This repo produces multi-concept learning roadmaps that bridge online courses to a build-from-scratch project (ZeroServe). Each roadmap is a single Markdown file converted to PDF.
roadmap/
├── FORMAT.md # Per-concept template (READ FIRST)
├── zeroserve-in-public.md # LLM inference engine curriculum (101 concepts)
App boots, routes navigate, mock data layer in place, AI/EditBus primitive types defined. Studio route renders an empty shell.
Before this phase, ovk-web/ is a bare Vite + React 19 + TanStack Router scaffold with several blockers that make every later phase harder:
src/router.tsx and src/main.tsx both create a router with subtly different options (defaultPreloadStaleTime only in router.tsx). One is dead code; later phases won't know which to extend.
A working engineer's reference for the HyperFrames open-source video framework (Apache 2.0). Covers the architecture from high-level package flow down to the deterministic capture loop, shows how to build videos programmatically with the SDK, and walks through a flagship contribution: centralizing the duplicated skill-layer TTS providers into the CLI and adding Gemini TTS via OpenRouter as a new audio provider.
All claims below are grounded in source files in the cloned repo at /Volumes/data/workspace/video/hyperframes/.
File references use path:line form. Where a fact is inferred rather than read directly, it is called out.
Goal: Build a mathematically perfect Transformer block from scratch, guided by
tiny-llm(MLX/Apple Silicon) andnanoGPT(PyTorch/eager).
graph LR
subgraph loop ["Autoregressive Loop"]
A[Input Tokens] --> B[Embedding Lookup]
B --> C[N x Transformer Blocks]
C --> D[Final RMSNorm]
D --> E[LM Head]
How to build MCP servers with FastMCP 3.x and mount them inside FastAPI.
Reference: The complete FastMCP documentation index is available at https://gofastmcp.com/llms.txt. Fetch it to discover all available pages, API references, and integration guides.
| version: '3' | |
| services: | |
| socks5: | |
| image: serjs/go-socks5-proxy | |
| container_name: socks5 | |
| restart: always | |
| # Host mode bypasses the Docker bridge, solving the IPv6 'network unreachable' bug | |
| network_mode: "host" | |
| env_file: .env |
Nomad generally uses less CPU and RAM than K3s for the same basic cluster functionality. Nomad is widely regarded as one of the most efficient and lightweight orchestrators available.
For your 5-VPS setup, Nomad's single-binary, minimal approach will result in lower resource utilization and operational overhead compared to K3s, which, despite being "lightweight," still runs the full set of Kubernetes control plane components.
The difference in resource consumption stems from the architectural complexity of the two systems.
| Resource | Nomad Agent (Server/Client) | K3s Agent / Server |
|---|
| #!/bin/sh | |
| curl -s https://bunnycdn.com/api/system/edgeserverlist -H "Accept: application/json" | jq -r .[] > /tmp/bunny_ips | |
| echo "" >> /tmp/bunny_ips | |
| curl -s https://bunnycdn.com/api/system/edgeserverlist/ipv6 -H "Accept: application/json" | jq -r .[] >> /tmp/bunny_ips | |
| for ip in `cat /tmp/bunny_ips`; do ufw allow proto tcp from $ip comment 'Bunny IP'; done; |
| curl -sL yabs.sh | bash -s -- -r9 | |
| 3x vCPU Cores | |
| 75 GB PURE SSD RAID-10 Storage | |
| 4 GB RAM | |
| 10,000GB Monthly Premium Bandwidth | |
| 1Gbps Public Network Port | |
| Full Root Admin Access | |
| 1 Dedicated IPv4 Address | |
| KVM / SolusVM Control Panel - Reboot, Reinstall, Manage rDNS, & much more |