Local AI Infrastructure Notes (5/15) — Home Server Dashboard Design with Next.js

Visualize only what agents send. Never control them directly.

Key Takeaways

  • In a multi-agent environment, tracking state via individual terminals does not scale. A dashboard becomes essential infrastructure as operational scope grows.
  • Next.js + SQLite is a practical fit for home server dashboards. Server-side rendering eliminates a separate frontend build, and there are no external DB dependencies.
  • The core design principle is read-only: adding control logic to a dashboard causes complexity to explode.

Why Multi-Agent Monitoring Matters

As the number of agents grows, answering questions like "which agents are currently active?", "when did the last insight arrive?", or "has a stock signal come in recently?" requires manually digging through multiple terminals and log files. Beyond a certain operational threshold, this approach becomes unworkable.

The dashboard's role is to aggregate that information into a single interface. The design driver is operability, not technical interest.

Body

1. Why Next.js + SQLite

Three criteria drove the stack selection:

  1. Single-process operation — Home server resources are constrained. Running a separate DB server process was not acceptable.
  2. Server-side rendering — API and frontend managed within a single project.
  3. Fast prototyping — With agent operations already urgent, spending time on infrastructure was not an option.

Next.js satisfies criteria 1 and 2. API Routes provide backend endpoints, and page rendering lives in the same project. SQLite satisfies criterion 3: a single file is the entire database. No server process like PostgreSQL or MySQL is required.

Alternatives evaluated and rejected:

  • Grafana + Prometheus: Monitoring-only with limited custom widget support. Poorly suited for unstructured data like agent messages.
  • Express + React: Frontend/backend separation is overengineering for a home server.
  • Static HTML + cron generation: No real-time capability.

2. Dashboard Sections

The dashboard is organized into four sections.

Agent Status Panel

Displays registered agents as cards showing current status: Active, Suspended, or Complete. Each card includes the agent name, role, tech stack, and last activity timestamp.

Status is determined from webhook receipt history. An agent is marked Active if an insight or message arrived within the most recent polling window; otherwise, the default status from the agent's configuration file is shown. The active threshold is adjustable per agent type in the config.

Insight Timeline

Lists insights sent by agents via webhooks in chronological order. Each entry shows the source agent, tags, and timestamp. Filters allow narrowing to a specific agent or tag.

This timeline is the dashboard's core. It answers "what is happening in the system right now" at a glance.

Stock Signal Widget

Stock-related signals are isolated in a dedicated widget. It displays ticker, signal type (Buy / Sell / Hold), score, and receipt time. Separation from the general insight timeline is justified by different update frequency and the need for immediate visual recognition.

Agent Message Log

Agent-to-agent messages are displayed as a log: sender, recipient, message type (task_request, insight, status_update, etc.), and ACK status. Unacknowledged messages are visually highlighted.

3. Data Flow — Webhook Ingestion → SQLite → Rendering

The full data flow is unidirectional:

Agent → POST /api/webhooks → SQLite write → Dashboard render

Agents POST insights and state changes as webhooks. The API endpoint receives them, writes to SQLite, and reads the latest data at page render time.

Real-time updates use polling. WebSocket was considered, but agent event frequency is on the order of a few times per minute — 30-second polling is sufficient. The added complexity of WebSocket yields no operational benefit at this scale.

4. API Endpoint Design

Three core endpoints make up the dashboard API.

POST /api/webhooks — Insight ingestion

{
  "content": "Qwen3 model benchmark complete. 15% performance improvement over previous baseline.",
  "tags": ["benchmark", "qwen3"],
  "source": "model-manager"
}

Fire-and-forget endpoint for agents. No ACK required. Incoming payloads are timestamped and written to SQLite.

GET /api/events — Event retrieval

Read endpoint called by the dashboard frontend. Supports query parameter filtering by source, tag, and time range. Pagination included.

POST /api/agent-messages — Inter-agent messages

{
  "fromAgent": "research-agent",
  "toAgent": "quality-agent",
  "messageType": "task_request",
  "intent": "fact-check requested",
  "body": { "text": "Source file verification requested" },
  "dedupeKey": "fc-20260405-001"
}

Unlike webhooks, this endpoint requires ACK: bidirectional communication. The receiving agent must confirm receipt. Unacknowledged messages are visually flagged in the dashboard.

5. Design Principle — "Visualize Only. Never Control."

The most important principle upheld throughout dashboard design: the dashboard visualizes what agents send; it does not control them directly.

Adding agent start/stop or task assignment functionality introduces three problems:

  1. Control creates state synchronization issues — If the dashboard stops an agent but the agent self-recovers, state diverges.
  2. Each agent has a different control interface — Presenting a unified control surface across Flutter apps, Python scripts, and Node.js servers carries high abstraction cost.
  3. Read-only means bugs are non-fatal — A bug in the dashboard does not affect agent operation.

This principle keeps dashboard development simple: receive, store, display. Three operations only.

Agent control, when needed, is handled directly within each agent's own project. The dashboard is for observing the system, not operating it.

Troubleshooting Notes

SQLite concurrent write contention. When multiple agents POST webhooks simultaneously, "database is locked" errors occur. Enabling WAL (Write-Ahead Logging) mode resolves this. A single line — PRAGMA journal_mode=WAL; — activates it, and concurrent read/write throughput improves significantly.

Polling interval tuning. An overly short polling interval generates excessive SQLite read load. Agent events do not require sub-second real-time delivery, so 30-second polling is operationally appropriate. High-frequency widgets can have independent polling intervals configured separately.

Agent name normalization. Agent identifiers containing non-ASCII characters or spaces cause repeated encoding issues in URL parameters, filters, and SQLite queries. The internal layer uses English slugs exclusively; the display layer maps slugs to human-readable names.

Summary

Core design decisions for the home server dashboard:

  1. Use Next.js + SQLite for a single-process dashboard.
  2. Organize into four sections: agent status, insight timeline, stock signals, message log.
  3. Maintain unidirectional flow: webhook ingestion → SQLite write → page render.
  4. Keep the dashboard read-only. Do not add agent control functionality.

The last point is the most critical. The temptation to add control to a dashboard is strong, but the moment it is added, a simple monitoring tool becomes a complex orchestration platform. Monitoring and control must be separated at the design layer.

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 »