A complete guide to building a fully self-hosted AI assistant using Google Gemma 4 open source models, SearXNG private search, and OpenClaw for agentic tool calling — all running locally with zero data leaving your machine.
The full call stack from typing hello in Claude Code to receiving a response. Data flow and AsyncGenerator chain architecture analysis across 11 files from cli.tsx to toolExecution.ts
A complete dissection of the 7 continue paths in query.ts while-true loop and StreamingToolExecutor 4-state machine with 3-tier concurrency model validated through a Rust prototype
Classifying 42 tools into 8 families and deep-diving into BashTool 6-layer security chain and AgentTool 4 spawn modes
Analyzing the toolHooks.ts runtime hook pipeline and 85 React hooks classification system plus CLAUDE.md 6-stage discovery and token budget management
Analyzing MCP client.ts 3348-line connection management and OAuth 2465-line authentication plus the plugin vs skill distinction and 5-layer skill discovery engine
A full retrospective on designing and building an independent agent harness in Rust that overcomes the 8 limitations discovered during Phase 1-2
An overview of Google Veo 3.1 video generation capabilities and the growing ecosystem of video AI background removal tools including VideoBGRemover and n8n automation workflows
Added OpenTelemetry distributed tracing to the FastAPI server, set up Grafana Alloy on EC2 to forward traces to Grafana Cloud Tempo, and built a traced_span context manager that auto-captures CPU time and memory deltas per span
A practical guide to adding distributed tracing to a FastAPI service using OpenTelemetry and Grafana Cloud, including the debugging journey from broken traces to working observability
