfaces a clean `ValueError` (similar to the threshold errors you already handle) would provide a friendlier CLI failure mode for bad inputs.
## Individual Comments
### Comment 1
<location path="src/aelfrice/cli.py" line_range="2694-2703" />
<code_context>
+ j_override = getattr(args, "relationships_jaccard", None)
+ c_override = getattr(args, "relationships_confidence", None)
+ mp_override = getattr(args, "relationships_max_pairs", None)
+ config = RelationshipDetectorConfig(
+ jaccard_min=(
+ float(j_override) if j_override is not None else config.jaccard_min
+ ),
+ residual_overlap_min=config.residual_overlap_min,
+ confidence_min=(
+ float(c_override)
+ if c_override is not None
+ else config.confidence_min
+ ),
+ max_candidate_pairs=(
+ int(mp_override)
+ if mp_override is not None
+ else config.max_candidate_pairs
+ ),
+ )
</code_context>
<issue_to_address>
**issue:** Threshold overrides can raise ValueError before the try/except, contradicting the documented exit semantics
Because `float(j_override)`, `float(c_override)`, and `int(mp_override)` run before the `try`, malformed CLI values (e.g. `--relationships-jaccard=foo`) raise `ValueError` that bypasses your handler and crash the command instead of exiting with 1 as documented. Please move the parsing / `RelationshipDetectorConfig` construction into the existing `try` or add a dedicated `try/except ValueError` that logs the same `aelf doctor --detect-stale: ...` message and returns 1.
</issue_to_address>Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.
Comment on lines
+2694
to
+2703
| config = RelationshipDetectorConfig( | ||
| jaccard_min=( | ||
| float(j_override) if j_override is not None else config.jaccard_min | ||
| ), | ||
| residual_overlap_min=config.residual_overlap_min, | ||
| confidence_min=( | ||
| float(c_override) | ||
| if c_override is not None | ||
| else config.confidence_min | ||
| ), |
There was a problem hiding this comment.
issue: Threshold overrides can raise ValueError before the try/except, contradicting the documented exit semantics
Because float(j_override), float(c_override), and int(mp_override) run before the try, malformed CLI values (e.g. --relationships-jaccard=foo) raise ValueError that bypasses your handler and crash the command instead of exiting with 1 as documented. Please move the parsing / RelationshipDetectorConfig construction into the existing try or add a dedicated try/except ValueError that logs the same aelf doctor --detect-stale: ... message and returns 1.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (1)
src/aelfrice/cli.py (1)
2690-2709: ⚡ Quick winExtract the shared config-resolution block into a helper to avoid drift.
The 20-line
load_relationship_detector_config()+RelationshipDetectorConfig(...)override block in_cmd_doctor_detect_stale(lines 2690–2709) is a verbatim copy of the identical block in_cmd_doctor_relationships(lines 2622–2641). If a new--relationships-*flag is introduced (e.g., aresidual_overlapoverride), it must be updated in both places.♻️ Proposed extraction
+def _resolve_relationship_detector_config( + args: argparse.Namespace, +) -> "RelationshipDetectorConfig": + from aelfrice.relationship_detector import ( + RelationshipDetectorConfig, + load_relationship_detector_config, + ) + config = load_relationship_detector_config() + j_override = getattr(args, "relationships_jaccard", None) + c_override = getattr(args, "relationships_confidence", None) + mp_override = getattr(args, "relationships_max_pairs", None) + return RelationshipDetectorConfig( + jaccard_min=( + float(j_override) if j_override is not None else config.jaccard_min + ), + residual_overlap_min=config.residual_overlap_min, + confidence_min=( + float(c_override) + if c_override is not None + else config.confidence_min + ), + max_candidate_pairs=( + int(mp_override) + if mp_override is not None + else config.max_candidate_pairs + ), + )Then in both handlers:
