Add flag for keeping the tui open persistently

[Eric162]

Feature Request

Describe the problem that led to you needing this feature

I want to use this in a multiplexer to just persistently watch the pr state as it gets updated

Describe the solution you'd like

• a flag like --repo that makes it so the plugin doesn't quit
• takes arguments of either $owner/$repo or https://github.com/$owner/$repo, but defaults to the current repo
• also show actions for PRs. Add a column with the PR# that is related
• add options (configurable in the config file) to define a time after which we stop displaying successful and unsuccessful checks. Default to 15 minutes for successful checks and 30 minutes for failed actions.
• ?? a lower refresh rate for finished checks

Describe alternatives you've considered

none - there's no flag for this

Additional context

[chicks-net]

I like this idea. I'm pondering how I'd like to implement it. I'll let you know when it is in a release.

[chicks-net]

Implementation Plan

Based on discussion, here are the design decisions and implementation plan.

Decisions

Topic	Decision
Scope	All open PRs with active (running/queued) checks on a repo
Flag	--repo taking owner/repo or URL, defaulting to current repo; implies persistent mode
Active PR filter	Only PRs with at least one running/queued check right now
Completed check fade-out	Disappear from display after configurable timeout (15m success / 30m failure default)
PR disappears	When all checks complete and fade out, PR vanishes entirely
Layout	Grouped by PR (header row per PR, checks indented below)
Fetch strategy	Single batched GraphQL query for all open PRs + check rollup
Polling interval	15s default in repo mode, new config repo_refresh_interval
PR list cap	50 most recently updated PRs
Config for fade-out	Top-level fade_success: 15m and fade_failure: 30m
--quick + repo mode	Incompatible (error if both used)
--debug + repo mode	Works
Snapshot mode + repo mode	Incompatible (error if used in non-terminal)
Positional arg + --repo	Incompatible (error if both provided)
PR header	Standard: PR #123: Title

Flag & CLI Changes (main.go)

• Add --repo flag accepting owner/repo or https://github.com/$owner/$repo
• Validation rules:
  • --repo + positional arg = error
  • --repo + --quick = error
  • --repo in non-terminal (snapshot mode) = error
  • --repo implies persistent mode (TUI never auto-quit)
• New runArgs mode: modeRepo
• New parseRepoArg() to extract owner/repo from owner/repo or URL format
• New GetCurrentRepo() in internal/github/detect.go — parse git remote get-url origin for SSH/HTTPS owner+repo extraction
• New runRepoMode() entry point

Config Changes (internal/config/config.go)

Add three top-level config values:

repo_refresh_interval: 15s
fade_success: 15m
fade_failure: 30m

Update Config struct, viper defaults, and .config.example.yaml.

GraphQL Batched Query (internal/github/)

New query struct for fetching multiple PRs with check rollup in a single call:

repository(owner: $owner, name: $repo) {
  pullRequests(first: 50, states: OPEN, orderBy: {field: UPDATED_AT, direction: DESC}) {
    nodes {
      number
      title
      commits(last: 1) {
        nodes {
          commit {
            statusCheckRollup {
              contexts(first: 100, after: $contextsCursor) { ... }
            }
          }
        }
      }
    }
  }
  rateLimit { remaining }
}

New function: FetchRepoCheckRunsGraphQL(ctx, token, owner, repo) -> map[int]PRCheckData where PRCheckData holds PR title, check runs, and head commit time.

This avoids N separate API calls per poll cycle.

New TUI Model (internal/tui/repomodel.go)

New RepoModel struct — separate from the single-PR Model:

• Fields: owner, repo, map of PR# → PR data (title, check runs, head commit time), rate limit, spinner, refresh interval (15s default from config), fade timeouts, exit code, quitting flag
• Init(): kick off initial repo fetch + spinner
• Update(): handle TickMsg, RepoChecksUpdateMsg, tea.KeyMsg (q to quit)
• Never auto-quit — only exits on q/ctrl+c

New TUI View (internal/tui/repoview.go)

Grouped layout per PR:

owner/repo — 3 active PRs — Updated 8s ago

PR #123: Fix authentication bug
  ✅ CI / build     0s    2m avg
  🔄 CI / test      1m    5m avg
  ⏳ CI / deploy    —     3m avg

PR #456: Update README
  ❌ Lint / check   4s    1m avg
  🔄 CI / build     3m    2m avg

Press q to quit

Fade-Out Logic (internal/tui/repoupdate.go)

On each RepoChecksUpdateMsg, for each PR's completed checks:

• If conclusion == "success" and completed > fade_success ago → remove from display
• If conclusion is a failure and completed > fade_failure ago → remove from display
• If a PR has zero visible checks remaining (all faded or filtered) → remove the PR group entirely
• "Active" filter: only include PRs that have at least one in_progress or queued check OR have checks within the fade-out window

Message Types (internal/tui/messages.go)

New messages:

• RepoChecksUpdateMsg — carries map of PR# → check data + rate limit
• RepoPRInfoMsg — PR list with metadata

Files to Create

File	Purpose
internal/github/repo.go	GetCurrentRepo(), ParseRepoArg(), FetchRepoCheckRunsGraphQL()
internal/tui/repomodel.go	RepoModel struct, NewRepoModel()
internal/tui/repoupdate.go	RepoModel.Update(), fade-out filter logic
internal/tui/repoview.go	RepoModel.View(), grouped PR rendering

Files to Modify

File	Change
main.go	Add --repo flag, validation, modeRepo, runRepoMode()
internal/config/config.go	Add RepoRefreshInterval, FadeSuccess, FadeFailure fields + defaults
internal/tui/messages.go	Add RepoChecksUpdateMsg, RepoPRInfoMsg
internal/tui/constants.go	Add repo-mode constants (50 PR cap, etc.)
.config.example.yaml	Document new config values

Execution Order

1. Config + CLI flag changes (foundation)
2. GetCurrentRepo() + ParseRepoArg() (repo detection)
3. Batched GraphQL query (data layer)
4. RepoModel + update logic (state management)
5. RepoModel.View() (rendering)
6. Integration in main.go (wire it all up)
7. Tests