Single-User Temporal Memory foundation by G9000 · Pull Request #67 · G9000/animaOS

G9000 · 2026-06-28T19:15:48Z

## Summary

Stacked integration branch for the Single-User Temporal Memory foundation. Current completed scope on this branch covers SUM-001 through SUM-011; all child tickets in the tracker are now done.
SUM-001: baseline memory truth audit, deterministic recall probes, scored retrieval pool fixes, sleep consolidation cursor metadata, and KG vault portability fixes.
SUM-002: evidence audit/backfill surfaces, episode quality safeguards, grounded salient details, and multilingual/degraded retrieval baseline improvements.
SUM-003: temporal knowledge graph with relation lifecycle/history, alias-aware traversal, legacy KG migration repair, current graph API filtering, and vault round-trip support.
SUM-004: structured user profile storage, extraction/promotion, prompt rendering, correction API, profile forget cleanup, eval reset handling, and vault preservation.
SUM-005: deterministic retrieval router/query plans, intent-specific lane composition, serializable route traces, search_long_memory(mode="auto"), and focused route accuracy regressions.
SUM-006: salience-aware decay and soft evolution, including additive salience metadata, decay-class heat scoring, soft evolution chains, duplicate-candidate reinforcement, sleep-time drift surfacing, salience migrations, and vault support.
SUM-007: cross-episode pattern synthesis, including time/topic/salience episode sampling, strict repeated-evidence parsing, evidence-backed pattern memories, compact prompt rendering, sleep-time orchestration, duplicate suppression, stale-episode cleanup, and derived-pattern forget cleanup.
SUM-008: foresight signals with encrypted soul storage, deterministic relative-date extraction, lifecycle sweeping, consolidation writes, timezone-aware prompt retrieval, cancellation matching, and vault portability.
SUM-009: procedural experience and skill memory with encrypted experience rows, embedding-based clustering, learned skill distillation, growth-log entries, background capture, prompt retrieval, duplicate suppression, and redacted tool-output summaries.
SUM-010: optional external adapter seams with a native retrieval backend contract, rebuild-from-canonical SQLCipher helpers, derived-index fallback behavior, adapter documentation, and deferred graph adapter scope.
SUM-011: stable anima_server.services.memory package boundary with domain contracts, salience/retrieval facades, temporal helpers, package-boundary tests, and architecture references.
Includes Codex review follow-up across the stacked PRs, including cursor guards, SUM-007 encrypted/stale/duplicate pattern fixes, SUM-008/SUM-009 foresight/procedural review fixes, SUM-010 native retrieval adapter contract coverage, and SUM-005 router review cleanup.

Architecture snapshot

flowchart TD
  User["User turn"] --> Agent["Agent runtime"]
  Agent --> Extract["Memory extraction and consolidation"]
  Extract --> Evidence["Evidence anchors"]
  Evidence --> SQLCipher["Canonical SQLCipher soul DB"]

  subgraph Contract["Shared contract boundary"]
    Core["packages/anima-core"]
    PyMemory["anima_server.services.memory"]
    Fixtures["Parity fixtures and tests"]
    Core --> Fixtures
    PyMemory --> Fixtures
  end

  subgraph Durable["Durable memory records"]
    Claims["Temporal facts / MemoryClaim"]
    Graph["Temporal relationships / KGRelation"]
    Items["MemoryItem and episodes"]
    Semantics["Salience, stability, decay semantics"]
    Patterns["Cross-episode patterns"]
    Foresight["ForesightSignal"]
    Procedural["AgentExperience, clusters, skills"]
  end

  SQLCipher --> Claims
  SQLCipher --> Graph
  SQLCipher --> Items
  SQLCipher --> Semantics
  SQLCipher --> Patterns
  SQLCipher --> Foresight
  SQLCipher --> Procedural