- config = load_relationship_detector_config() - j_override = getattr(args, "relationships_jaccard", None) - c_override = getattr(args, "relationships_confidence", None) - mp_override = getattr(args, "relationships_max_pairs", None) - config = RelationshipDetectorConfig( - jaccard_min=( - float(j_override) if j_override is not None else config.jaccard_min - ), - residual_overlap_min=config.residual_overlap_min, - confidence_min=( - float(c_override) - if c_override is not None - else config.confidence_min - ), - max_candidate_pairs=( - int(mp_override) - if mp_override is not None - else config.max_candidate_pairs - ), - ) + config = _resolve_relationship_detector_config(args)🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@src/aelfrice/cli.py` around lines 2690 - 2709, The shared config-resolution code for relationship detector overrides is duplicated between _cmd_doctor_detect_stale and _cmd_doctor_relationships; extract that logic into a helper (e.g., a new function like resolve_relationship_detector_config) that calls load_relationship_detector_config and applies overrides from args (relationships_jaccard, relationships_confidence, relationships_max_pairs, etc.) returning a RelationshipDetectorConfig, then replace the inline blocks in both handlers to call this helper so future flags only need one update.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@tests/test_relationship_detector_stale_writer.py`:
- Around line 185-217: The current tautological assertion in
test_writer_returns_report_counts should be replaced with a concrete equality:
assert that report.n_pairs_audited equals the expected total contradicting pairs
from the setup (one sub-confidence + one high-confidence = 2). In the test
function test_writer_returns_report_counts, replace the tautology involving
report.n_sub_confidence with a direct check assert report.n_pairs_audited == 2
(and keep or remove the existing >= check as optional) so the test actually
verifies the intended total audited count.
---
Nitpick comments:
In `@src/aelfrice/cli.py`:
- Around line 2690-2709: The shared config-resolution code for relationship
detector overrides is duplicated between _cmd_doctor_detect_stale and
_cmd_doctor_relationships; extract that logic into a helper (e.g., a new
function like resolve_relationship_detector_config) that calls
load_relationship_detector_config and applies overrides from args
(relationships_jaccard, relationships_confidence, relationships_max_pairs, etc.)
returning a RelationshipDetectorConfig, then replace the inline blocks in both
handlers to call this helper so future flags only need one update.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: e7657785-a59c-4bbf-8f77-9a80f7b131cd
📒 Files selected for processing (4)
docs/edge_rerank.mdsrc/aelfrice/cli.pysrc/aelfrice/relationship_detector.pytests/test_relationship_detector_stale_writer.py
Comment on lines
+185
to
+217
| def test_writer_returns_report_counts(store: MemoryStore) -> None: | ||
| """Mixed store: report counts add up correctly. | ||
|
|
||
| Store contains: | ||
| - one sub-confidence contradicting pair → 1 edge written | ||
| - one high-confidence contradicting pair → 0 edges (excluded) | ||
| - one unrelated pair → 0 edges | ||
| """ | ||
| # Sub-confidence pair (score 0.4). | ||
| _make_belief(store, belief_id="s1", content=_CONTENT_A, | ||
| created_at="2026-01-01T00:00:00Z") | ||
| _make_belief(store, belief_id="s2", content=_CONTENT_B, | ||
| created_at="2026-02-01T00:00:00Z") | ||
| # High-confidence pair (score 0.5 — NOT sub-confidence). | ||
| _make_belief(store, belief_id="h1", content=_CONTENT_HI_A) | ||
| _make_belief(store, belief_id="h2", content=_CONTENT_HI_B) | ||
| # Unrelated pair. | ||
| _make_belief(store, belief_id="u1", content="sphinx fts indexing latency") | ||
| _make_belief(store, belief_id="u2", content="compression ratio on disk") | ||
|
|
||
| report = write_potentially_stale_edges(store) | ||
|
|
||
| # Exactly 1 sub-confidence contradicts pair qualifies. | ||
| assert report.n_sub_confidence == 1 | ||
| assert report.n_edges_written == 1 | ||
| assert report.n_edges_skipped_existing == 0 | ||
| assert report.n_edges_skipped_self_pair == 0 | ||
| # Total contradicts audited = sub + high. | ||
| assert report.n_pairs_audited == report.n_sub_confidence + ( | ||
| report.n_pairs_audited - report.n_sub_confidence | ||
| ) | ||
| # n_pairs_audited must be >= n_sub_confidence. | ||
| assert report.n_pairs_audited >= report.n_sub_confidence |
There was a problem hiding this comment.
Tautological assertion tests nothing — replace with the intended concrete check.
