conversations-by-date — L1 temporal browse
Framework context
L1 retrieval scoped to a single date. Returns distinct conversation_ids with titles and sources. Pure function of (date, corpus).
When to invoke
- User cites a specific date: "last Tuesday", "yesterday", "on 2026-04-01"
- Pair with
daily-memory-filefor narrative reconstruction - Pre-step before tunnel-state when the domain is unknown but the date is known
Input
date: str # YYYY-MM-DD
limit: int = 30
Output contract
Markdown list of distinct conversations from that date. Each row: title, source, model, created timestamp, conv_id citation.
Does NOT do
- Handle relative date strings ("yesterday") — caller must resolve to YYYY-MM-DD
- Return messages — only conversation summaries
- Fabricate empty-day output — explicit "No conversations found on {date}"
Verification checklist
- DATE cast on both sides of comparison (timezone-safe)
- DISTINCT on conversation_id
- ORDER BY created DESC
Source: mordechaipotash/brain-mcp — distributed by TomeVault.