Agent Operations Retrospective (6/7) — 10 Omissions Found by Full Inventory Audit

memcore — SQLite로 다시 짠 메모리 코어, 51 tests pass

Subtitle: A full cross-reference of 65 scripts reveals 10 missing entries around the memory layer

핵심 요약

  • Full cross-reference of 65 files in ~/Project/scripts/ confirmed that 10 peripheral memory scripts were missing from the previously documented inventory.
  • The 10 core memory pipeline scripts (retain/dialectic/decay/lint/entities/topics/prefetch/bank_migrate) are fully migrated.
  • Three items must be resolved before Hermes goes live: vector embedding / decisions queue / archive & housekeeping.

This post documents the full cross-reference audit pattern for detecting the "migrate the core, drop the periphery" mistake in migration work — and classifies the 10 omitted scripts it surfaced.

1. memcore 라이브러리

Full Cross-Reference Audit — How It Works

Migration inventories are typically built around "what looks necessary." This approach gets the core pipeline across but drops housekeeping, archive, and cleanup scripts. The system runs fine initially, then archive/TTL/stale cleanup accumulates as deferred debt and fails days later.

The procedure to catch this is straightforward.

  1. Snapshot the full file list of the source directory (~/Project/scripts/).
  2. Map each entry against the migration inventory (the porting spec).
  3. Classify unmapped files into three tiers: core / peripheral / diagnostic & utility.
  4. Re-classify peripheral tier into: pre-launch / post-launch / low-priority.

Applying this procedure to 65 files in ~/Project/scripts/ confirmed the core inventory was already migrated, but 10 peripheral scripts remained unmapped.

The 10 Omissions — By Category

Immediate Add — memcore v0.2

  1. decision-prepare.py — Generates LLM decision requests (TOPIC_CLASSIFY, CONFLICT_RESOLVE, CHANGELOG_SUMMARIZE)
  2. decision-apply.py — Applies LLM responses to bank/
  3. conflict-apply.py — Applies W-tag conflict resolution
  4. memory-archive.sh — Moves processed memory/ entries to archived/
  5. memory-optimize.py — Reflect Phase 2 optimization
  6. memory-warning-report.py — Daily audit Phase C memory warning
3. bank/ → memcore 마이그레이션

Medium Priority

  1. recall-cleanup.py — Recall TTL cleanup
  2. session-cleanup.py — Stale/orphan session cleanup
  3. session-archive.py — Harvests content from .deleted sessions
  4. monthly-review.py — Monthly memory review data collection

Diagnostic & Utility (Low Priority)

reflect-trace.py, section-extract.py, source-ingest.py, skill-candidate-stage.py. These are not launch blockers; they are classified separately as observation and collection utilities.

Decision Tree — Pre/Post-Launch Split

The audit results drove four architectural decisions.

  • (a) Vector embedding: Required. Preserves semantic search when DB scales.
  • (b) Sync interval: 30 min → 24 h relaxation is dual-run period only. Reverts to real-time sync_turn when Hermes becomes primary.
  • (c) Heartbeat escalation: Port after Hermes launch.
  • (d) docs-snapshot: Configured separately for Hermes.

Priority Table — Before and After Hermes Launch

# Task Timing Est.
1 Vector embedding layer (sqlite-vss or separate .ann + bge-m3 / Hermes auxiliary_client) Pre-launch 2–3h
2 memcore.decisions (LLM decision queue) Pre-launch 2h
3 memcore.archive + housekeeping Pre-launch 1–2h
4 Hermes live launch validation Launch —
5 memcore.heartbeat (proactive 1310→300) Post-launch 2h
6 Daily Audit → Hermes cron Post-launch 1h

Migration Baseline (Unchanged)

Items that were not disturbed by the audit. Recording these alongside the omission list prevents self-correction from skewing in one direction.

  • memcore.dialectic (U-tag 3-phase)
  • memcore.decay (confidence decay)
  • memcore.lint (Retain tag validator)
  • memcore.entities (stale detection)
  • memcore.topics (registry + M2M)
  • memcore.ingest (retain-extract / merge unified)
  • memcore.prefetch (FTS5 hybrid)
  • memcore.bank_migrate (READ-ONLY, --incremental)
  • MemcoreProvider adapter (Hermes MemoryProvider implementation)
  • LaunchAgent auto sync (daily 03:30)
  • Data migration: 38 files → 236 sections + 4 entities + 15 topics + 498 links (0 errors)
  • Tests: 51/51 passed

Scope and Limitations

The full cross-reference audit is effective under these conditions.

