tacit.knowledge

tacit Knowledge Base Query

You are retrieving relevant knowledge from the user's local tacit knowledge database using the tacit CLI.

Available Commands

tacit list [duration]                    — List entries created within duration (default: 1h). Supports: 30m, 1h, 24h, 1d, 7d, 2w
tacit search [--duration <d>] <pattern>  — Full-text search across all entries (title + summary + content). Optional --duration limits to entries created within that window.
tacit get <file-path>                    — Print full content of a specific entry

Process

When the user invokes /tacit.knowledge <prompt>, extract the time window from the prompt if mentioned (e.g. "today", "this week", "last hour"). If no time is specified, default to 1h for list.

Step 1 — Run list agent and search agent IN PARALLEL

List agent — retrieves recent entries:

tacit list <duration>

• Use the time window from the prompt if specified; otherwise use 1h
• Parse the output: each entry has [datetime] title, File: <path>, and optionally Summary: <first line>
• Collect all file paths from the output

Search agent — finds topically relevant entries:

1. Extract 2–4 keywords from the user's prompt (nouns, domain terms, key verbs)
2. Run tacit search <keyword> for each keyword to maximize recall. If a time window was identified, pass --duration <d> to narrow results (e.g. tacit search --duration 1h <keyword>)
3. Collect all unique file paths from results
4. Deduplicate across keywords

Each agent returns: a list of { file_path, title, created_at, summary_snippet } objects.

Step 2 — Merge results

In the main thread:

• Union the file paths from both agents, deduplicated
• If a file path appears in both, it is high-confidence relevant — prioritize it
• Limit to the 5 most relevant entries to avoid context overload (prefer recent + multi-source matches)

Step 3 — Fetch full content for top entries

Run a single command with all selected file paths at once:

tacit get <file-path-1> <file-path-2> <file-path-3> ...