feat(wonder): wonder_ingest + wonder_gc lifecycle (closes #229, #548) - #559
Merged
Conversation
Reviewer's GuideImplements the full lifecycle for speculative “wonder” phantoms: ingesting them as speculative beliefs with audit and relation edges, and garbage-collecting stale ones via a new soft-delete mechanism on beliefs, plus the necessary enums, schema changes, and tests. Sequence diagram for wonder_ingest speculative belief creationsequenceDiagram
participant WonderLifecycle as wonder_lifecycle
participant MemoryStore as memory_store
participant DB as sqlite_db
WonderLifecycle->>WonderLifecycle: compute_constituent_key(constituent_belief_ids)
WonderLifecycle->>MemoryStore: get_belief_by_content_hash(content_hash)
MemoryStore->>DB: SELECT * FROM beliefs WHERE content_hash = content_hash AND valid_to IS NULL
DB-->>MemoryStore: existing_belief_or_none
alt belief_exists
MemoryStore-->>WonderLifecycle: existing_belief
WonderLifecycle->>WonderLifecycle: increment skipped
else new_belief
MemoryStore-->>WonderLifecycle: None
WonderLifecycle->>WonderLifecycle: create Belief(type=speculative, origin=speculative, alpha=0.3, beta=1.0, retention_class=snapshot)
Note over WonderLifecycle: alpha=0.3, beta=1.0, retention_class=snapshot
WonderLifecycle->>MemoryStore: insert_belief(belief)
MemoryStore->>DB: INSERT INTO beliefs(..., type, origin, alpha, beta, retention_class, valid_to)
loop for each constituent_belief_id
WonderLifecycle->>MemoryStore: insert_edge(Edge(src=phantom_belief, dst=constituent, type=RELATES_TO))
MemoryStore->>DB: INSERT INTO edges(src, dst, type, weight)
end
WonderLifecycle->>WonderLifecycle: audit_meta = generator@score
WonderLifecycle->>MemoryStore: record_corroboration(belief_id, source_type=wonder_ingest, session_id, source_path_hash=audit_meta)
MemoryStore->>DB: INSERT INTO belief_corroborations(..., source_type, source_path_hash)
WonderLifecycle->>WonderLifecycle: increment inserted and edges_created
end
WonderLifecycle-->>WonderLifecycle: return WonderIngestResult(inserted, skipped, edges_created)
Sequence diagram for wonder_gc speculative belief garbage collectionsequenceDiagram
participant WonderLifecycle as wonder_lifecycle
participant MemoryStore as memory_store
participant DB as sqlite_db
WonderLifecycle->>WonderLifecycle: compute cutoff_ts = now - ttl_days
WonderLifecycle->>MemoryStore: query_wonder_gc_candidates(cutoff_ts, alpha_default=0.3, beta_default=1.0, alpha_epsilon, beta_epsilon)
MemoryStore->>DB: SELECT id FROM beliefs b WHERE type = speculative AND origin = speculative AND valid_to IS NULL AND created_at < cutoff_ts AND alpha <= alpha_default + alpha_epsilon AND beta <= beta_default + beta_epsilon AND NOT EXISTS feedback_history(b.id) AND NOT EXISTS RESOLVES_edges(b.id)
DB-->>MemoryStore: candidate_ids
MemoryStore-->>WonderLifecycle: candidate_ids
WonderLifecycle->>WonderLifecycle: scanned = len(candidate_ids)
alt dry_run
WonderLifecycle-->>WonderLifecycle: return WonderGCResult(scanned, deleted=0, surviving=scanned)
else non_dry_run
WonderLifecycle->>WonderLifecycle: now_ts = current_time
loop for each candidate_id
WonderLifecycle->>MemoryStore: soft_delete_belief(belief_id=candidate_id, ts=now_ts)
MemoryStore->>DB: UPDATE beliefs SET valid_to = now_ts WHERE id = candidate_id AND valid_to IS NULL
MemoryStore->>MemoryStore: _bump_belief_version(candidate_id)
MemoryStore->>DB: COMMIT
MemoryStore->>MemoryStore: _fire_invalidation()
end
WonderLifecycle-->>WonderLifecycle: return WonderGCResult(scanned, deleted=scanned, surviving=0)
end
Entity-relationship diagram for speculative belief GC predicateserDiagram
BELIEFS {
TEXT id
TEXT type
TEXT origin
REAL alpha
REAL beta
TEXT created_at
TEXT valid_to
}
FEEDBACK_HISTORY {
INTEGER id
TEXT belief_id
TEXT created_at
}
EDGES {
INTEGER id
TEXT src
TEXT dst
TEXT type
}
BELIEF_CORROBORATIONS {
INTEGER id
TEXT belief_id
TEXT source_type
TEXT source_path_hash
}
BELIEFS ||--o{ FEEDBACK_HISTORY : has_feedback
BELIEFS ||--o{ EDGES : participates
