SideButton Knowledge Module

SideButton Portal — Jobs — Knowledge Module

SideButton knowledge module — UI selectors, data model, and page states documenting Portal — Jobs.

Available free v1.1.0 Browser

$ sidebutton install sidebutton

PRIVACY CONVENTION: This skill file must describe the PAGE — elements, selectors, layouts, states, and public UI copy (placeholder text, button labels, default headings). It must NOT record tenant-specific data visible to a logged-in user: real ticket keys (use PROJ-123), real agent hostnames (use agent-host or {hostname}), real IP addresses (use 203.0.113.10), real workflow/automation IDs (use {id}), real user names or emails, real Jira host (use {tenant}.atlassian.net), real registry names (use {registry-name}), or real automation display names. When re-verifying, generalize observed values before writing them into this file.

VERIFIED 2026-04-26: H1 "Jobs". Filter tabs are 6 link elements: All, Running, Waiting, Completed, Failed, Cancelled. Active filter pill is dark grey solid (NOT green/blue as earlier notes claimed). Controls: textbox "Search tickets…", combobox with label "Workflow" (id workflow-filter-select, 29 options including legacy agent_* and dash IDs), combobox with label "Agent". NEW (Apr 2026): job rows show cost + model inline in the format {Workflow} #{id} {effort} 1 step {duration} ${cost} {model_short} — visible on jobs created after cost-tracking landed; older jobs omit ${cost} {model}. Token-usage span on each row reads "N in · N,NNN out · N turns · claude-opus-4-7" (model with claude- prefix in span; without prefix in row text). NEW (Apr 2026): LLM USAGE card on detail page — large dark card showing model name (top-right), Cost / Turns / Input Tokens / Output Tokens stat columns, Cache read / Cache write / Prompt sha row, and PER STEP breakdown table. NEW: yellow "Status mismatch" inline alert above Temporal timeline when portal status diverges from orchestrator status. NEW: floating "Report a Bug" button in viewport bottom-right (global to portal, anchored to viewport). Effort badges use distinct background colors per level (NOT all green): MAX = purple, HIGH = orange/amber, MEDIUM = dark grey. Auth0 login required.

Portal — Jobs

What This Is

Job execution history page -- lists all workflow jobs that have been dispatched to agents. Shows a filterable, searchable list of jobs with status badges, workflow names, ticket references, agent assignments, step counts, durations, and relative timestamps. Each job links to a detail page showing input parameters, step execution timeline with task progress, expandable step logs, and output/error cards. Jobs are created when workflows are dispatched to agents (via Run Workflow, Run Job, Queue dispatch, or Automation trigger). Supports cancel (running/waiting) and resend (failed/cancelled) actions.

URL Patterns

Pattern	Description
`/portal/jobs`	Job list (all)
`/portal/jobs?status=running`	Filtered to running jobs
`/portal/jobs?status=waiting`	Filtered to waiting jobs
`/portal/jobs?status=completed`	Filtered to completed jobs
`/portal/jobs?status=failed`	Filtered to failed jobs
`/portal/jobs?status=cancelled`	Filtered to cancelled jobs
`/portal/jobs/{id}`	Job detail view
`/portal/jobs/{id}?from_status={status}`	Job detail with back-link context to filtered list

Page Structure

Job List (`/portal/jobs`)

+--[Sidebar]--+--[Main Content]------------------------------------------+
|  Chat       |  H1: "Jobs"                                               |
|  Agents     |  "{N} jobs"  (e.g. "2551 jobs")                          |
|  Jobs *     |                                                           |
|  Queue      |  [Filter Tabs: All | Running N | Waiting | Completed | Failed | Cancelled] |
|  Automations|                                                           |
|  Workflows  |  [Search: "Search tickets..."]  Workflow [▼]  Agent [▼]   |
|  Settings   |                                                           |
|             |  ITEM                              AGENT         STATUS   |
|             |  [dot] Workflow Name  TICKET  #ID                         |
|             |        [EFFORT]  N step  Nm Ns  $X.XX  model-short        |
|             |                          [avatar] agent-name              |
|             |                                   [Status Badge]  time ago |
|             |                                   [Cancel/Resend]          |
|             |  ...                                                      |
|             |  Loading more...                                          |
|             |                                       [Report a Bug ⌶]    |
+-------------+-----------------------------------------------------------+

New in Apr 2026: each row's secondary line includes inline $X.XX cost and a model-short tag (e.g. opus-4-7, sonnet-4-6). These render only on jobs created after cost-tracking shipped — older jobs render the secondary line without $cost model. Each row also exposes a hidden <span> with the verbose "N in · N,NNN out · N turns · claude-{model}" summary (visible in the accessibility tree, used as a tooltip).

