Search & Sequences

Backlog.md provides a unified fuzzy-search service built on Fuse.js, covering tasks, documents, and decision records.

Run backlog search followed by a keyword to search across all projects:

backlog search "user login"

Filters

Option Description Example
--type <type> Restrict result type (task, document, decision) --type task
--status <status> Filter by task status (repeatable or comma-separated for multiple) --status "In Progress"
--exclude-status <status> Exclude tasks with the given status --exclude-status "Done"
--unassigned Show only tasks with no assignee --unassigned
--priority <priority> Filter by priority (high, medium, low) --priority high
--modified-file <path> Filter by substring of a modified file path --modified-file src/api.ts
--completed Extend the search scope to historical tasks in the completed corpus --completed
--limit <number> Limit the total number of returned results --limit 10
--plain Plain-text output --plain

Combined Search Examples

# Search for tasks containing "api" and in "In Progress" status
backlog search "api" --status "In Progress"

# Search for high-priority bugs
backlog search "bug" --priority high --type task

# Search for changes related to a specific file
backlog search "auth" --modified-file src/auth.ts

# Search only documents and decisions
backlog search "architecture" --type document --type decision

# Exclude completed tasks
backlog search "api" --exclude-status "Done"

# Show only unassigned high-priority tasks
backlog search "api" --unassigned --priority high

# Show the first 20 results, then the next page
backlog search "api" --max-count 20
backlog search "api" --max-count 20 --skip 20

# Output only the total match count
backlog search "api" --count

List Paging

backlog search and the eight list commands task list, draft list, milestone list, doc list, doc search, decision list, and memo list share the same grep-style paging window options:

Option Description Example
--max-count <n> Output at most n items --max-count 20
--skip <n> Skip the first n items (next page) --skip 20
--count Output only the total match count (a single number) --count
  • The window is applied after filtering, sorting, and --limit, in output order; consecutive pages do not overlap or skip items.
  • When output is truncated, it ends with Showing <first>-<last> of <total> items. Next: backlog <command> --skip <n>; copy the suggested Next command to fetch the next page.
  • --count outputs only a number and cannot be combined with --json (returns a non-zero exit code).
  • The JSON envelope appends total / nextSkip fields only when the window truncates the output; if nothing is truncated, the output is byte-for-byte unchanged.
  • For milestone list, the window unit is a milestone; the --with-no-milestone "No Milestone" section is always placed first, does not consume the quota, and is not counted by --count.

backlog instructions overview contains a List Paging Quick Reference for these options; each command chapter (tasks, drafts, milestones, documents, decisions, memos) documents paging for that command.

In the interactive backlog board or backlog task list interface, typing a keyword filters the list in real time. TUI search responds instantly and updates results as you type, without requiring Enter.

After starting the Web interface with backlog browser, press Ctrl+K (Windows/Linux) or Cmd+K (macOS), or click the search trigger in the sidebar, to open the global search dialog. It searches tasks, drafts, documents, decisions, Wiki pages, and memos in one go. Results are grouped by type, with the keyword and resource ID highlighted, and support keyboard navigation and scroll-position memory. Query and type filters are synchronized to /search?q=...&type=... links that can be refreshed or shared. See Global Search for details.

CLI backlog search plain-text output covers only tasks, documents, and decisions; memos and Wiki results are Web-only types and appear only in Web search and JSON output.

Search Score Threshold Alignment

CLI, TUI, MCP, and Web UI share the same Fuse.js score threshold of 0.45, filtering out low-relevance results. This prevents short numbers such as 63 from accidentally matching unrelated task IDs while preserving Fuse semantics for text queries.

Sequences

What Sequences Are

Sequences are groups of tasks that can run in parallel, automatically computed from task dependencies. They help you understand, without manual scheduling:

  • Critical path: which tasks must finish before others can start
  • Parallelizable work packages: tasks in the same sequence can be advanced simultaneously by different team members
  • Project bottlenecks: the longest dependency chain often determines overall duration

Viewing Sequences

backlog sequence list

By default this opens the sequence view in interactive TUI mode. Use --plain for plain-text output:

backlog sequence list --plain

Reading the Sequence Output

Unsequenced:
  back-12 - Optimize homepage load speed

Sequence 1:
  back-1 - Design database tables
  back-2 - Set up project skeleton

Sequence 2:
  back-3 - Implement user registration API
  back-4 - Implement user login API

Sequence 3:
  back-5 - Deploy to test environment
  • Unsequenced: tasks that have no dependencies, are not depended on by any other task, and have no ordinal. They block nothing and are blocked by nothing.
  • Sequence N: the Nth layer of parallelizable tasks. Sequence 1 has no prerequisites; Sequence 2 depends on one or more tasks in Sequence 1, and so on.
  • Tasks marked Done are not shown in sequences.

TUI Sequence View Operations

backlog sequence list now uses a dual-panel layout:

  • Left sidebar: lists Unsequenced tasks and Sequence 1, Sequence 2, etc.; click to switch the focused sequence.
  • Right task list: shows tasks in the currently selected sequence (or Unsequenced). Use arrow keys or j/k to browse; press Enter to view task details.
  • Move mode: press m in the task list to adjust task order and automatically update dependencies. The move preview is computed by the pure function buildMovePreview, so what you see is what you get.
  • Press Escape or q to exit.

The old recursive select item implementation could stack-overflow on deep sequences; this has been eliminated with pure functions such as buildSequenceRows.

Web UI Sequence Page

In the browser interface, the sequence page supports drag-and-drop reordering. Dropping a task card at a different sequence position automatically updates the related dependencies.

Sequence Quick Reference in CLI Instructions

backlog instructions overview includes a quick reference for sequence commands:

backlog sequence list --plain

This command prints a plain-text sequence hierarchy. Sequences are derived from task dependencies; tasks within the same sequence can run in parallel, while different sequences have a happens-before relationship. See the "Viewing Sequences" and "TUI Sequence View Operations" sections above for more details.

What Sequences Mean for Project Planning

Use Case Action
Planning a Sprint Assign sequence layers to different iteration cycles
Identifying Risk Delays on the critical path directly impact overall progress
Resource Allocation Assign tasks in the same sequence to different developers to run in parallel
Dependency Review Run backlog sequence list regularly to check for unreasonable hard dependencies

results matching ""

    No results matching ""