Skip to content

ROADMAP.md

Latest commit

 

History

History
75 lines (51 loc) · 4.86 KB

ROADMAP.md

File metadata and controls

75 lines (51 loc) · 4.86 KB
Raw

Roadmap

The DevOps Error Library is built on one durable foundation: a tree of plain Markdown files under errors/, compiled by errlib into a JSON search index. Every item on this roadmap reads that same corpus and index — no rewrites, no lock-in. We grow the content, and tooling layers on top.

We're at ~167 documented errors across 14 technologies today, architected to scale to 10,000+. Here's how we get there.

Want to help? See CONTRIBUTING.md. The single highest-leverage thing you can do is add accurate error pages.

🟢 Near-term — grow & harden the corpus

The priority is breadth and depth of high-quality, verified errors, plus the tooling that keeps the corpus trustworthy.

Per-technology targets

Stretch goals for the most-requested technologies (current counts in errors/README.md):

Technology Now Target
Kubernetes 16 200
Docker 12 150
Terraform 16 200
OpenStack (all services) 21 250
Linux 12 150
GitLab 12 100
Prometheus / Grafana 22 150
RabbitMQ / Redis 20 120
PostgreSQL / MySQL 20 150
Ceph / Linstor 16 120

These are directional, not gates — accuracy still beats quantity. Batches are coordinated via tracking issues (see CONTRIBUTING.md).

Corpus quality

  • Category indexes — keep scripts/gen_indexes.py output fresh on every merge (per-category README.md + the root index).
  • Link checking — CI step that verifies every relative Related Errors link and external References URL resolves.
  • Freshness tracking — surface pages with a stale last_reviewed date so they get re-verified.
  • Front-matter & schema tightening — broaden errlib validate checks (tag hygiene, related slugs that actually exist, severity sanity).
  • New technologies — add the next wave of folders as contributors bring verified errors (CI/CD runners, Helm, ArgoCD, Nginx, HAProxy, etcd, cloud-provider APIs, …).

🟡 Mid-term — make the corpus searchable everywhere

Once the index is rich, expose it through more surfaces — all reading errlib's JSON index.

  • CLI search tool packaging — publish errlib to PyPI (pip install devops-error-library) and ship prebuilt binaries so the local fuzzy search is a one-line install. (The CLI itself already exists via errlib search.)
  • Static docs site — generate a browsable, SEO-friendly site directly from the Markdown corpus (one URL per error), so pages rank for "Kubernetes / Terraform / Docker / OpenStack troubleshooting" searches.
  • Web search UI — a hosted search front-end over the JSON index: paste a log line, get matching pages, filter by technology/severity/tag.
  • REST API — a thin HTTP service over the JSON index (/search, /errors/<slug>, /stats) so other tools and dashboards can query the library programmatically.
  • GitHub Action — drop-in Action that, on a failed CI job, searches the library for the failure's error string and comments the matching error pages on the run — troubleshooting where the failure happens.

🔵 Long-term — intelligence & ecosystem

With a large, structured, verified corpus and an API, build the smart layer on top.

  • AI troubleshooting suggestions — grounded, retrieval-based suggestions that cite specific library pages (never ungrounded guesses), bringing the structured corpus to the AI Incident Response Assistant.
  • MCP server — a Model Context Protocol server so AI agents and IDE assistants can query the library as a first-class tool (search_errors, get_error), with citations back to the corpus.
  • VS Code extension — search and read errors without leaving the editor; surface matching pages from terminal output and log files.
  • Community voting — let readers mark whether a page resolved their issue, feeding a quality signal back to maintainers.
  • Error popularity rankings — aggregate search and voting data into "most-hit errors per technology" and trending pages, guiding where to deepen coverage next.

🧭 Guiding principles

  1. Markdown is the source of truth. Every feature reads errors/**/*.md and the generated JSON index. If a feature needs a separate database of content, we've taken a wrong turn.
  2. Accuracy over quantity. Growth never lowers the bar — verified logs, read-only diagnostics, real fixes.
  3. Offline-first. Core value (the corpus + errlib search) always works with no network and no account.
  4. Open by default. Content stays CC BY 4.0; tooling stays MIT.

Have an idea or want to own a roadmap item? Open an issue or read CONTRIBUTING.md. 🚀