Job Detail (`/portal/jobs/{id}`)

+--[Sidebar]--+--[Main Content]------------------------------------------+
|             |  < Back to jobs                                           |
|  Jobs *     |                                                           |
|             |  H1: workflow-slug — ticket/source  [Status Badge]       |
|             |                                          [Cancel/Resend]  |
|             |  Job #N  [avatar] agent  [EFFORT]  Nm Ns  Created {date} |
|             |  pipeline-id (green monospace)                            |
|             |                                                           |
|             |  [INPUT card]                                             |
|             |    hint: focus: /customers/ID/tasks                       |
|             |    (or trigger_* fields for automation jobs)              |
|             |                                                           |
|             |  [LLM USAGE card]  ← NEW Apr 2026                         |
|             |    LLM USAGE                          claude-opus-4-7     |
|             |                                                           |
|             |    COST    TURNS   INPUT TOKENS   OUTPUT TOKENS           |
|             |    $9.79   128     208            62,084                  |
|             |                                                           |
|             |    Cache read 11,565,708  Cache write 392,651             |
|             |    Prompt sha 8065ecd866c2                                |
|             |                                                           |
|             |    PER STEP                                               |
|             |    #1  agent_sd_coverage      62,292 tok  128t  $9.79     |
|             |                                                           |
|             |  [ERROR card] (failed jobs only)                          |
|             |    Step N (step-name) failed: error message               |
|             |                                                           |
|             |  [Status mismatch alert] (yellow inline, when divergent)  |
|             |    ⚠ Status mismatch: portal says "failed" but            |
|             |       orchestrator says "COMPLETED"                       |
|             |                                                           |
|             |  [Temporal Timeline] OR [Temporal unreachable]            |
|             |    Job: STATUS | N events | size KB | short_id           |
|             |    Init          Nms    N activities                      |
|             |    Agent Setup   Nms    N activities                      |
|             |    Execution     Nm Ns                                    |
|             |      runAndMonitor  Nm Ns                                |
|             |        workflow: agent_workflow  agent: agent-name        |
|             |    Teardown      Nms    N activities                      |
|             |                                                           |
|             |  [OUTPUT card] (completed jobs only)                      |
|             |    "Workflow X completed on agent-name"                   |
|             |                                                           |
|             |                                       [Report a Bug ⌶]    |
+-------------+-----------------------------------------------------------+

