Search, Filters & Activity

Pensum ships a handful of views that don’t get their own pages individually because they’re smaller and self-explanatory. They share a doc here for discoverability.

Search

Open via the command palette: Pensum: Search tasks.

Live full-text search across every task in your vault. Type into the search box and results update as you type — no Enter required. Hits are highlighted inline.

What it searches

• Task description text
• Task tags
• The project the task belongs to (matched against the inline #project/name tag or containment)
• The file path of the source file

What it returns

A flat list of matching tasks, sorted by priority then due date. Click any task to open the edit modal. Right-click for quick actions. Same interaction model as every other view.

When to use Search

• You remember roughly what a task said but not where it lives
• You’re trying to find every mention of a specific tag, person, or project
• You want a fast “is there a task about X anywhere?” check

For more structured queries (status, date range, multiple filters combined), use the Filters view.

Filters

Open via the command palette: Pensum: Open filters view.

Filters is a builder for custom task queries. You compose a query out of conditions (status, priority, project, tags, date ranges, description text), name it, save it, and it persists across sessions. The view shows results for the active filter and updates live as the underlying tasks change.

Available conditions

Condition	Operators	Example
Status	is / is not	status is open
Priority	is / is not / >= / <=	priority >= high
Project	is / is not / is any of / is none of	project is acme
Tags	has / has any of / has none of	tags has work
Due date	before / after / on / between / is empty / is not empty	due before next week
Scheduled date	(same operators as due)	scheduled is empty
Description	contains / does not contain / matches regex	description contains "review"

Multiple conditions combine with AND by default. The builder UI lets you toggle to OR or add nested groups for more complex logic.

Saving and managing filters

• Save the current builder state with a name — appears in the filter dropdown for later use
• Edit a saved filter inline; changes persist
• Delete filters you don’t need
• Set as default so it’s the filter that opens when you reopen the view

Saved filters are stored in plugin settings (install-scoped). They don’t sync across devices via Obsidian Sync today — copy/paste between devices if you need that.

Useful filter examples

What you want	How to build it
Everything from Acme this week	project is acme AND due between today and next week
Untriaged tasks older than 3 days	project is empty AND created before 3 days ago
Anything I tagged urgent that’s not done	tags has urgent AND status is open
High-priority work I’ve actively started	priority >= high AND status is in-progress
Stale stubs to review	(use the Curate view instead — it’s purpose-built for this)

When to use Filters

• You have a query you run regularly and want it one click away
• You need conditions Search can’t express (multiple AND / OR clauses, date ranges, negation)
• You’re building a dashboard of saved queries for your weekly review

Activity

Open via the command palette: Pensum: Open activity view.

Activity is split into two tabs:

Tasks tab (default)

Your completion history grouped by day, most recent first. For each day:

• Day label (Today, Yesterday, or date)
• Count of tasks completed that day
• List of the completed tasks, each clickable to open the source file

Above the list:

• Range toggle — 7 days / 14 days / 30 days / All
• Stats — completed count, active days (days with at least one completion), unique projects touched, current streak (consecutive days with at least one completion)

The streak is useful for accountability with yourself — momentum is real, and seeing “7d streak” is a small but reliable nudge.

AI tab

Every AI operation Pensum has run, in reverse chronological order. Each entry shows:

• Status dot — green (success), red (failed), gray (aborted)
• Feature — smart-capture, smart-triage, summarization, action-extraction, stub-expansion, transcription
• Timestamp
• Mode + model — e.g. managed · claude-haiku-4-5 or byo · gpt-4o
• Cost indicator — token counts for text AI, audio minutes for transcription, latency in seconds
• Error code + message — when failed, the structured error returned by the provider or proxy

This exists because AI ops often happen in the background while you’re in another note. Without a log, you’d never see what ran, what cost, or what failed. The activity log is the transparency mechanism.

Above the list:

• Entry count
• Clear button — wipes the log (kept up to 500 entries automatically)

The status-bar dot at the bottom of Obsidian reflects the AI tab’s recent state: gray if the last 5 minutes were quiet, green if the most recent op succeeded, red with a count if there were recent failures. Click the dot to jump directly to the AI tab.

Agent Captures

Open via the command palette: Pensum: Open agent captures.

When an AI agent (via MCP) captures a task on your behalf, it lands here with agent-suggested-pending provenance rather than going straight into your active task graph. Agent Captures is the review surface.

For each pending capture:

• Accept — the task moves into the active graph with provenance flipped to agent-suggested-accepted
• Edit — open the edit modal, make changes, then accept (provenance: agent-suggested-edited)
• Discard — drop the suggestion entirely

This is the agent-side counterpart to Smart Triage’s review flow. The principle: AI never silently mutates your task graph; you always have a review step before agent-originated content becomes part of your real work.

The “require review” behavior can be turned off in Settings → Agents → Agent capture review if you trust your agents and want them to write directly into the inbox. The default is on — review required.