  • The source tree is concentrated in one folder or one repository (distributed source trees produce unwieldy file snapshots).
  • Migration targets are script or utility units (function- or class-level porting is better served by AST-based tooling).
  • Unmapped files can be manually classified by someone with domain knowledge (without it, the boundary between diagnostic/utility and peripheral tiers blurs).

The limitation is explicit. The cross-reference only catches "file present / absent." It does not catch semantic omissions — cases where functionality was ported but constraints, configuration, or defaults differ. Those require separate behavior-based validation (tests, sample input comparison).

Summary

An inventory is never complete on the first pass. A list built by selecting only what looks essential will always drop the peripheral tier. "Scan the entire script folder one more time" is the single cheapest and most effective safety net in migration audit work.

Open question: how far can full cross-reference be automated? File list comparison is scriptable, but the three-tier classification and pre/post-launch split currently require human judgment. Whether an LLM-based classifier can absorb that boundary is the next thing to observe.

Series Index

  1. Dissecting OpenClaw — Agent and Cron Structure Analysis
  2. What to Migrate, What to Drop — 4-Category Classification Table
  3. Code Review: Selecting 8 Scripts Worth Porting
  4. Migration Plan — Porting Spec Grouped into 11 Categories
  5. memcore — Memory Core Rebuilt on SQLite, 51 Tests Pass
  6. 10 Omissions Found by Full Inventory Audit — Script Audit Pattern (this post)

댓글

댓글 쓰기

이 블로그의 인기 게시물

Agent Memory Engine (2/10) — Building an AI Agent Memory System with SQLite Alone

← 1/10 SQLite Memory Engine f… 📚 Series Index 3/10 memcore Completed → A memory engine that runs without daemons, without dependencies, anywhere 핵심 요약 This post covers how to design an AI agent memory engine (memcore) on a single-file SQLite backend. - Structure: 25 modules, ~6,464 lines, 22 DB tables - Principles: no daemon, single-file portability, no required external dependencies, host-agnostic - Core techniques: 4-layer hybrid search (topic · vector · FTS5 · LIKE), U-tag dialectic-based confidence evolution, ingestion-layer bias rejection rules What This Post Covers One approach to building AI agent memory without a server or cloud vector DB. Specifically: (1) the limitations of text-file-based memory, (2) vector DB alternatives compared and selection criteria, (3) how to layer search tiers, (4) how to evolve confidence scores over time, and (5) how to block biased memory accumulation at the ingestion stage. Design Principles: Why Single-File SQLite Text-file-based ...
Read more »

"ML Foundations (9/9) — PyTorch vs TensorFlow, and the Road to Local LLMs"

이미지
← 8/9 Deep Learning Architec… 📚 Series Index (series end) The final part. So far we've focused on models. Now we focus on the tools that actually run them . We'll lay out the philosophical difference between PyTorch and TensorFlow, and trace how Hugging Face transformers , llama.cpp , MLX , and Ollama built the bridge to running large language models on your own machine. By the end you should have the full mental model of "download a pretrained LLM and serve it locally." 0. Learning Objectives Explain the eager-vs-graph difference between PyTorch and TensorFlow. Explain, in graph terms, how autograd automates the backward pass. Use Hugging Face transformers ' AutoModel , AutoTokenizer , and pipeline abstractions. Describe llama.cpp's GGUF quantization, INT4 inference, and CPU-first flow. Describe Apple MLX's use of unified memory and how it differs from PyTorch. Run a local LLM with Ollama and call it through an OpenAI-compatible API. ...
Read more »

"RAG Core Study (14/26) — Evaluation Sets with RAGAS & DeepEval"

이미지
Series overview: Series index 📚 Series Index Without an evaluation set, every RAG improvement is a story, not evidence. RAG systems fail in more than one place: retrieval, context selection, answer generation, and grounding. That means "looks better" is not a metric. Part 14 explains how to build a golden dataset , how RAGAS and DeepEval fit into the workflow, and why teams must separate retrieval quality from answer quality if they want tuning decisions to hold up. 0. Prerequisites Part 13 reranking — system quality now depends on multi-stage ranking. Part 1 grounding — a correct-looking answer is not enough. Part 12 Hybrid — multiple retrieval settings need comparison on fixed data. 1. Learning Objectives Build a small but useful golden dataset for RAG. Distinguish retrieval metrics from answer metrics . Use RAGAS and DeepEval in the right roles. Avoid the common traps in judge-model-based evaluation. 2. 핵심 요약 An evaluation set for RAG is ...
Read more »