Lines 213–215:
assert report.n_pairs_audited == report.n_sub_confidence + (
report.n_pairs_audited - report.n_sub_confidence
)This reduces to X == X + (X − X) → X == X, which is unconditionally true and provides zero test coverage. The comment above it ("Total contradicts audited = sub + high") reveals the intent: verify the total contradicting-pair count equals 2 (one sub-confidence + one high-confidence pair from the test setup). The second assertion on line 217 (n_pairs_audited >= n_sub_confidence) is similarly always true by construction.
🐛 Proposed fix
- # Total contradicts audited = sub + high.
- assert report.n_pairs_audited == report.n_sub_confidence + (
- report.n_pairs_audited - report.n_sub_confidence
- )
- # n_pairs_audited must be >= n_sub_confidence.
- assert report.n_pairs_audited >= report.n_sub_confidence
+ # 1 sub-confidence + 1 high-confidence (score==0.5 NOT < 0.5) = 2 total.
+ assert report.n_pairs_audited == 2
+ assert report.n_sub_confidence == 1🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@tests/test_relationship_detector_stale_writer.py` around lines 185 - 217, The
current tautological assertion in test_writer_returns_report_counts should be
replaced with a concrete equality: assert that report.n_pairs_audited equals the
expected total contradicting pairs from the setup (one sub-confidence + one
high-confidence = 2). In the test function test_writer_returns_report_counts,
replace the tautology involving report.n_sub_confidence with a direct check
assert report.n_pairs_audited == 2 (and keep or remove the existing >= check as
optional) so the test actually verifies the intended total audited count.
|
This PR is now behind Auto-rebase was removed because the bot has no signing key; rebasing as the bot strips author signatures and the |
Collaborator
Author
|
[claim:review:Toug:2026-05-05T17:37:45Z] |
…onfidence contradicts (#387) Add PotentiallyStaleWriteReport dataclass and write_potentially_stale_edges() function. For each contradicts pair with score < confidence_min, one POTENTIALLY_STALE edge is emitted src=newer→dst=older (created_at lex order, id tiebreak). Pre-insert get_edge check keeps the call idempotent. Also add format_write_report() helper. Update module docstring to reflect that the POTENTIALLY_STALE writer ships here; CONTRADICTS write-path remains R2-deferred.
Add --detect-stale argparse flag to aelf doctor. Dispatch routes to the new _cmd_doctor_detect_stale handler, which opens the store and calls write_potentially_stale_edges with the same --relationships-jaccard / --relationships-confidence / --relationships-max-pairs overrides shared with --relationships. Prints format_write_report output on success. Update _cmd_doctor_relationships docstring to note the CONTRADICTS write-path remains R2-deferred while POTENTIALLY_STALE ships via #387.
…ort counts (#387) Seven unit tests against a real MemoryStore(:memory:): sub-confidence pairs emit edges, high-confidence pairs are excluded, writer is idempotent, newer→older direction is correct, created_at tiebreak on lex id order, refines/unrelated pairs produce zero edges, report count fields are internally consistent.
Expand the Producer side section to name the CLI entry point (aelf doctor --detect-stale), describe the staleness signal (sub-confidence contradicting pairs, score < confidence_min), the edge direction rule (newer belief → older belief, created_at lex order + id tiebreak), idempotency guarantee, and the shared tuning flags.
robotrocketscience
force-pushed
the
feat/issue-387-potentially-stale-writer
branch
from
May 5, 2026 17:39
9f0295a to
68dafc0
Compare
yoshi280
removed
attn:review
Collaborator
Author
|
[release:review:Toug:2026-05-05T17:41:36Z] |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #387.
Summary
Wires the
POTENTIALLY_STALEedge writer intoaelf doctor, closingthe missing acceptance criterion #2 from #387. The substrate (#421)
shipped the edge-type constant, BFS skip-during-expansion pin, the
edge-type-keyed rerank consumer, and the bench-gate stub at
tests/bench_gate/test_edge_rerank_potentially_stale.py. This PRadds the producer side.