BELIEFS ||--o{ BELIEF_CORROBORATIONS : has_corroboration
%% Wonder GC selects from BELIEFS where
%% type = speculative
%% origin = speculative
%% valid_to IS NULL
%% created_at < cutoff_ts
%% alpha and beta at ingest defaults
%% and no related FEEDBACK_HISTORY
%% and no related EDGES of type RESOLVES
Class diagram for wonder lifecycle and belief soft-deleteclassDiagram
class Belief {
+str id
+str content
+str content_hash
+float alpha
+float beta
+str type
+int lock_level
+str locked_at
+float demotion_pressure
+str created_at
+str last_retrieved_at
+str session_id
+str origin
+float hibernation_score
+str activation_condition
+str retention_class
+str valid_to
}
class Edge {
+str src
+str dst
+str type
+float weight
}
class Phantom {
+str content
+tuple~str~ constituent_belief_ids
+str generator
+float score
}
class WonderIngestResult {
+int inserted
+int skipped
+int edges_created
}
class WonderGCResult {
+int scanned
+int deleted
+int surviving
}
class MemoryStore {
+insert_belief(belief_id, content, content_hash, alpha, beta, type, lock_level, locked_at, demotion_pressure, created_at, last_retrieved_at, session_id, origin, hibernation_score, activation_condition, retention_class, valid_to) void
+get_belief_by_content_hash(content_hash) Belief
+insert_edge(src, dst, type, weight) void
+record_corroboration(belief_id, source_type, session_id, source_path_hash) void
+soft_delete_belief(belief_id, ts) void
+query_wonder_gc_candidates(cutoff_ts, alpha_default, beta_default, alpha_epsilon, beta_epsilon) list~str~
}
class WonderLifecycleModule {
+wonder_ingest(store, phantoms, session_id) WonderIngestResult
+wonder_gc(store, ttl_days, dry_run) WonderGCResult
-float _INGEST_ALPHA
-float _INGEST_BETA
-_constituent_key(constituent_belief_ids) str
}
Belief "1" <.. "*" Edge : src_or_dst
Belief "1" <.. "*" Phantom : constituent
WonderLifecycleModule ..> Belief
WonderLifecycleModule ..> Edge
WonderLifecycleModule ..> Phantom
WonderLifecycleModule ..> WonderIngestResult
WonderLifecycleModule ..> WonderGCResult
WonderLifecycleModule ..> MemoryStore
MemoryStore o--> Belief
MemoryStore o--> Edge
class EnumsAndConstants {
<<enumeration>>
+str BELIEF_SPECULATIVE
+str EDGE_RESOLVES
+str EDGE_RELATES_TO
+str CORROBORATION_SOURCE_WONDER_INGEST
+str ORIGIN_SPECULATIVE
+str RETENTION_SNAPSHOT
+int LOCK_NONE
}
EnumsAndConstants ..> Belief
EnumsAndConstants ..> Edge
EnumsAndConstants ..> WonderLifecycleModule
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
You’ve run out of usage credits. Purchase more in the billing tab. ⌛ 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 (6)
✨ 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 |
robotrocketscience
added
author-Faraday
![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
query_wonder_gc_candidates, consider using the enum constants (e.g.,BELIEF_SPECULATIVE,ORIGIN_SPECULATIVE,EDGE_RESOLVES) instead of string literals ('speculative', 'RESOLVES') to avoid drift if the model constants ever change. - In
soft_delete_belief,_bump_belief_versionis called unconditionally even when the UPDATE affects zero rows; you could check the cursor rowcount and only bump/commit when a change actually occurred to avoid unnecessary invalidations. - The type hint for
storeinwonder_ingest/wonder_gcisMemoryStore, which is more specific than needed; consider annotating against a protocol or the broader store interface so these functions work cleanly with other store implementations.
Prompt for AI Agents
Please address the comments from this code review:
## Overall Comments
- In `query_wonder_gc_candidates`, consider using the enum constants (e.g., `BELIEF_SPECULATIVE`, `ORIGIN_SPECULATIVE`, `EDGE_RESOLVES`) instead of string literals ('speculative', 'RESOLVES') to avoid drift if the model constants ever change.
- In `soft_delete_belief`, `_bump_belief_version` is called unconditionally even when the UPDATE affects zero rows; you could check the cursor rowcount and only bump/commit when a change actually occurred to avoid unnecessary invalidations.