Key Elements

Job List

Element	Selector / Ref	Notes
Page heading	`heading "Jobs"`	H1
Job count	Below heading	"N jobs" summary
Filter: All	`link "All"`	Green pill (`bg-emerald-500/20 text-emerald-400`) when active
Filter: Running	`link "Running"`	Shows count badge (e.g. "Running 3") when jobs are running
Filter: Waiting	`link "Waiting"`
Filter: Completed	`link "Completed"`
Filter: Failed	`link "Failed"`	Blue pill when active as selected filter
Filter: Cancelled	`link "Cancelled"`	6th status filter tab
Search textbox	`textbox "Search tickets…"`	Searches by ticket reference (e.g. "PROJ-123")
Workflow filter	`combobox` / `label "Workflow"` / `#workflow-filter-select`	Native `<select>`, default "All workflows". Options populate dynamically from job data — currently 29 options including legacy `agent_*` IDs (e.g. `agent_se_work`) and dash IDs (e.g. `se-work`); display labels are Title Case
Agent filter	`combobox` / `label "Agent"`	Native `<select>`, default "All agents"
Column header: ITEM	Above job rows	Left column
Column header: AGENT	Above job rows	Middle column
Column header: STATUS	Above job rows	Right column
Job row	`link` (each job)	Clickable, navigates to `/portal/jobs/{id}`
Status dot (green)	Left of workflow name	Completed
Status dot (red)	Left of workflow name	Failed
Status dot (blue, pulsing)	Left of workflow name	Running
Status dot (grey)	Left of workflow name	Cancelled / Waiting
Workflow name	Bold text in row	Title case display name (e.g. "SE Rca Fix", "QA Validate", "Pull Repos")
Ticket reference	Orange/cyan text after name	e.g. "PROJ-123" — only present when job has a linked ticket
Job ID	Grey text	e.g. "#3121"
Agent avatar	`img` inside agent link	Small green circle with agent avatar image
Agent name link	`link "agent-name"`	Links to `/portal/agents`, shows agent hostname
Effort badge	`[MAX]` / `[HIGH]` / `[MEDIUM]`	Pill badge — color varies per level: MAX = purple, HIGH = orange/amber, MEDIUM = dark grey (NOT all green; earlier docs were wrong)
Step count	Grey text	"N step" (singular)
Duration	Grey text	"Nm Ns" format (e.g. "6m 31s", "5m 46s") or "Ns" for short durations
Inline cost	Green text after duration	`$X.XX` (e.g. `$9.79`, `$0.127`) — only present on jobs with cost-tracking data; older jobs omit
Inline model tag	Grey text after cost	Short model name (e.g. `opus-4-7`, `sonnet-4-6`, `opus-4-6`) — paired with inline cost
Token usage span	`<span>` inside row link, e.g. `span "208 in · 62,084 out · 128 turns · claude-opus-4-7"`	Full LLM usage summary (with `claude-` prefix). Visible in accessibility snapshot — likely a tooltip on hover
Completed badge	Status area	Green dot + "Completed"
Failed badge	Status area	Red dot + "Failed"
Running badge	Status area	Blue dot + "Running"
Cancelled badge	Status area	Grey dot + "Cancelled"
Waiting badge	Status area	Amber dot + "Waiting"
Relative time	Right side	"Nm ago", "Nh ago", "Nd ago"
Cancel button	`button "Cancel"`	On running/waiting jobs — cancels the job
Resend button	`button "Resend"`	On failed and cancelled jobs — re-dispatches the job
Resended button	`button "Resended"`	Shown after a job has already been resent — disabled/grey state
Loading more	Bottom of list	"Loading more..." text, infinite scroll trigger (50 items/page)

Job Detail

Element	Selector / Ref	Notes
Back link	`link "Back to jobs"`	Chevron left + text, returns to job list
Workflow heading	H1 text	Format: "workflow-slug — ticket-key" or "workflow-slug — queued"
Status badge	Next to heading	"Completed" / "Failed" / "Running" / "Cancelled" with colored dot
Cancel button	`button "Cancel"`	On running/waiting jobs only, top-right, blue outline
Resend button	`button "Resend"`	On failed/cancelled jobs, top-right, blue outline. Changes to "Resended" after use
Job number	Below heading	"Job #N"
Agent avatar link	`link > img "agent-name"`	Green circle avatar, links to `/portal/agents`
Agent name	Below heading	Agent hostname next to avatar
Effort badge	Below heading	"MAX" / "HIGH" / "MEDIUM" green pill
Duration	Below heading	"Nm Ns" format (ticks live on running jobs)
Created date	Below heading	Full date string, e.g. "Created 4/2/2026, 9:33:59 PM"
Pipeline ID	Below metadata	Green/teal monospace text, e.g. "pipeline-job-4227"
Temporal Workflow ID	`span "Temporal Workflow ID"`	Visible in accessibility tree, links to Temporal tracking
INPUT card	Dark card below metadata	"Input" header, key-value pairs of input parameters
INPUT trigger variables	Inside INPUT card	Automation-triggered jobs show: trigger_ticket_url, trigger_project, trigger_assignee, trigger_status, trigger_issue_key, trigger_issue_type
INPUT ticket URL	Inside INPUT card	Clickable cyan link when input contains a URL (e.g. Jira ticket)
INPUT hint	Inside INPUT card	"hint:" field showing dispatch instructions (e.g. "Jira webhook: jira:issue_updated", "focus: /customers/ID/tasks")
LLM USAGE card (NEW Apr 2026)	Dark card below INPUT	"LLM USAGE" header (left) + full model name `claude-opus-4-7` / `claude-sonnet-4-6` / `claude-opus-4-6` (top-right). Stat columns + cache + per-step block. Hidden when no LLM telemetry exists for the job
LLM Cost stat	First column under "COST" label	Large green dollar value (e.g. `$9.79`, `$0.127`)
LLM Turns stat	Second column under "TURNS"	Plain number (e.g. `128`, `8`)
LLM Input Tokens stat	Third column under "INPUT TOKENS"	Plain number — non-cached prompt input (e.g. `208`, `14`)
LLM Output Tokens stat	Fourth column under "OUTPUT TOKENS"	Plain number with thousands comma (e.g. `62,084`, `2,264`)
Cache read	Below stat row	Label "Cache read" + monospace number (e.g. `11,565,708`)
Cache write	Below stat row	Label "Cache write" + monospace number (e.g. `392,651`, `11,008`)
Prompt sha	Below stat row	Label "Prompt sha" + 12-char hex prefix (e.g. `8065ecd866c2`)
PER STEP table	Bottom of LLM USAGE card	Heading "PER STEP" + one row per executed step
Per-step row	Inside PER STEP	Format: `#{idx}` + `step_name` (monospace) + `{N,NNN} tok` + `{N}t` (turns) + `${cost}` (green)
ERROR card	Red background card	Shown on failed jobs only, "Error" header + "Step N (step-name) failed" + error message
Status mismatch alert (NEW Apr 2026)	Yellow `⚠` inline alert above timeline	Reads "Status mismatch: portal says 'X' but orchestrator says 'Y'" — appears when portal/Temporal status differ
Temporal Timeline	Below ERROR/alert	Full job execution timeline from Temporal orchestrator
Timeline summary	First line of timeline	"Job: STATUS \| N events \| size KB \| short_id"
Timeline: Init phase	Timeline row	Duration (ms) + activity count (e.g. "3 activities")
Timeline: Agent Setup phase	Timeline row	Duration (ms) + activity count (e.g. "4 activities")
Timeline: Execution phase	Timeline row	Duration (Nm Ns) — main execution time
Timeline: runAndMonitor	Sub-row under Execution	Duration + "workflow: {workflow_id} agent: {agent-name}"
Timeline: Teardown phase	Timeline row	Duration (ms) + activity count (e.g. "5 activities")
Temporal unreachable	Replaces timeline	Shown when Temporal workflow history is unavailable (old jobs, cancelled jobs, service down)
OUTPUT card	Dark card at bottom	Shown on completed jobs. "Output" header + monospace green completion message of shape `Workflow {workflow_slug} completed on {agent_hostname}`

