DEV Community: sahaabhishek

Local dashboard for monitoring AI agent sessions — no cloud, no account

sahaabhishek — Sat, 16 May 2026 23:23:44 +0000

I've been running multiple AI coding agents in parallel — Claude Code sessions, sub-agents spawned mid-task, OpenCode running alongside — and I couldn't keep track of any of it.

So I built CrewView: a local dashboard that gives you a real-time view of every AI agent session running on your machine.

Here's what it does:

→ Sessions list — every conversation, with status, duration, and a prompt preview
→ Kanban board — split across Running / Needs Approval / Needs Input
→ Live view — agent cards updating in real time, with sub-agents shown nested under their parent
→ Sub-agent tracking — when Claude Code spawns a Task agent, it appears as a child session
→ Workflow builder — define multi-step agent pipelines and trigger scheduled runs
→ Multi-source — Claude Code (reads local JSONL files) and OpenCode (HTTP API) at the same time

The entire thing is a single Go binary with the React dashboard embedded inside. Install with one command:

curl -fsSL https://raw.githubusercontent.com/ovsec/crewview/main/install.sh | sh

No Docker, no cloud account, no telemetry you didn't opt into. All data stays in a local SQLite file.

It's MIT licensed and on GitHub: github.com/ovsec/crewview

If you're spending serious time with AI coding agents, I'd love to hear what visibility you wish you had. Happy to add it.

AI #DeveloperTools #OpenSource #ClaudeCode #AIAgents

Meet httpOwl: Run HTTP Requests From Any Webpage

sahaabhishek — Wed, 13 May 2026 13:01:26 +0000

If you use httpYac or the IntelliJ .http format, you've probably run into this situation: someone shares a request example in a GitHub issue, a wiki page, or a PR description. To test it, you copy-paste into cURL or open your IDE.

httpOwl eliminates that step. It's a browser extension that detects httpYac-format request blocks on any webpage and lets you run them inline.

How it works

Open any page containing .http format blocks — GitHub blob views, Confluence docs, internal wikis, raw files
httpOwl auto-detects them and places small pill buttons next to each request
Click → the request fires through the extension's service worker (no CORS issues)
Results appear inline: status code, response time, assertion pass/fail
Click again to open a detail overlay with full request/response headers and body

Features

Auto-detection — parses ###-separated blocks from <pre>, <code>, <textarea> elements
Heading+code strategy — handles Confluence/wikis where ### renders as HTML headings
Inline pill buttons — subtle glass UI with states: idle, running, success, error
Assertion engine — evaluates ?? lines client-side: status == 200, body.items isArray, header content-type includes json, duration checks, regex, etc.
Detail overlay — full request/response with JSON pretty-printing and copy-as-cURL
Side panel — request list with HTTP method filtering, multi-select run, environment variables
HTML report — download a self-contained dark-themed report after running requests
CORS bypass — fetches originate from the extension, not the page

Variable system

# $shared
API_KEY=abc123
BASE_URL=https://api.example.com

# dev
BASE_URL=https://dev.api.example.com
TOKEN=dev-token

Variables with {{KEY}} syntax are substituted in URL, headers, and body. Requests with unresolved variables are visibly muted and disabled until you define them.

Technical stack

Manifest V3
TypeScript + Vite
Playwright for screenshot/icon generation
Shadow DOM for UI isolation

Open source

MIT licensed. Contributions welcome.

Chrome Web Store: (Edge support coming soon)
httpOwl Extension

GitHub: https://github.com/ovsec/httpyac-browser

httpYac: https://httpyac.github.io

I built a resume compiler in Go — write YAML, get every format

sahaabhishek — Mon, 27 Apr 2026 21:34:48 +0000

I got tired of fighting Word docs and Canva templates every time I updated my resume.

Your resume is structured data. A name, jobs, dates, skills. Once I realized that, I stopped treating it like a document and started treating it like a data model with renderers. That's resumelang.

How it works

Write one resume.yml:

# yaml-language-server: $schema=https://resumelang.dev/schema/v1.json
resumelang: v1

meta:
  theme: sap          # 13 built-in themes
  language: en
  page_size: a4
  sections:
    - summary
    - experience
    - skills
    - education
    - projects

person:
  name:     Jane Doe
  title:    Senior Software Engineer
  email:    jane@example.com
  github:   janedoe
  linkedin: janedoe

summary: |
  Experienced engineer with 8 years building distributed systems.

experience:
  - company:  Acme Corp
    role:     Lead Engineer
    start:    "2021"
    end:      ""
    highlights:
      - Scaled API to 10M requests/day
      - Reduced CI deploy time by 60%
    tags: [Go, Kubernetes, Postgres]

skills:
  - category: Languages
    skills: [Go, TypeScript, Python, Rust]

Run one command:

resumelang build resume.yml

Get:

dist/
├── resume.html   # themed, print-ready
├── resume.json   # JSON Resume schema
├── resume.md     # GitHub-flavored markdown
└── resume.txt    # ATS-safe plain text

The theme system

13 built-in themes: SAP, Vercel, Linear, Aurora, Material, Brutalist, Terminal, Editorial, Dossier, Academic, Finance, Timeline, Minimal.

Switch with one line:

meta:
  theme: vercel

Your content doesn't move. Only the rendering changes.

Each theme is a folder with a validated theme.yml spec:

spec: v1
name: aurora
version: 1.0.0
author: ovsec

supports:
  layouts: [timeline, grid]
  blocks: [bullets, tags, text]

tokens:
  colors:
    accent: "#6366f1"
  spacing:
    section_gap: 24
  typography:
    heading_weight: 600
    body_size: 14

The compiler validates every theme at load time against the spec. Missing template → build error. Unknown component → warning. This means community themes that pass resumelang theme validate are guaranteed to render correctly.

Why Go

Single binary, zero runtime dependencies
Same binary serves the web editor, runs the CLI, and powers the GitHub Action
Cross-compiles to Linux/macOS/Windows without Docker
go install github.com/ovsec/resumelang@latest and you're done

The web editor

The editor is Go Fiber + HTMX + CodeMirror 5. No React, no build step.

Edit YAML on the left, live preview on the right
Theme gallery rendered server-side via HTMX lazy loading
Share modal with gzip-encoded URLs and AES-GCM password encryption
Decryption happens entirely in the browser — server never sees resume content
Sign in with GitHub or LinkedIn to save resumes to a dashboard

GitHub Actions path

Drop this into .github/workflows/resume.yml:

on:
  push:
    paths: ["resume.yml"]

jobs:
  publish:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-go@v5
        with: { go-version: "1.22" }
      - run: go install github.com/ovsec/resumelang@latest
      - run: resumelang build resume.yml --out dist
      - uses: actions/upload-pages-artifact@v3
        with: { path: dist }
      - uses: actions/deploy-pages@v4

Push resume.yml, get a live GitHub Pages site. Every change is a commit. git diff shows exactly what changed.

What's on the roadmap

Native PDF export (Puppeteer server-side)
resumelang lint — ATS keyword score, date-gap detection
LinkedIn import (resumelang import linkedin export.zip)
AI tailoring — paste a JD, rewrite highlights to match (Pro)
Link analytics — see when recruiters open your resume (Pro)

Contribute

The theme ecosystem is community-driven. If you have design opinions, the spec is documented at docs.resumelang.dev/themes. Submit a theme, open an issue, or just star the repo.

🔗 resumelang.dev
📖 docs.resumelang.dev
⭐ github.com/ovsec/resumelang

What would you want from a YAML-first resume tool? Drop it in the comments.