"ML Foundations (8/9) — Deep Learning Architectures: CNN, RNN, Attention"

이미지
← 7/9 Deep Learning Training 📚 Series Index 9/9 PyTorch vs TensorFlow → What did we keep adding on top of an MLP? Vision went CNN, sequences went RNN/LSTM, and eventually both converged on Attention and the Transformer. Each architecture is easier to remember if you read it as what it gave up to gain something else . This part is the one-line summary of the decisive inflection points: LeNet → AlexNet → ResNet → LSTM → Attention → Transformer. 0. Learning Objectives Explain why CNN's convolution, pooling, and weight sharing match images so well. Trace what got unblocked from LeNet → AlexNet → VGG → ResNet as depth grew. State BPTT for RNNs and the vanishing/exploding gradient problem. Write the LSTM gate equations (forget, input, output) with intuition. Write the attention formula in query/key/value form and the scaled dot-product variant. State the precise reasons Transformers replaced RNNs (parallelism + long-range dependencies). 1. 핵심 요약 CNN : weight sharin...
Read more »

"ML Foundations (7/9) — Deep Learning Training: Optimizers, Regularization, Initialization"

← 6/9 Neural Networks 📚 Series Index 8/9 Deep Learning Architec… → Building an MLP, as we did in Part 6, is not the same as getting it to train well . The right optimizer, regularization, initialization, and learning rate — when these line up, deep networks converge. When they don't, the network refuses to learn at all. This part is the catalog of those crafts, with formulas, paper citations, and code in one place. 0. Learning Objectives Compare and write the update rules for SGD, Momentum, Nesterov, and Adam. Explain how Dropout, BatchNorm, and LayerNorm work and where they belong in a model. Derive the variance formulas for Xavier and He initialization and match them to activations. Implement step, cosine, and warmup learning-rate schedules in PyTorch. Explain why gradient clipping is effectively required for RNNs and Transformers. Diagnose the most common training failures (NaN, plateau, overfitting) and apply first-line fixes. 1. 핵심 요약 SGD : \(w \leftarro...
Read more »

OpenClaw to Hermes Migration (2/13) — What to Preserve, Partially Port, or Discard

← 1/13 Current Structure Inve… 📚 Series Index 3/13 What Moves to Hermes → A code review record classifying operational scripts into three tiers: preserve, partial-preserve, and discard. What This Post Covers An asset classification framework (preserve / partial-preserve / discard) for migrating a legacy agent system to a successor A concrete case applying the Strangler Fig + Subset Migration strategy to a memory pipeline Classification results evaluated across LOC, external dependencies, and replaceability axes A method for slimming down a bloated single file (1,310 LOC) using the migration as a natural entry point Problem Definition: Port Everything or Cut Deep? The OpenClaw inventory comprises multiple agents, multiple scheduler/daemon processes, and multiple operational scripts. When moving to the successor system, Hermes, the choice is not binary. A full port carries accumulated technical debt along with it; a minimal port discards hard-won domain logic. Code-revie...
Read more »

AI Agents I Built (5/7) — Building an Automated Blogger API Publishing System

← 4/7 My Life Organizer Agent 📚 Series Index 6/7 Dual → Designing a Markdown-Based Auto-Publish Pipeline with Blogger API v3 + OAuth 2.0 Key Summary Blogger API v3 + OAuth 2.0 enables a pipeline that converts Markdown files into publishable HTML posts blogger_publish.py handles the full flow — frontmatter parsing → body cleanup → HTML conversion → CSS injection → API post — in a single script For bulk publishing, quota management (delay, retry, TOC suppression) is the critical design concern Platform Selection Criteria API access is mandatory for AI agent-driven content publishing. A comparison of platforms: Platform API Cost AdSense Decision WordPress.com REST API Paid plan required Paid only Cost overhead Ghost REST API Self-hosted or paid Manual setup Ops overhead Blogger v3 API Free Native integration Adopted Blogger is free, exposes a REST API, and integrates natively with AdSense. For personal blog automation, this combination i...
Read more »