Data Model

Job

Field	Type	Description
id	number	Job ID (auto-increment)
pipeline_id	string	Pipeline tracking ID (e.g. "pipeline-job-3121")
name	string	Display name format: "{workflow_id} — {ticket_key\|queued}"
status	string	"pending" / "running" / "waiting" / "completed" / "failed" / "cancelled"
effort_level	string	"max" / "high" / "medium" (optional — may be absent)
input	string (JSON)	Key-value map of input parameters (ticket_url, hint, etc.)
output	string	Completion message (on completed jobs)
error	string	Error message (on failed jobs)
ticket_url	string	Full Jira ticket URL (optional)
ticket_key	string	Jira ticket key e.g. "PROJ-123" (optional)
current_step	number	Index of currently executing step
steps	string (JSON)	Array of pipeline step definitions
started_at	string (datetime)	When job started executing
completed_at	string (datetime)	When job finished
created_at	string (datetime)	Job creation timestamp
slack_channel_id	string	Slack channel for notifications (optional)
slack_thread_ts	string	Slack thread timestamp (optional)

Temporal Timeline (Job Detail)

The job detail page fetches execution history from the Temporal orchestrator. When available, it renders a phase-based timeline. When unavailable, it shows "Temporal unreachable".

Field	Type	Description
status	string	Temporal workflow status: "COMPLETED" / "FAILED" / "RUNNING" / "CANCELLED"
event_count	number	Total events in workflow history (e.g. "83 events")
data_size	string	History data size (e.g. "14.9 KB")
short_id	string	Short hex workflow run ID prefix (e.g. "019d500e")

Timeline Phases

Phase	Description
Init	Workflow initialization (~15-20ms, 3 activities)
Agent Setup	Agent environment preparation (~24-25ms, 4 activities)
Execution	Main workflow execution — contains runAndMonitor sub-entry showing workflow ID + agent name
Teardown	Cleanup and reporting (~38-48ms, 4-5 activities)

LLM Usage (Job Detail card — NEW Apr 2026)

The detail page renders an LLM USAGE card aggregating token, cost, and cache telemetry from the agent's LLM calls. The card is omitted when telemetry is unavailable.