- The type hint for `store` in `wonder_ingest`/`wonder_gc` is `MemoryStore`, which is more specific than needed; consider annotating against a protocol or the broader store interface so these functions work cleanly with other store implementations.
## Individual Comments
### Comment 1
<location path="src/aelfrice/store.py" line_range="1507" />
<code_context>
+ SELECT b.id
+ FROM beliefs b
+ WHERE b.type = 'speculative'
+ AND b.origin = 'speculative'
+ AND b.valid_to IS NULL
+ AND b.created_at < ?
</code_context>
<issue_to_address>
**issue:** Use the ORIGIN_SPECULATIVE constant instead of hardcoding the origin string in the GC query.
Hardcoding `b.origin = 'speculative'` makes this query fragile if the origin wire value changes or differs from the literal. Using the existing constant (via interpolation or a parameter) keeps the query consistent with the rest of the codebase and prevents subtle divergence.
</issue_to_address>Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.
| SELECT b.id | ||
| FROM beliefs b | ||
| WHERE b.type = 'speculative' | ||
| AND b.origin = 'speculative' |
There was a problem hiding this comment.
issue: Use the ORIGIN_SPECULATIVE constant instead of hardcoding the origin string in the GC query.
Hardcoding b.origin = 'speculative' makes this query fragile if the origin wire value changes or differs from the literal. Using the existing constant (via interpolation or a parameter) keeps the query consistent with the rest of the codebase and prevents subtle divergence.
Owner
Author
|
[claim:review:planck:2026-05-10T05:05:23Z] |
|
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 |
robotrocketscience
added
attn:merge-conflict
Owner
Author
|
Reviewed and ready to merge — code is solid (see below) — but Code review: ✅ approved.
One subtle thing worth a comment but not a blocker: Action: Faraday, please rebase on Released claim. Setting — planck |
Owner
Author
|
[claim:review:rogue1:2026-05-10T05:08:06Z] |
Owner
Author
|
[release:review:rogue1:2026-05-10T05:08:11Z] |
Owner
Author
|
[release:review:planck:2026-05-10T05:08:21Z] |
…RCE_WONDER_INGEST + Belief.valid_to (#548) Pre-stages the type/edge/corroboration enum entries and the Belief.valid_to soft-delete field that the wonder lifecycle module depends on. RESOLVES gets edge_valence=0.0 (marker, not evidential).
…+ lifecycle allowlist (#548) ALTER TABLE beliefs ADD COLUMN valid_to TEXT (NULL=active). Partial index idx_beliefs_speculative_gc on (origin, created_at) WHERE valid_to IS NULL for GC scan performance. New MemoryStore.soft_delete_belief() (idempotent via WHERE valid_to IS NULL) and query_wonder_gc_candidates() (NOT EXISTS guards on feedback_history and RESOLVES edges). aelfrice.wonder.lifecycle added to INSERT_BELIEF_ALLOWLIST per the PR #478 gate; allowlist guard test updated.
New src/aelfrice/wonder/lifecycle.py. wonder_ingest persists Phantom candidates as type='speculative' beliefs (alpha=0.3, beta=1.0, origin=ORIGIN_SPECULATIVE), writes RELATES_TO edges to each constituent, and records a wonder_ingest corroboration row. Idempotent via SHA-256 of sorted constituent_belief_ids as content_hash (not text — phantoms with identical text from different pairs are distinct). wonder_gc soft-deletes stale phantoms via store.soft_delete_belief(). Preserves any phantom with a RESOLVES edge or alpha-update beyond epsilon band. Closes the cli.py:884 TODO against #229.
14 tests covering wonder_ingest (belief schema, session_id propagation, RELATES_TO edges, audit corroboration row, idempotency on constituent-pair key, distinct-pair non-dedup) and wonder_gc (dry-run, non-dry-run, fresh-skip, RESOLVES preservation incoming/outgoing, alpha-update preservation, idempotency).
robotrocketscience
force-pushed
the
feat/issue-548-wonder-lifecycle
branch
from
May 10, 2026 05:08
4c0d2ab to
e45fde3
Compare
robotrocketscience
added
attn:review
Owner
Author
|
[claim:review:rogue1:2026-05-10T05:11:53Z] |
Owner
Author
|
Review verdict: approve, merging. Read
Two non-blocking nits (file as follow-up, do not block this PR)
MechanicalMerging by local FF push. |
robotrocketscience
removed
the
attn:review
Owner
Author
|
[release:review:rogue1:2026-05-10T05:14:31Z] |
This was referenced May 10, 2026
This was referenced May 10, 2026
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
1 participant
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 #229. Sub-task C1+C2 of umbrella #542.