  Extract --> Claims
  Extract --> Graph
  Extract --> Items
  Extract --> Foresight
  Agent --> Procedural

  subgraph Sleep["Sleep and background cognition"]
    PatternSynth["Pattern synthesis"]
    ForesightSweep["Foresight lifecycle sweep"]
    SkillDistill["Experience clustering and skill distillation"]
    Decay["Salience-aware decay"]
  end

  Items --> PatternSynth --> Patterns
  Foresight --> ForesightSweep --> Foresight
  Procedural --> SkillDistill --> Procedural
  Semantics --> Decay --> Items

  subgraph Recall["Recall path"]
    Query["Memory query"]
    Router["Retrieval router"]
    Plan["Intent-specific query plan"]
    Lanes["Profile, graph, episodes, transcripts, foresight, skills"]
    Trace["Recall trace and score breakdown"]
  end

  Agent --> Query --> Router --> Plan --> Lanes
  Claims --> Lanes
  Graph --> Lanes
  Items --> Lanes
  Patterns --> Lanes
  Foresight --> Lanes
  Procedural --> Lanes
  Lanes --> Trace --> Prompt["Fenced memory prompt blocks"]
  Prompt --> Agent

  subgraph Cleanup["Privacy and eval cleanup"]
    Forget["User forget"]
    EvalReset["Eval reset"]
  end

  Forget --> Claims
  Forget --> Items
  Forget --> Foresight
  EvalReset --> Claims
  EvalReset --> Graph
  EvalReset --> Items
  EvalReset --> Foresight
  EvalReset --> Procedural

sequenceDiagram
  participant User
  participant Forget as forget_memory
  participant Evidence as MemoryItemEvidence
  participant Foresight as foresight_signals
  participant Prompt as Memory prompt blocks

  User->>Forget: forget memory
  Forget->>Evidence: read source runtime message IDs
  Forget->>Foresight: delete same-user signals with overlapping source IDs
  Forget->>Evidence: delete memory evidence and claims
  Prompt->>Foresight: build foresight block
  Foresight-->>Prompt: forgotten future event no longer appears

Affected areas

apps/server: memory retrieval, retrieval router plans, retrieval backend seams, sleep-time consolidation, evidence/provenance, episodes, temporal KG, structured profile, salience/heat scoring, pattern synthesis, foresight signals, procedural experience memory, prompt rendering/budgeting, agent post-turn hooks, forgetting, vault, API routes, and Alembic migrations.
packages/anima-core: additive shared memory contract types for temporal record status, endpoint kinds, salience classes, recall score breakdowns, temporal facts, and temporal relationships.
apps/server/tests and packages/anima-core/tests: deterministic baseline, multilingual retrieval, retrieval rebuild contracts, evidence, episode, KG, profile, salience, pattern synthesis, foresight, procedural experience, router/package-boundary, Rust memory contract, agent service hooks, forgetting, vault, runtime DB, and review-regression coverage.
docs/architecture/memory, docs/prds/memory, tickets/single-user-temporal-memory-v2: baseline audit docs, optional adapter documentation, package-boundary documentation, and completed child-ticket tracker updates.