Field	Type	Description
model	string	Full Claude model id, e.g. `claude-opus-4-7`, `claude-sonnet-4-6`, `claude-opus-4-6` (rendered top-right of card)
cost	string	Total cost in USD, formatted with `$` prefix and 2-3 decimals (e.g. `$9.79`, `$0.127`)
turns	number	Total LLM round-trips (e.g. `128`)
input_tokens	number	Non-cached input tokens (e.g. `208`)
output_tokens	number	Output tokens, comma-separated (e.g. `62,084`)
cache_read	number	Cache hit token count, comma-separated (e.g. `11,565,708`)
cache_write	number	Cache write token count, comma-separated (e.g. `392,651`)
prompt_sha	string	12-char hex prefix of prompt SHA (e.g. `8065ecd866c2`)
per_step	array	Per-step usage breakdown

Per-Step Usage Row

Field	Type	Description
index	number	1-based step index, displayed as `#1`, `#2`, ...
step_name	string	Step identifier (e.g. `agent_sd_coverage`, `agent_pull_repos`) — monospace
tokens	number	Total tokens consumed in this step (e.g. `62,292 tok`)
turns	number	Turns in this step, suffixed `t` (e.g. `128t`)
cost	string	Step cost (e.g. `$9.79`) — green text

Input Trigger Variables (Automation-triggered jobs)

Jobs triggered by Jira automations include trigger context in the INPUT card:

Field	Type	Description
trigger_ticket_url	string	Full Jira ticket URL from trigger event
trigger_project	string	Jira project key (e.g. "MVP")
trigger_assignee	string	Jira account ID of assignee
trigger_status	string	Jira issue status at trigger time (e.g. "To Do", "In Review")
trigger_issue_key	string	Jira issue key (e.g. "MVP-1601")
trigger_issue_type	string	Jira issue type (e.g. "Bug", "Story")

States & Variations

State	Trigger	Visual Indicator
All filter active	Default / click "All"	Active filter pill is dark grey solid background (NOT green/blue — earlier docs were wrong)
Running filter	`?status=running`	Only running jobs shown, Running tab active. Count badge ("Running 3") only when jobs are actually running
Completed filter	`?status=completed`	Only completed jobs shown
Failed filter	`?status=failed`	Only failed jobs shown
Waiting filter	`?status=waiting`	Only waiting jobs shown
Cancelled filter	`?status=cancelled`	Only cancelled jobs shown
Running count badge	Running jobs exist	Running tab shows count badge (e.g. "Running 3")
Job completed	Finished successfully	Green dot + "Completed" badge
Job failed	Error occurred	Red dot + "Failed" badge
Job running	In progress	Blue pulsing dot + "Running" badge, live-ticking duration
Job waiting	Blocked at gate	Amber dot + "Waiting" badge
Job pending	Not yet started	"Pending" badge
Job cancelled	User cancelled	Grey dot + "Cancelled" badge
Effort: Max	`effort_level = "max"`	Purple "MAX" pill badge
Effort: High	`effort_level = "high"`	Orange/amber "HIGH" pill badge
Effort: Medium	`effort_level = "medium"`	Dark grey "MEDIUM" pill badge
Inline cost+model present	Job has cost-tracking telemetry	Row shows `${cost} {model_short}` after duration (e.g. `$9.79 opus-4-7`)
Inline cost+model absent	Older job (pre-cost-tracking)	Row stops at duration; no `$` or model in row text
LLM USAGE card present	Detail page, telemetry available	Card with cost/turns/tokens/cache/per-step breakdown renders below INPUT
LLM USAGE card absent	Detail page, no LLM telemetry	LLM USAGE card omitted entirely (e.g. cancelled before any LLM call)
Status mismatch alert	Portal status ≠ Temporal status	Yellow `⚠` inline alert above timeline reads `Status mismatch: portal says "X" but orchestrator says "Y"`
Ticket reference present	Job linked to ticket	Orange/cyan ticket key text next to workflow name
Cancel available	Job is running or waiting	"Cancel" button shown in list row and detail top-right
Resend available	Job is failed or cancelled	"Resend" button shown in list row and detail top-right
Already resent	Resend was clicked	"Resended" button shown (disabled/past tense)
Detail: Temporal timeline	Temporal history available	Phase-based timeline: Init → Agent Setup → Execution → Teardown with durations and activity counts
Detail: Temporal unreachable	Temporal history unavailable	"Temporal unreachable" text replaces timeline (old jobs, cancelled jobs, service down)
Detail: Temporal status mismatch	Portal vs orchestrator differ	ERROR card may show "Status mismatch: portal says 'failed' but orchestrator says 'COMPLETED'"
Detail: ERROR card	Job has failure	Red "Error" card above timeline with "Step N (step-name) failed" + error detail
Detail: OUTPUT card	Job completed	Dark "Output" card below timeline with green completion message
Detail: trigger variables	Automation-triggered job	INPUT card shows trigger_ticket_url, trigger_project, trigger_assignee, trigger_status, trigger_issue_key, trigger_issue_type
Empty: All tab	No jobs dispatched, "All" active	Lightning bolt icon + "No jobs yet" heading + "Run a workflow from the agents page to create a job" hint
Empty: Running tab	No running jobs, "Running" active	Lightning bolt icon + "No running jobs" heading + same hint
Empty: Waiting tab	No waiting jobs, "Waiting" active	Lightning bolt icon + "No waiting jobs" heading + same hint
Empty: Completed tab	No completed jobs, "Completed" active	Lightning bolt icon + "No completed jobs" heading + same hint
Empty: Failed tab	No failed jobs, "Failed" active	Lightning bolt icon + "No failed jobs" heading + same hint
Empty: Cancelled tab	No cancelled jobs, "Cancelled" active	Lightning bolt icon + "No cancelled jobs" heading + same hint
Agent filter hidden	No agents registered or no jobs	Agent dropdown not rendered in filter bar; only Workflow filter shown
Workflow filter empty	No jobs exist	Workflow dropdown only shows "All workflows" — options populate dynamically from job data
Infinite scroll	Scroll to bottom	"Loading more..." appears, loads next 50 jobs

