drone @memory rollover run                 # Execute rollover for files over limits
drone @memory rollover status              # Show per-branch rollover statistics
drone @memory rollover check               # Dry run — what needs rollover
drone @memory rollover sync-lines          # Update line count metadata

drone @memory search "query"               # Semantic search across all branch memories
drone @memory search "query" --branch X    # Filter by branch
drone @memory search "query" --n 10        # Limit results

drone @memory symbolic demo                # Mock analysis demonstration
drone @memory symbolic fragments "query"   # Search stored symbolic fragments
drone @memory symbolic extract <file>      # Extract fragments via LLM (requires API key)
drone @memory symbolic bootstrap           # Populate fragments from session JSONLs
drone @memory symbolic hook-test           # Test hook with sample conversation text

drone @memory templates push-templates     # Push template updates to all branches
drone @memory templates diff-templates     # Show template differences per branch
drone @memory templates template-status    # Show template version and push status

drone @memory lint                         # Audit .trinity entries for over-limit violations (read-only)
drone @memory lint @devpulse               # Lint a specific branch

drone @memory verify FPLAN-XXXX            # Check if plan is vectorized in ChromaDB
drone @memory watch                        # Auto-rollover watcher daemon (Ctrl+C to stop)

Architecture

memory/
├── apps/
│   ├── memory.py                # Entry point — auto-discovers modules
│   ├── modules/                 # 6 modules
│   │   ├── lint.py              # Entry limit violation scanner (read-only)
│   │   ├── rollover.py          # Rollover orchestration, status, sync-lines
│   │   ├── search.py            # Semantic query routing
│   │   ├── symbolic.py          # Fragmented memory extraction and search
│   │   ├── templates.py         # Template push, diff, status
│   │   └── verify.py            # Plan vectorization check
│   └── handlers/                # 14 handler groups
│       ├── archive/             # indexer.py
│       ├── intake/              # plans_processor.py, pool_processor.py
│       ├── json/                # json_handler.py, memory_files.py, entry_limits.py, lint_handler.py, config_loader.py
│       ├── learnings/           # manager.py
│       ├── monitor/             # detector.py, memory_watcher.py
│       ├── rollover/            # extractor.py, orchestrator.py
│       ├── schema/              # normalize.py
│       ├── search/              # query_executor.py, vector_search.py
│       ├── storage/             # chroma.py, chroma_subprocess.py
│       ├── symbolic/            # chroma_client, deduplicator, extractor, hook, retriever, storage
│       ├── templates/           # pusher.py, differ.py, spawn_pusher.py
│       ├── tracking/            # line_counter.py, tab_renderer.py
│       ├── vector/              # embedder.py, embed_subprocess.py
│       └── central_writer.py
├── templates/                   # LOCAL.template.json, OBSERVATIONS.template.json
├── tests/                       # 978 tests
├── .chroma/                     # ChromaDB vector store
└── memory_json/                 # Operation logs + custom_config/memory.config.json

Rollover Pipeline

detector.check_all_branches()        # scan AIPASS_REGISTRY.json + external registries
→ _should_rollover(file)             # v1: line_count >= max_lines (600)
                                     # v2: len(sessions) >= max_sessions (20)
→ orchestrator.execute_rollover()
  → create_rollover_backup()         # safety copy to branch/.backup/
  → extract_items()                  # v2: max(excess, 1) oldest entries
  → embed via subprocess             # fastembed (ONNX) in memory .venv
  → upsert in ChromaDB               # content-hash IDs (sha256[:16]), no duplicates
  → trim source file                 # write back with oldest removed

Rollover writes safety copies (rollover_backup_*.json) into <branch>/.backup/ — a shared runtime namespace (see @backup's README for all writers).

Subprocess Isolation

All ML operations (fastembed, chromadb) run via subprocess. The main process never imports these libraries. Python interpreter resolved via _get_memory_python() (env var AIPASS_MEMORY_PYTHON → memory/.venv/bin/python → sys.executable).

State-Tabs (`*_meta` keys)

Every .trinity/local.json and .trinity/observations.json carries inline *_meta banner strings that tell the editing agent what rollover rules apply to each section. Example:

"sessions_meta": "⟦ rollover ON → oldest archived to @memory · keep 15 · summary ≤300 chars ⟧"

Source of truth: memory.config.json — rollover counts (defaults + per-branch overrides) and entry char limits. Tab strings are generated, never hand-written.

Sections: todos_meta (rollover OFF — operational, never trimmed), key_learnings_meta, sessions_meta, observations_meta (all rollover ON).

Two value flows

Scenario	How tabs arrive
Live branches	`refresh_all_tabs()` walks the registry, renders tabs from config with per-branch overrides, writes them into `.trinity/` files. Wired after rollover, sync-lines, and push-templates.
New branches	Templates carry `{{TODOS_META}}`, `{{KEY_LEARNINGS_META}}`, `{{SESSIONS_META}}`, `{{OBSERVATIONS_META}}` placeholders. `spawn_pusher` propagates these (unresolved) from memory templates → spawn template sets. At branch creation, @spawn calls `render_all_meta_tabs()` to get rendered defaults and resolves the placeholders.

Public API

from aipass.memory.apps.handlers.tracking.tab_renderer import render_all_meta_tabs

tabs = render_all_meta_tabs()
# → {"TODOS_META": "⟦ rollover OFF ...", "KEY_LEARNINGS_META": "⟦ rollover ON ...", ...}

Returns defaults (not per-branch overrides) — appropriate for template resolution at branch creation.

Integration Points

Depends on:

prax — logging via get_system_logger()
api — API key for symbolic extraction (get_api_key())
AIPASS_REGISTRY.json — branch discovery for rollover scanning
External *_REGISTRY.json — scanned via AIPASS_CALLER_CWD

Provides to:

All branches — rollover archival when .trinity/ files hit limits
All branches — semantic search across archived memories
All branches — .trinity/ template distribution and sync

ML dependencies (memory .venv/ only):

fastembed — ONNX embeddings (model: sentence-transformers/all-MiniLM-L6-v2)
chromadb — vector storage and semantic search
numpy

Quality

Tests: 978 passed, 0 failures, 0 skips
Seedgo: 100%

Known Issues

search requires fastembed in memory .venv/ — fails without it
rollover status shows 0 branches when registry path not resolved
memory_threshold_exceeded trigger event registered but never fired

Last Updated: 2026-06-25

← Back to AIPass

README.md

MEMORY

Commands