Acceptance check (#387)
EDGE_POTENTIALLY_STALEinmodels.py, BFS pin atBFS_EDGE_WEIGHTS[POTENTIALLY_STALE] = 0.0)aelf doctor --detect-staletests/bench_gate/test_edge_rerank_potentially_stale.pyfrom #421 runs lab-side (skips whenAELFRICE_CORPUS_ROOTunset, per directory-of-origin). Public CI verifies the stub's collection + skip path; lab-side run records the +1pp drop number in a follow-up comment on #387.docs/edge_rerank.md§ Producer sideDesign
Picked the staleness signal that the codebase had already pre-decided
(
relationship_detector.py:22-26,cli.py:2606-2608pre-PR docstrings):Staleness signal: sub-confidence (
score < confidence_min)LABEL_CONTRADICTSpairs fromrelationships_audit(). High-confidencecontradicting pairs are explicitly NOT touched — those belong to the
deferred
CONTRADICTSwrite-path hook (R2 in #201, still bench-gated).Edge direction: for each qualifying pair
(a, b), onePOTENTIALLY_STALE edge is emitted with
src = newer belief, dst = older belief. "Newer" is the belief whose(created_at, id)tuple islexicographically greater (ISO timestamps, tie-break on lex id order).
Semantics: the newer belief casts doubt on the older one.
Idempotent: a pre-insert
get_edge(src, dst, POTENTIALLY_STALE)check skips pairs whose edge already exists. Repeated invocations are
safe.
Stdlib only. No new dependencies. Determinism is preserved end-to-end:
relationships_auditalready sorts pairs by(belief_a_id, belief_b_id),and the writer iterates in that order.
CLI surface
The tuning flags share semantics with
--relationships. Models thewrite-mode shape after
--classify-orphans(the existing precedentfor write-mode doctor flags).
Sample output:
Test plan
7 new unit tests in
tests/test_relationship_detector_stale_writer.py:test_writer_emits_edges_for_sub_confidence_pairs— sub-confidencepair (always/rarely, score 0.4) yields exactly one edge.
test_writer_skips_high_confidence_pairs— negation-only pair(score exactly 0.5 ==
confidence_min) yields zero edges. Thefilter is strict less-than (
< confidence_min), so 0.5 ishigh-confidence. This boundary behaviour is intentional and the
test pins it.
test_writer_idempotent— second call writes 0 edges,n_edges_skipped_existingreflects the prior insertion.test_writer_direction_newer_to_older— newer (latercreated_at)is
src, older isdst. Reverse direction does NOT exist.test_writer_direction_tiebreak_on_id— identicalcreated_at,lex-greater id wins as
src.test_writer_skips_refines_and_unrelated— onlyLABEL_CONTRADICTSpairs are eligible.
test_writer_returns_report_counts— mixed store; counts add up.Local:
uv run pytest tests/ --ignore=tests/e2e -q→ 2508 passed, 41 skipped.Out of scope
CONTRADICTSwrite-path hook is unchanged (stillR2-deferred per Semantic contradiction detector (relationship_detector) — v2.0 evaluation #201).
aelf doctordoes not auto-run--detect-stale; the operator opts in.+1pp drop number is recorded by the lab-side run.
Files changed
src/aelfrice/relationship_detector.py—+150(newwrite_potentially_stale_edges,PotentiallyStaleWriteReport,format_write_report,__all__extension, docstring reconciliation).src/aelfrice/cli.py—+85(new_cmd_doctor_detect_stale,argparse flag, dispatch in
_cmd_doctor, docstring update).tests/test_relationship_detector_stale_writer.py—+217(new file).docs/edge_rerank.md—+24(Producer-side § expanded).Summary by Sourcery
Add a POTENTIALLY_STALE edge writer and expose it via the
aelf doctor --detect-staleCLI to emit edges for sub-confidence contradicting belief pairs.New Features:
write_potentially_stale_edgesand reporting helpers to generate POTENTIALLY_STALE edges from low-confidence contradicting relationships.aelf doctor --detect-stalecommand-line option to scan for stale beliefs and write POTENTIALLY_STALE edges with shared tuning flags from--relationships.Enhancements:
Tests:
Summary by CodeRabbit
New Features
aelf doctor --detect-stalecommand now identifies contradicting belief pairs and emits stale edge markersDocumentation
Tests