Common Tasks

1. View job list

Navigate to /portal/jobs
See all jobs with status, workflow name, ticket ref, agent, step count, duration, time

2. Filter jobs by status

Click filter tab (Running / Waiting / Completed / Failed / Cancelled)
List filters to matching jobs, URL updates to ?status={status}

3. Search by ticket reference

Click the "Search tickets..." textbox
Type ticket reference (e.g. "PROJ-123")
List filters to matching jobs

4. Filter by workflow

Click the "Workflow" dropdown (native <select>)
Select a workflow name (e.g. "SE Rca Fix", "QA Validate")
List filters to jobs for that workflow

5. Filter by agent

Click the "Agent" dropdown (native <select>)
Select an agent name
List filters to jobs for that agent

6. View job detail

Click any job row in the list
See workflow name, ticket ref, status, input parameters, step execution timeline

7. View Temporal timeline

On job detail page, see the timeline section below INPUT/ERROR
Timeline shows 4 phases: Init, Agent Setup, Execution, Teardown with durations
Execution phase shows runAndMonitor sub-entry with workflow ID and agent name
If Temporal is unreachable, shows "Temporal unreachable" instead

8. Cancel a running or waiting job

On a running/waiting job (list or detail), click the "Cancel" button
Job transitions to "Cancelled" status

9. Resend a failed or cancelled job

On a failed/cancelled job (list or detail), click the "Resend" button
Job is re-dispatched — button changes to "Resended" (disabled)

Tips

Jobs are created when workflows are dispatched via Run Workflow button, Run Job on agent, Queue dispatch, or Automation trigger
The effort level maps to LLM model: Max/High = Opus, Medium = Sonnet — confirmed by inline model_short text on rows (opus-4-7 for max/high, sonnet-4-6 for medium)
Inline $cost model_short was added in Apr 2026 cost-tracking. Use it as a quick filter heuristic when scanning lists for expensive runs
The LLM USAGE card is the canonical source for total spend per job — sum the Cost field across rows for fleet-level totals. The PER STEP table itemizes per pipeline step
Cache read typically dwarfs input_tokens because of system-prompt + skill-pack caching; large cache_read with low input_tokens is the expected, healthy pattern
Prompt sha identifies the canonical prompt build — same SHA across jobs means identical static prompt content, useful for reproducibility checks
The token-usage span on each row ("N in · N,NNN out · N turns · claude-{model}") is the same data the LLM USAGE card aggregates; both are absent on jobs without telemetry
Pipeline ID format is pipeline-job-{id}
Job list shows relative timestamps ("Nm ago", "Nh ago"); detail page shows absolute timestamps
The Temporal timeline shows 4 phases (Init, Agent Setup, Execution, Teardown) — Execution contains the actual workflow run with agent name
Workflow names appear as Title Case in the list (e.g. "SE Rca Fix") but as slugs in detail headings (e.g. "se-rca-fix")
Heading format on detail page: "workflow-slug — ticket-key" (when ticket linked) or "workflow-slug — queued" (from queue)
Agent names link to /portal/agents page
Automation-triggered jobs show trigger context variables in INPUT (trigger_ticket_url, trigger_project, trigger_assignee, trigger_status, trigger_issue_key, trigger_issue_type)
INPUT card may contain hint: field showing extra dispatch instructions or automation trigger context (e.g. "Jira webhook: jira:issue_updated")
Queued jobs (non-automation) show only standard params (ticket_url, hint) without trigger_* variables
Infinite scroll loads 50 items per page
Job links from filtered views include ?from_status={status} for back-navigation context

