Agent Memory Engine (6/10) — Comparing 5 Vector DBs Across 7 Dimensions: sqlite-vec

The decision matrix itself is the content

ํ•ต์‹ฌ ์š”์•ฝ

  • This post provides a 7-dimension evaluation framework for selecting a vector DB in a single-user agent system.
  • Five candidates (sqlite-vec / ChromaDB / Qdrant / Weaviate / custom FAISS) evaluated across 7 dimensions: Portability / Rollback / Performance / Integration / Dependencies / Operational Complexity / Cost.
  • Result: sqlite-vec + bge-m3-mlx-fp16 (1024d) ranked first. The decisive dimensions were not search speed but rollback capability and single-file portability.

What This Post Covers

A vector DB selection is a decision that follows you for at least a year. Most comparison resources focus on search speed or QPS, which misaligns with the decision axes that matter in a single-user system. This post covers three things:

  1. A scoring template for comparing 5 candidates across 7 dimensions
  2. Which dimensions actually drive the decision in a single-user environment (spoiler: not performance)
  3. How local embeddings (bge-m3 + MLX fp16) fit into this framework

The Five Candidates

Candidate One-liner
1 sqlite-vec SQLite extension. Single file. C implementation.
2 ChromaDB Python-friendly. Embedded mode available.
3 Qdrant Rust server. HNSW. Production-oriented.
4 Weaviate Go server. Modular backend.
5 Custom FAISS Index managed directly in Python.

7-Dimension Scorecard (5-point scale)

Dimension sqlite-vec ChromaDB Qdrant Weaviate FAISS
Portability 5 3 2 2 4
Rollback Ease 5 3 2 2 3
Performance (single-user baseline) 4 4 5 5 5
Hermes Integration Difficulty 4 3 3 3 3
External Dependencies 5 3 1 1 4
Operational Complexity 4 3 2 2 3
Cost (self-hosted) 5 4 3 3 5
Total 32/35 23/35 18/35 18/35 27/35

(Initial tally was sqlite-vec 28/35; recalculation yields 32/35. Either way, the ranking is identical.)

Why Performance Is Not the Deciding Dimension

Query load in a single-user system runs at roughly 5–20 requests per second. The performance headroom of an HNSW-based distributed server carries no practical meaning at this throughput. Performance is therefore included in the scorecard, but classified as an unweighted dimension.

The two dimensions that determine the outcome are as follows.

Dimension 1: Portability

sqlite-vec stores everything in one DB file.

  • Backup: cp memcore.db backup.db — a single line.
  • Migration: copy the file to another machine.
  • Version control: compare snapshots via a single file hash.

Every other candidate separates "server process + data directory + config files," making backup and migration multi-step operations. In a single-user environment, this complexity accumulates as long-term operational cost.

Dimension 2: Rollback

The highest risk during migration is inability to revert after a partial failure. sqlite-vec transactions carry the same semantics as SQLite transactions:

BEGIN;
-- migration queries
ROLLBACK;  -- full revert in one line

Distributed-index candidates (Qdrant, Weaviate) can emit partial commits during index rebuilds, making a clean revert from a mid-run failure non-trivial. This difference is decisive in a single-user system.

Embedding Model Selection: bge-m3 + MLX fp16

The embedding model must be locked in alongside the vector DB. Selection criteria and rationale:

Axis Choice Rationale
Model bge-m3 Robust on Korean/English mixed input; 1024 dimensions balance storage cost and expressiveness
Runtime MLX fp16 Apple Silicon native; no separate GPU required
Hosting Local No external API calls → zero token cost, zero network dependency

For a single-user system, this combination is a reasonable default at the current point in time. The single tradeoff: model update cadence is slower than an external API.

Scope and Limitations

This framework is valid only under the following conditions:

  • Single user or small team (QPS < 50)
  • Total vector count in the low millions or below (sqlite-vec recommended upper bound)
  • Hardware capable of local execution (Apple Silicon or equivalent)

Outside these conditions, the weights across dimensions shift. In multi-tenant / high-QPS / tens-of-millions-vector environments, the penalties for Qdrant and Weaviate recover, and the portability premium diminishes.

Summary

Vector DB comparisons are often compressed into a single QPS benchmark. However, in a single-user system, the deciding dimension is not performance — it is how simply you can back up and roll back. sqlite-vec ranked first in the 7-dimension matrix because it scored maximum points on those two dimensions; the remaining five dimensions served as tiebreakers.

The open question: when QPS crosses from single digits into the high double digits, or when vector count exceeds tens of millions, which dimension in this framework breaks first — and that is the scenario requiring re-evaluation against the same matrix.

Series overview: Series index

๋Œ“๊ธ€

๋Œ“๊ธ€ ์“ฐ๊ธฐ

์ด ๋ธ”๋กœ๊ทธ์˜ ์ธ๊ธฐ ๊ฒŒ์‹œ๋ฌผ

"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 »

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 (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 »

"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 (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 »

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 »

"ML Foundations (6/9) — Neural Networks: From Perceptron to MLP"

์ด๋ฏธ์ง€
← 5/9 Regularization & Model… ๐Ÿ“š Series Index 7/9 Deep Learning Training → A neural network is logistic regression from Part 4 stacked into layers. Each layer passes through a new nonlinear activation, expanding the family of functions the network can represent. In this part we put how a neural network learns onto one page. Perceptrons, XOR, MLPs, backpropagation, activation functions, and the universal approximation theorem all meet here. 0. Learning Objectives Write the perceptron's learning rule and explain why it cannot solve XOR. Express an MLP's forward pass as one line of matrix products. Derive backpropagation by hand on a two-layer MLP. Compare the definitions, derivatives, and trade-offs of sigmoid, tanh, ReLU, and GELU. State the universal approximation theorem (Cybenko 1989 / Hornik 1991) precisely. Train an MLP end-to-end in PyTorch. 1. ํ•ต์‹ฌ ์š”์•ฝ Perceptron (1958): \(\hat{y} = \mathrm{step}(w^\top x + b)\). Converges only on linearly separabl...
Read more »