Validation

Focused suites, red/green review regressions, lint, build, health smoke, Alembic checks, and repeated full backend runs are recorded in the child ticket validation sections and stacked PR history.
Latest SUM-001 review loop on this branch: test_sleep_agent.py 20 passed; focused memory/vault/BM25/baseline suite 74 passed; bun run lint passed; bun run build passed.
SUM-006 was restacked onto this branch through PR SUM-006 add salience-aware decay and soft evolution #75; its ticket notes validation was not run in that restack session.
SUM-007 was restacked onto this branch through PR SUM-007 Cross-episode pattern synthesis #76. Its latest recorded checks include focused pattern/forgetting review regressions passing, apps/server/tests/test_forgetting.py 33 passed, and bun run lint:server passed.
SUM-008 and SUM-009 were restacked onto this branch through PR SUM-008 Foresight signals and procedural skill memory #77. Recorded PR SUM-008 Foresight signals and procedural skill memory #77 validation includes test_foresight.py + test_agent_experience.py 8 passed, adjacent agent/prompt/sleep/service set 64 passed, full backend test:server -- --maxfail=1 -q 1775 passed and 1 skipped, bun run lint:server passed, bun run build passed, bun run db:server:heads reported 20260703_0002 (head), and health smoke for GET /health returned 200 ok.
PR SUM-008 Foresight signals and procedural skill memory #77 added a narrow guard to the inherited SUM-006 salience migration so stamped legacy soul databases missing memory_items can reach metadata repair; this addressed the earlier SUM-007 full-suite failure path.
SUM-010 was restacked onto this branch through PR SUM-010 optional external adapter seams #78. Recorded PR SUM-010 optional external adapter seams #78 validation includes test_memory_retrieval_rebuild.py 8 passed, adjacent BM25/hybrid retrieval suite 25 passed, bun run lint:server passed, git diff --check passed, bun run build passed after fresh worktree dependency install, and FastAPI GET /health smoke returned 200 ok.
SUM-011 package-boundary validation on this branch: package-boundary tests 4 passed, bun run lint:server passed, git diff --check passed with CRLF notices only, bun run build:server passed, and rename checks found no services.memory_v2 package references.
SUM-005 router/contract validation on this branch: router/package-boundary tests 13 passed, cargo test -p anima-core memory_contract 4 passed, full cargo test -p anima-core passed, touched Rust files passed rustfmt --check, bun run lint:server passed, bun run build:server passed, bun run db:server:current passed, git diff --check passed with CRLF notices only, and temporary-data FastAPI GET /health returned 200 ok.

Notes

Parent tracker is now done; every child ticket from SUM-001 through SUM-011 is marked complete.
SUM-005 implements query-plan lane composition and maps automatic tool routing into the existing wide-evidence execution modes. It does not introduce a separate per-lane retrieval executor yet.
SUM-007 uses existing MemoryItem and MemoryItemEvidence storage with category=pattern and source=pattern_synthesis; no schema migration was required.
SUM-008 and SUM-009 add new core migrations for foresight signals and procedural experience memory, plus vault export/import coverage for foresight/procedural tables with import-time re-encryption.
SUM-010 keeps SQLCipher as canonical memory storage. External vector adapters are documented as optional, rebuildable derived indexes; graph adapter seams remain deferred until temporal KG semantics are stable enough for a separate contract.
This PR intentionally replaces the earlier SUM-001-only description because SUM-002, SUM-003, SUM-004, SUM-005, SUM-006, SUM-007, SUM-008, SUM-009, SUM-010, and SUM-011 have since been merged into the same stacked branch.

Memory package boundary follow-up planning

Adds a separate backlog initiative, memory-package-boundary, for hardening the services.memory ownership boundary after this foundation lands.
Includes implementation plan docs/superpowers/plans/2026-07-03-memory-package-boundary-hardening.md and MPB-000 through MPB-007 tickets.
This is planning only: the SUM implementation remains complete, and MPB tickets are intentionally backlog for follow-up work.

G9000 · 2026-06-28T19:37:37Z

@codex review pwease. Don't nitpick.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 46dd1f422f

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

G9000 · 2026-06-29T02:42:29Z

@codex review pwease. Don't nitpick.

chatgpt-codex-connector · 2026-06-29T02:47:16Z

Codex Review: Didn't find any major issues. Chef's kiss.

