meta: widget catalog & roadmap

[unhappychoice]

Index page for every widget roadmap in splashboard. Individual fetcher / renderer candidates live in the sub-issues below; this issue holds cross-cutting context (composition model, prioritization rubric, ReadStore recipes, ecosystem reference).

No separate issue per fetcher / renderer. A PR that ships one ticks the box on the relevant sub-issue and references that issue. Things that do get their own issue are cross-cutting features (splashboard fetch CLI #39, splashboard cache CLI #40, theme system #17). ReadStore fetcher shipped via #43.

---

Sub-issues

Renderers

• meta: renderer roadmap #61 — meta: renderer roadmap — primitives (charts, gauges, QR, etc.) + animation families (animated_postfx via tachyonfx, animated_reveal TTE-style)

Fetchers

• meta: fetcher roadmap - Local / System #62 — meta: fetcher roadmap - Local / System — clock_*, system_*, disk_*, net_*, docker_*, k8s_*, code_*, project_*
• meta: fetcher roadmap - VCS / Forge #63 — meta: fetcher roadmap - VCS / Forge — git_*, github_*, gitlab_*, bitbucket_*, gitea_*, sourcehut_*, azure_devops_*, gerrit_*
• meta: fetcher roadmap - Coding & Build #64 — meta: fetcher roadmap - Coding & Build — ci_*, deploy_*, leetcode_*, stackoverflow_*, oss_*, package downloads, trending
• meta: fetcher roadmap - Cloud / Infra / Ops #65 — meta: fetcher roadmap - Cloud / Infra / Ops — aws_*, gcp_*, terraform_*, cloudflare_*, tailscale_*, uptime monitors, pagerduty_*, cloudwatch_* / datadog_* / sentry_*
• meta: fetcher roadmap - Communication & PM #66 — meta: fetcher roadmap - Communication & PM — slack_*, discord_*, email_*, matrix_*, jira_*, linear_*, asana_*, todoist_*, notion_*
• meta: fetcher roadmap - Social & Media #67 — meta: fetcher roadmap - Social & Media — RSS / news, mastodon_*, bluesky_*, youtube_*, twitch_*, spotify_*, plex_*, reading / watching, novelty image / text feeds
• meta: fetcher roadmap - Life & Ambient #68 — meta: fetcher roadmap - Life & Ambient — weather_*, calendar_*, holiday_*, travel_*, anki_*, health_*, strava_*, finance, monetization, analytics, home / IoT, gaming, public-API feeds

---

Composition model

A widget is Fetcher × Renderer × Layout slot. Fetchers produce Payload; renderers consume it. Most fetchers have a natural renderer pairing, but the two are swappable — same clock fetcher + text renderer = digital clock; + analog_clock renderer = analog face. Same github_contributions fetcher + heatmap renderer = 草; habit-tracker ReadStore + same heatmap renderer = habit grid.

Shapes are the only coupling between fetchers and renderers. Each Body variant corresponds to one Shape: Lines, Entries, Ratio, NumberSeries, PointSeries, Bars, Image, Calendar, Heatmap. See AGENTS.md for the full contract and the checklist to add a new shape.

---

Layering per item

• Fetcher kind — Fetcher (cache-backed, async) vs RealtimeFetcher (sync, per-frame). "Right now" values (clock, system_cpu, clock_countdown, system_uptime) are realtime; everything else is cached.
• Safety class — Safe (renders in untrusted local config), Network (gated if URL is config-provided). Fixed-host authenticated fetchers stay Safe — see project_trust_model.md.
• Built-in vs ReadStore — ReadStore (shipped feat: $HOME/.splashboard/ home-dir layout + ReadStore fetcher #43) renders any file the user writes at $HOME/.splashboard/store/<id>.<ext>. It's the escape hatch, not the default: prefer a dedicated built-in whenever the widget warrants curated UX (parser, per-user auth, protocol handling, rate-limit strategy, stateful update). ReadStore fits when the data is ad-hoc and the user is willing to produce the file themselves.

---

Ecosystem reference

Before building a renderer from scratch, check awesome-ratatui. Many primitive gaps already have community crates that we can wrap with a thin Renderer impl — a few lines mapping Payload → widget input.

Two animation paradigms, with distinct upstream libraries:

• Post-process effects — render the widget normally, then apply a shader-like pass (fade, color shift, glitch, slide, dissolve, sweep). tachyonfx (50+ effects) is the clear winner.
• Reveal animations — the widget's final form emerges from a particle-like animation (matrix rain decrypting into text, beams, fireworks). TTE is the algorithmic reference; not expressible via tachyonfx's buffer-post-process paradigm. tui-rain covers matrix/rain.

ansi-to-tui covers ANSI passthrough.

Full primitive-gap list lives in #61.

---

ReadStore recipes (not dedicated fetchers)

Config conventions for the shipped read_store fetcher; the user produces the file and read_store renders it. Listed here for discoverability, not for implementation tracking — they don't need a sub-issue because no code ships for each one.

• habit_tracker — daily 0/1 grid → heatmap
• goal_progress — single ratio
• journal_prompt — random line
• gratitude_list
• word_of_day
• reading_queue_random
• bookmark_random
• portfolio_summary
• spending_this_month
• invoice_due
• most_recent_file — MRU list
• command_history_top

---

Composition helpers (layout-level, not fetchers)

Multi-widget layouts that compose existing pieces. Tracked here rather than any fetcher sub-issue.

• combined_status_row — multiple badges in one row
• daily_digest — weather + calendar + notifications
• random_widget_of_day — N configured, pick 1
• rotating_message — different text per cd

---

Prioritization rubric

For when/if we pick something up — not a commitment.

1. Renderer primitives first — a single renderer (heatmap, badge, timeline) unlocks many dependent fetchers. Highest leverage per unit effort. Lives in meta: renderer roadmap #61.
2. Daily-driver coverage — does a fresh install fill a satisfying home splash? Gaps: weather_*, calendar_*, full system_* rollout.
3. Per-dir killer feature — does cd into project show meaningfully more than without splashboard? Gaps: forge parity across gitlab_* / bitbucket_* / etc., ci_*.
4. Dedicated built-ins over ReadStore — ReadStore is the escape hatch (feat: $HOME/.splashboard/ home-dir layout + ReadStore fetcher #43), not the preferred path for anything that warrants curated UX. Build a dedicated fetcher when the widget earns it.

---

Rejected designs (don't reintroduce)

• Plugin protocol (core: Plugin protocol (subprocess-based I/O spec) #5) — closed. Splashboard is a curated splash renderer, not a dashboard framework. Subprocess plugins can't be bounded in blast radius. Custom widgets land as built-in PRs or use ReadStore.
• Command widget (widget: command / shell oneliner (generic subprocess fetcher) #20) — closed. No command = "..." fetcher, in any config scope. Local-vs-global rule carve-outs are not worth the footgun potential.
• Screensaver sub-mode — out of scope. Animation lives within the existing 2-second ANIMATION_WINDOW.
• Exec safety class — reserved enum arm, currently unreachable since core: Plugin protocol (subprocess-based I/O spec) #5 / widget: command / shell oneliner (generic subprocess fetcher) #20 are closed.