Gotchas

Cancel works on running AND waiting: Both running and waiting jobs can be cancelled
Resend button: Apr 2026 — current accessibility tree exposes only button "Resend" on failed/cancelled rows. Earlier docs claimed a "Resended" disabled state appears post-click; this could not be reverified non-destructively in the Apr 2026 pass. Treat the post-click label as needs verification and avoid double-clicking Resend in automation
No delete: There is no UI to delete jobs from history
Effort badge colors are NOT all green: Earlier docs were wrong. As of Apr 2026: MAX = purple, HIGH = orange/amber, MEDIUM = dark grey. Color-based selectors must use per-level CSS classes, not a single shared green class
Inline cost/model only on telemetry-present jobs: Older jobs and any job created before cost-tracking shipped omit the $cost model_short segment. Don't rely on its presence to detect "completed"; rely on the explicit Status badge instead
LLM USAGE card may be absent: Cancelled-before-execution jobs and very old jobs render the detail page without the LLM USAGE card. Tests must guard with if card present rather than assuming presence
Status mismatch alert is a real failure signal: When the yellow ⚠ Status mismatch alert renders, the portal DB and Temporal disagree about the job outcome. The portal-side status drives the badge and Resend availability; the Temporal-side status drives the timeline. Investigators should check both to find the truth
"Report a Bug" floating button: Anchored to viewport bottom-right on every portal page (not just Jobs). Don't snapshot it as Jobs-specific; it's global to /portal/*
Status dots: Colored dots match status: green=completed, red=failed, blue=running (pulsing), amber=waiting, grey=cancelled
Workflow/Agent dropdowns: Native <select> elements — use select_option() to interact
Search scope: "Search tickets..." searches by ticket reference only, not by workflow name or agent
Temporal unreachable: Older jobs, cancelled jobs, and jobs during Temporal outages show "Temporal unreachable" instead of the phase timeline — OUTPUT/ERROR cards still display from portal DB
Temporal status mismatch: A job's portal status and Temporal status can differ — e.g., portal shows "Failed" but Temporal shows "COMPLETED". The ERROR card shows the mismatch detail
Cancelled jobs: Often show "Temporal unreachable" rather than a timeline — no step-level detail available
Live duration: Running jobs show a ticking duration counter
hint field: The INPUT card may show a hint: field with automation context (e.g. "Jira webhook: jira:issue_updated") — this helps trace how the job was triggered
Empty state per filter: Each filter tab shows its own empty state heading — "No running jobs", "No completed jobs", etc. Only the All tab says "No jobs yet". The hint text is always the same
Dynamic dropdown options: Workflow and Agent filter dropdowns populate dynamically from existing job data. An account with 0 jobs shows only "All workflows" and no Agent dropdown at all
Workflow filter dual IDs: Historical jobs use legacy agent_* prefixed IDs (e.g. agent_se_work); newer jobs use dash-separated IDs (e.g. se-work). Both appear in the Workflow filter dropdown since it populates from job data
Workflow filter select ID: The select element has id="workflow-filter-select" — useful for targeted automation
Page title format: List page: "Jobs - SideButton Fleet". Detail page: "Job #{id} - SideButton Fleet"
No Show details expandable: The old <details> expandable with RESOLVED INPUT/OUTPUT is gone — replaced by Temporal timeline phases

← All SideButton modules SideButton overview