Reviewed commit: 8cd77ccb59

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

…baseline # Conflicts: # docs/architecture/README.md

G9000 · 2026-06-29T14:52:31Z

@codex review pwease. Don't nitpick.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d78a4ccb8f

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

G9000 · 2026-06-29T15:04:40Z

@codex review pwease. Don't nitpick.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 48bb32d702

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

G9000 · 2026-06-29T15:16:14Z

@codex review pwease. Don't nitpick.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 50bc3293b4

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

G9000 · 2026-06-29T15:24:54Z

@codex review pwease. Don't nitpick.

chatgpt-codex-connector · 2026-06-29T15:31:31Z

Codex Review: Didn't find any major issues. 🎉

Reviewed commit: 5361542c34

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

* memory: add evidence audit and episode quality safeguards * SUM-003 temporal knowledge graph v2 (#70) * memory: add temporal knowledge graph v2 * memory: fix temporal kg review regressions * api: filter superseded graph relations * db: repair legacy kg schema columns * memory: preserve kg alias reads and confidence * memory: resolve kg context aliases * memory: resolve kg alias pruning * memory: filter kg pruning candidates * db: guard legacy kg downgrade * memory: prevent kg self supersession * SUM-004 structured user profile (#71) * memory: add structured user profile * memory: address profile review feedback * memory: invalidate profile cache after forget * memory: hard-delete forgotten profile state * memory: preserve profile fields with surviving evidence * memory: link profile forget cleanup to runtime provenance * memory: narrow runtime profile forget provenance * memory: preserve manual profile corrections * memory: reject forgotten profile candidates * memory: canonicalize profile corrections * memory: narrow profile claim reconciliation * memory: tighten profile forget matching * memory: preserve profile vault state * memory: dedupe claim profile reconciliation * memory: restore profile vault links safely * memory: delete profile evidence on forget * memory: preserve newer profile reconciliation * memory: return bad request for blank profile corrections * memory: preserve profile candidate timestamps * memory: scope profile forget cleanup * memory: prioritize profile prompt block * memory: truncate profile prompt lines * memory: restore prior profile on forget

* memory: add salience-aware decay * memory: address salience review comments * memory: refresh heat after salience reinforcement * memory: preserve neutral salience heat defaults * memory: preserve duplicate candidate salience * memory: refresh promoted duplicate candidate evidence * memory: preserve candidate audit history * memory: repair legacy salience schema * memory: rebase salience migrations * memory: preserve zero evidence strength restore * memory: update duplicate reinforcement test * memory: mark reinforced candidates distinctly * memory: treat reinforced candidates as terminal

* memory: synthesize cross-episode patterns * memory: tighten pattern synthesis prompt wording * docs: record SUM-007 full-suite result * memory: skip duplicate pattern evidence * memory: decrypt pattern episode arcs * memory: address pattern synthesis review * memory: clean stale pattern synthesis references * memory: clean deleted pattern indexes * memory: decrypt pattern evidence for forget cleanup * memory: delete derived pattern evidence

* memory: add foresight and procedural learning * memory: fix foresight review regressions * memory: address foresight and skill review follow-ups * memory: fix foresight extraction and cluster state review * memory: prioritize dated foresight in prompts * memory: address timezone and resume review * memory: use local dates for foresight views * memory: redact procedural tool outputs * memory: address foresight and procedural review * memory: cover vault review fixes * memory: tighten foresight cancellation matching

G9000 · 2026-07-03T09:18:24Z

@codex review

G9000 · 2026-07-03T09:40:00Z

@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 36d4184488

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

G9000 · 2026-07-03T10:26:40Z

@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 85344a7435

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

G9000 · 2026-07-03T10:48:33Z

@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: cd75e712a9

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

G9000 · 2026-07-03T11:17:02Z

@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: acdb0db8d9

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

G9000 · 2026-07-03T11:38:47Z

@codex review

chatgpt-codex-connector · 2026-07-03T11:49:24Z

Codex Review: Didn't find any major issues. Swish!

Reviewed commit: 08bd6d45e8

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

* memory: baseline temporal memory audit * memory: address baseline review feedback * memory: reset kg vault restore sequences * memory: fix default sleep cursor runtime factory * memory: guard sleep cursor on candidate backlog * memory: include extraction failures in sleep cursor guard * SUM-002 evidence baseline and episode quality (#68) * memory: add evidence audit and episode quality safeguards * SUM-003 temporal knowledge graph v2 (#70) * memory: add temporal knowledge graph v2 * memory: fix temporal kg review regressions * api: filter superseded graph relations * db: repair legacy kg schema columns * memory: preserve kg alias reads and confidence * memory: resolve kg context aliases * memory: resolve kg alias pruning * memory: filter kg pruning candidates * db: guard legacy kg downgrade * memory: prevent kg self supersession * SUM-004 structured user profile (#71) * memory: add structured user profile * memory: address profile review feedback * memory: invalidate profile cache after forget * memory: hard-delete forgotten profile state * memory: preserve profile fields with surviving evidence * memory: link profile forget cleanup to runtime provenance * memory: narrow runtime profile forget provenance * memory: preserve manual profile corrections * memory: reject forgotten profile candidates * memory: canonicalize profile corrections * memory: narrow profile claim reconciliation * memory: tighten profile forget matching * memory: preserve profile vault state * memory: dedupe claim profile reconciliation * memory: restore profile vault links safely * memory: delete profile evidence on forget * memory: preserve newer profile reconciliation * memory: return bad request for blank profile corrections * memory: preserve profile candidate timestamps * memory: scope profile forget cleanup * memory: prioritize profile prompt block * memory: truncate profile prompt lines * memory: restore prior profile on forget * SUM-006 add salience-aware decay and soft evolution (#75) * memory: add salience-aware decay * memory: address salience review comments * memory: refresh heat after salience reinforcement * memory: preserve neutral salience heat defaults * memory: preserve duplicate candidate salience * memory: refresh promoted duplicate candidate evidence * memory: preserve candidate audit history * memory: repair legacy salience schema * memory: rebase salience migrations * memory: preserve zero evidence strength restore * memory: update duplicate reinforcement test * memory: mark reinforced candidates distinctly * memory: treat reinforced candidates as terminal * SUM-007 Cross-episode pattern synthesis (#76) * memory: synthesize cross-episode patterns * memory: tighten pattern synthesis prompt wording * docs: record SUM-007 full-suite result * memory: skip duplicate pattern evidence * memory: decrypt pattern episode arcs * memory: address pattern synthesis review * memory: clean stale pattern synthesis references * memory: clean deleted pattern indexes * memory: decrypt pattern evidence for forget cleanup * memory: delete derived pattern evidence * SUM-008 Foresight signals and procedural skill memory (#77) * memory: add foresight and procedural learning * memory: fix foresight review regressions * memory: address foresight and skill review follow-ups * memory: fix foresight extraction and cluster state review * memory: prioritize dated foresight in prompts * memory: address timezone and resume review * memory: use local dates for foresight views * memory: redact procedural tool outputs * memory: address foresight and procedural review * memory: cover vault review fixes * memory: tighten foresight cancellation matching * memory: add optional retrieval adapter seams (#78) * memory: add memory v2 package boundary * memory: rename package boundary * memory: add retrieval router plans * docs: plan memory package boundary hardening * memory: clean durable rows on reset and forget * docs: add memory architecture diagrams * memory: align salience evidence contract * memory: narrow foresight forget cleanup * vault: defer kg relation self-links on import

