tim

Create an edge (relationship) between two entries.

Follow an edge chain from a starting entry (BFS traversal).

Re-confirm entries as still valid without editing them. Stamps metadata.verified_at — clears the stale annotation on reads and the stale count in tim_health.

Delete an entry (soft: mark irrelevant, hard: tombstone).

Batch hard-delete entries by id (max 100). Skips missing ids.

Sync operations: push staging records, pull from remote, or check status.

Suppress entries matching a pattern: hidden from tim_search, tim_read, tim_show, tim_section_children, tim_remember, and tim_load_project. Optional TTL (e.g. "24h", "7d").

Run health diagnostics: broken links, orphans, FTS integrity, counts, memory coverage and embedding backlog.

Content statistics: entry counts, content size aggregates, length buckets, and breakdown by metadata.kind.

List direct children of a project section in compact form.

Export TIM database to markdown or .hmem SQLite format.

Import entries from a .hmem SQLite file. Always call tim_import_manifest first, then tim_import with dryRun:true, then live import, then tim_import_audit. Never use raw SQL for hmem migration.

Read a .hmem file without importing it: format, labels, titles, and node counts.

Return one project structure: root, direct sections, duplicate sections, loose direct children, and imported hmem provenance.

Report-first duplicate detector: scans a project for near-duplicate entries (identical/similar titles or embeddings) and enqueues them as pending curation candidates — never deletes anything. Review the returned pairs, then consolidate via tim_move_entry/tim_delete, or reject.

Post-import audit for hmem migrations: missing sections, duplicate sections, loose imported nodes, health issues, and repair suggestions.

Preview tim_move_entry impact without writing: current parent/depth, target parent/depth, descendant count, and cycle risk.

Ensure a canonical project section exists and optionally move specified children into it. Uses TIM writes/moves, never SQL edits.

Run comprehensive diagnostics: config, DB, API connectivity.

Start a TIM session (idempotent). With projectId (or default P0000 Inbox), creates nested Sessions/Summary/Exchanges tree. Without a resolvable project the response lists recently active projects — follow its ACTION line to bind one.

Append exchange entries to a session log.

Return the next unsummarized batch of exchanges for a session (UUIDs + user/agent bodies). Summarizer reads this, writes a Batch node under Summary.

Scan ALL sessions and return every unsummarized batch. Use at startup for cleanup sweep of stale batches (crashed summarizer, missed triggers). No parameters needed.

Return batch-summary nodes that have only structural tags (#session-summary, #batch-summary) and no content hashtags. Use for re-tagging failed or legacy summaries.

Write an idempotent Batch summary node under the session Summary tree. Used by tim-summarizer CLI.

Record a git commit under the project Commits section. Idempotent by hash. Links to session via relates/implements when sessionId given.

Atomically rename an entry ID and update all references (edges, parent_id, staging, metadata).

Move an entry under a new parent and cascade depth updates to descendants. Preview with tim_dry_run_move before moving imported or ambiguous nodes.

Batch-update irrelevant and/or favorite flags on multiple entries (flags only, never content).

Add tags to an entry (deduplicated). Deprecated status/priority tags (#todo, #done, #priority-*) are skipped — use metadata.task.status / metadata.task.priority.

Remove tags from an entry.

Rename a tag across all entries (exact match only, safe for substring collisions).

Create a project in exactly one mode. Every project representing files on disk MUST pass its absolute path; memoryOnly:true is only for an intentionally virtual/database-only project and is never a shortcut for an unknown cwd.

Load a project by label or alias and bind the session once. Rejects a different project if the session is already bound — pass bind:false for cross-project reads (replaces tim_read_project).

[DEPRECATED — use tim_load_project with bind:false] Read a project brief + tree WITHOUT binding the session (cross-project lookup). Use tim_load_project to start working on a project.

Show error statistics: total errors, top errors, error rate, alert thresholds (>5 identical errors in 1h).

Log an error entry. Used by CLI tools and summarizer for structured error tracking.