OpenClaw to Hermes Migration (4/13) — Code Review That Identified 8 Assets Worth Migrating

Average 175 LOC, Zero External Dependencies, One File Accounting for 30% of Migration Value

ํ•ต์‹ฌ ์š”์•ฝ

  • Line-by-line review of preserve (๐Ÿ’Ž) candidates: 8 preserved, 2 partially preserved, 1 discarded.
  • Migration difficulty is low — average 175 LOC, standard library only, zero external dependencies.
  • user-pattern-stage.py alone accounts for approximately 30% of the migration's total value. No equivalent exists in Hermes.

This post covers asset review techniques for legacy → new platform migrations: how to classify assets as preserve/partially-preserve/discard using LOC, external dependency count, and functional replaceability as axes, and the pattern of reducing migration scope through LOC-based triage.

Prerequisites for Asset Review

Before entering the code review, migration targets, method, and boundaries must be fixed — otherwise classification decisions become unstable. The premises for this migration are as follows.

  • Strategy: Strangler Fig + Subset Migration. Progressively encircle the original while migrating only a subset.
  • Memory backend: SQLite + Python library (memcore). No separate daemon.
  • Folder boundary: ~/Hermes/ and ~/Project are separate. The original is locked as a read-only backup.
  • Agent roles: Hermes Manager (new) and OpenClaw Manager (backup-only) are separated.
  • Cutover model: Dual execution → verification → cutover.

These premises are what make the question "is it worth migrating?" meaningful, as opposed to merely "can it be migrated?"

Classification Criteria

Each asset is evaluated on three axes.

  1. LOC — Primary indicator of migration cost.
  2. External dependencies — Standard library only, or external packages attached?
  3. Replaceability on the new platform — Discard if equivalent functionality already exists; preserve if it does not; partially preserve if only some overlap exists.

๐Ÿ’Ž Preserve — 8 Assets

File LOC Migration Form Value
user-pattern-stage.py 260 memcore.dialectic module Highest. U-tag 3-phase, anti-bias guard
confidence-decay.py 82 memcore cron tick opinions confidence decay 0.02 per tick, remove if <0.30
topics-validate.py 131 memcore consistency check bank ↔ topics consistency
topics-expand.py 186 memcore auto-registration auto-registers new files
bank-lint.py 176 memcore validator Retain tag format validation
entity-audit.py 152 memcore stale detection 30-day stale, equivalent via SQLite timestamp query
docs-snapshot.sh 141 Hermes cron payload docs change detection
release-check.sh 66 Hermes cron payload version change detection

The highest-value asset is user-pattern-stage.py. It separates hypotheses about the user into three stages — observe → hypothesize → validate — with an anti-bias guard that prevents assertions before validation is complete. The approach is inspired by Honcho-class techniques but reimplemented for the target data model. The new platform has no equivalent dialectic structure. This single file represents approximately 30% of the justification for the migration itself.

๐ŸŸก Partially Preserve — 2 Assets

File LOC Handling
bank-size-watch.py 176 Weakened by SQLite transition. Extract only the MemTree split logic portion
proactive-exception-alerts.py 1310 Too large. Extract only the 11 escalation-reason classification logic (~300 LOC); delegate alert routing and context collection to Hermes gateway/hook

proactive-exception-alerts.py is a textbook case of partial preservation. Opening the 1310 LOC reveals that the classification logic is approximately 300 lines; the remainder is delivery infrastructure. Since the new platform already provides delivery infrastructure, only the classification logic is extracted and migrated, while the infrastructure is delegated. The result is a reduction from 1310 → 300 LOC, simultaneously lowering both migration cost and maintenance debt.

This pattern is reusable: when a large script mixes "classification core / delivery shell," and the new platform provides the shell, extract only the core. In many cases, 70–80% of LOC turns out to be shell.

๐Ÿ—‘ Discard — 1 Asset

File LOC Replacement
heartbeat-tick.py 262 Hermes cron + on_turn_start hook provides 100% coverage

Assets with a confirmed 1:1 replacement path are not preserved. They are terminated with only a backup retained. The discard criterion is "100% functional coverage + equivalent operational model."

Measurements and Patterns

  1. Why migration difficulty is low: Average 175 LOC, standard library only, zero external dependencies. Replacing the common I/O layer (lib_common.py) with memcore.io is sufficient for most files to operate as-is.
  2. Effort estimate: The code migration for the 8 assets itself takes 1–2 days. The dominant cost is not code — it is verification (testing and dual execution).
  3. Concentration of value: Among 11 candidates, a single file (user-pattern-stage.py) accounts for approximately 30% of total value. Prioritizing review time on this file yields the best efficiency.
  4. Dietable assets: One file can be reduced from 1310 → 300 LOC. There are cases where migration simultaneously functions as debt reduction.

Limitations and Open Questions

  • LOC is only an approximation of migration cost. Files with high state dependency carry large verification costs even at low LOC.
  • The "30% value" estimate is a qualitative judgment based on functional overlap and replaceability. Converting it to a quantitative metric would require actual call frequency data and a dependency graph.
  • Having chosen Strangler Fig, observing data inconsistencies during the dual-execution period is the next task.

Applicability

This review pattern is reusable under the following conditions.

  • Legacy code is spread across dozens of scripts and carries few external dependencies.
  • The new platform provides common infrastructure such as delivery, scheduling, and I/O — making "extract the core only" separation feasible.
  • A decision structure exists that allows confidently discarding assets with 100% replaceability.

The next post covers reclassifying these 11 assets and remaining functionality into 11 categories, and scheduling them across a 5-week migration timeline.

Series overview: Series index

๋Œ“๊ธ€

๋Œ“๊ธ€ ์“ฐ๊ธฐ

์ด ๋ธ”๋กœ๊ทธ์˜ ์ธ๊ธฐ ๊ฒŒ์‹œ๋ฌผ

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

← 1/10 SQLite Memory Engine f… ๐Ÿ“š Series Index 3/10 memcore Completed → A memory engine that runs without daemons, without dependencies, anywhere ํ•ต์‹ฌ ์š”์•ฝ 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"

์ด๋ฏธ์ง€
← 8/9 Deep Learning Architec… ๐Ÿ“š 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"

์ด๋ฏธ์ง€
Series overview: Series index ๐Ÿ“š 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. ํ•ต์‹ฌ ์š”์•ฝ An evaluation set for RAG is ...
Read more »

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

์ด๋ฏธ์ง€
← 7/9 Deep Learning Training ๐Ÿ“š 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. ํ•ต์‹ฌ ์š”์•ฝ CNN : weight sharin...
Read more »

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

← 6/9 Neural Networks ๐Ÿ“š 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. ํ•ต์‹ฌ ์š”์•ฝ SGD : \(w \leftarro...
Read more »

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

← 1/13 Current Structure Inve… ๐Ÿ“š 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 ๐Ÿ“š 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 »