G9000 added 2 commits June 29, 2026 03:13

memory: baseline temporal memory audit

8ae5f8c

memory: address baseline review feedback

46dd1f4

chatgpt-codex-connector Bot reviewed Jun 28, 2026

View reviewed changes

Comment thread apps/server/src/anima_server/services/vault.py

G9000 marked this pull request as ready for review June 29, 2026 02:28

memory: reset kg vault restore sequences

8cd77cc

G9000 mentioned this pull request Jun 29, 2026

SUM-002 evidence baseline and episode quality #68

Merged

Merge remote-tracking branch 'origin/main' into codex/sum-001-memory-…

d78a4cc

…baseline # Conflicts: # docs/architecture/README.md

G9000 mentioned this pull request Jun 29, 2026

SUM-003 temporal knowledge graph v2 #70

Merged

chatgpt-codex-connector Bot reviewed Jun 29, 2026

View reviewed changes

Comment thread apps/server/src/anima_server/services/agent/sleep_agent.py

memory: fix default sleep cursor runtime factory

48bb32d

chatgpt-codex-connector Bot reviewed Jun 29, 2026

View reviewed changes

Comment thread apps/server/src/anima_server/services/agent/sleep_agent.py

memory: guard sleep cursor on candidate backlog

50bc329

chatgpt-codex-connector Bot reviewed Jun 29, 2026

View reviewed changes

Comment thread apps/server/src/anima_server/services/agent/sleep_agent.py

memory: include extraction failures in sleep cursor guard

5361542

G9000 changed the title ~~SUM-001 baseline temporal memory audit~~ Single-User Temporal Memory v2 foundation Jul 1, 2026

This was referenced Jul 1, 2026

SUM-005 retrieval router and query plans #72

Merged

SUM-006 add salience-aware decay and soft evolution #75

Merged

G9000 mentioned this pull request Jul 2, 2026

SUM-007 Cross-episode pattern synthesis #76

Merged

G9000 mentioned this pull request Jul 2, 2026

SUM-008 Foresight signals and procedural skill memory #77

Merged

G9000 mentioned this pull request Jul 3, 2026

SUM-010 optional external adapter seams #78

Merged

G9000 added 4 commits July 3, 2026 15:27

memory: add optional retrieval adapter seams (#78)

faf0f0d

memory: add memory v2 package boundary

621bc66

memory: rename package boundary

49d2165

memory: add retrieval router plans

bfae87b

G9000 changed the title ~~Single-User Temporal Memory v2 foundation~~ Single-User Temporal Memory foundation Jul 3, 2026

docs: plan memory package boundary hardening

36d4184

chatgpt-codex-connector Bot reviewed Jul 3, 2026

View reviewed changes

Comment thread apps/server/src/anima_server/services/eval_reset.py

Comment thread apps/server/src/anima_server/services/agent/forgetting.py

memory: clean durable rows on reset and forget

85344a7

docs: add memory architecture diagrams

0a208b8

chatgpt-codex-connector Bot reviewed Jul 3, 2026

View reviewed changes

Comment thread packages/anima-core/src/memory_contract.rs Outdated

memory: align salience evidence contract

cd75e71

chatgpt-codex-connector Bot reviewed Jul 3, 2026

View reviewed changes

Comment thread apps/server/src/anima_server/services/agent/forgetting.py

memory: narrow foresight forget cleanup

acdb0db

chatgpt-codex-connector Bot reviewed Jul 3, 2026

View reviewed changes

Comment thread apps/server/src/anima_server/services/vault.py Outdated

vault: defer kg relation self-links on import

08bd6d4

G9000 merged commit b89232d into main Jul 3, 2026

G9000 deleted the codex/sum-001-memory-baseline branch July 3, 2026 17:17

Conversation

G9000 commented Jun 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Architecture snapshot

Affected areas

Validation

Notes

Memory package boundary follow-up planning

Uh oh!

G9000 commented Jun 28, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

G9000 commented Jun 29, 2026

Uh oh!

chatgpt-codex-connector Bot commented Jun 29, 2026

Uh oh!

G9000 commented Jun 29, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

G9000 commented Jun 29, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

G9000 commented Jun 29, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

G9000 commented Jun 29, 2026

Uh oh!

chatgpt-codex-connector Bot commented Jun 29, 2026

Uh oh!

G9000 commented Jul 3, 2026

Uh oh!

G9000 commented Jul 3, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

Uh oh!

G9000 commented Jul 3, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

G9000 commented Jul 3, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

G9000 commented Jul 3, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

G9000 commented Jul 3, 2026

Uh oh!

chatgpt-codex-connector Bot commented Jul 3, 2026

Uh oh!

G9000 commented Jun 28, 2026 •

edited

Loading