Implements the wonder lifecycle that the TODO at
src/aelfrice/cli.py:884flagged as deferred to v2.x:wonder_ingestpersistsPhantomcandidates fromaelf wonderas speculative beliefs, andwonder_gcsoft-deletes them on a TTL when no evidence accrues.What lands
src/aelfrice/wonder/lifecycle.py— new module:wonder_ingest(store, phantoms, session_id=None) -> WonderIngestResult— writes eachPhantomas atype='speculative'belief withorigin=ORIGIN_SPECULATIVE, prior α=0.3 / β=1.0,RELATES_TOedges to each constituent, and awonder_ingestcorroboration row whosesource_path_hashcarries"<generator>@<score:.4f>"for audit. Idempotent via SHA-256 of sortedconstituent_belief_idsascontent_hash(not text — phantoms with identical text from different pairs are distinct candidates).wonder_gc(store, ttl_days=14, dry_run=False) -> WonderGCResult— soft-deletes viastore.soft_delete_belief(). Preserves any phantom with aRESOLVESedge or an α-update beyond the epsilon band.src/aelfrice/models.py— three enum entries + one field:BELIEF_SPECULATIVE = "speculative"added toBELIEF_TYPES.EDGE_RESOLVES = "RESOLVES"added toEDGE_TYPESandEDGE_VALENCE(0.0 — marker, not evidential).CORROBORATION_SOURCE_WONDER_INGEST = "wonder_ingest"added toCORROBORATION_SOURCE_TYPES.Belief.valid_to: str | None— soft-delete timestamp (NULL = active).src/aelfrice/store.py— additive schema + GC primitives:ALTER TABLE beliefs ADD COLUMN valid_to TEXTmigration.idx_beliefs_speculative_gc ON beliefs(origin, created_at) WHERE valid_to IS NULLfor GC scan performance.soft_delete_belief(belief_id, ts=None)— idempotent viaWHERE valid_to IS NULL.query_wonder_gc_candidates(*, cutoff_ts, ...)— returns IDs satisfying all GC predicates (type/origin/age/priors/no-feedback/no-RESOLVES).aelfrice.wonder.lifecycleadded toINSERT_BELIEF_ALLOWLISTper the PR feat(view-flip): #265 PR-B — scanner→worker migration + insert_belief gate #478 gate.Tests — 14 cases in
tests/test_wonder_lifecycle.pycovering all 7 acceptance criteria from [v2.1] C1+C2: wonder_ingest + wonder_gc lifecycle (closes #229, #542 sub-task) #548 (belief schema, RELATES_TO edges, audit row, ingest idempotency, distinct-pair non-dedup, GC dry-run, GC non-dry-run, RESOLVES preservation incoming + outgoing, α-update preservation, GC idempotency). Existing guard tests (test_corroborations.py,test_insert_belief_gate.py) updated to include the new enum entries — no weakening (still exact-match frozenset comparisons).Acceptance vs the issue body
wonder_ingestwrites a belief with the documented schema (origin, type, alpha, beta).wonder_ingestwrites RELATES_TO edges to all constituents.wonder_ingestwrites the audit corroboration row with the right tag.wonder_ingestis idempotent (same constituent set → no duplicates).wonder_gc --dry-runreports candidates, doesn't act.wonder_gcnon-dry-run setsvalid_to.wonder_gcpreserves phantoms withRESOLVESedges (incoming OR outgoing).wonder_gcpreserves phantoms with α updates.wonder_gcis idempotent.Out of scope (handled in sibling sub-issues)
aelf wonder --persist,aelf wonder gc) — sibling [v2.1] C3: aelf wonder --persist + gc CLI surface (#542 sub-task) #549 (C3).aelf confirm) — sibling [v2.1] C4: phantom promotion trigger (#542 sub-task) #550 (C4), depends on this PR.wonder_ingest) — sibling [v2.1] E4: skill layer integration — subagent dispatch → wonder_ingest (#542 sub-task) #552 (E4), depends on this PR.Notes
audit_logtable exists today (and per [phantom-prereqs T2] Implicit retrieval-driven feedback — sweeper + grace window #191/feat: deferred-feedback sweeper — implicit retrieval-driven posterior signal (#191) #256 implementation,feedback_historyis the live audit surface for posterior updates). Usedbelief_corroborationswithsource_type='wonder_ingest'for the audit row — consistent with how the v1.5.0 corroboration design treats re-assertion events as first-class signals.github/main(v2.0.1 → v2.1.0). All commits in this branch land on top of currentgithub/maindirectly; no version-rollback noise included.Summary by Sourcery
Introduce lifecycle support for speculative "wonder" beliefs, including ingestion, soft-deletion, and schema wiring for auditability and GC.
New Features:
Enhancements:
Tests: