Files
claw-email/docs/superpowers/specs/2026-04-28-references-threading-design.md
T
2026-04-28 04:51:38 +07:00

45 lines
1.8 KiB
Markdown

# Threading: Add References Header Fallback
## Problem
Ingress threading only resolves parent via `In-Reply-To`. When a client sets `References` without `In-Reply-To`, the bridge creates a new `thread_id` instead of attaching to the existing conversation, breaking thread continuity and downstream dispatch.
## Requirements
From `requirements.md:74`:
- If `In-Reply-To` or `References` is present, find parent task in DB and inherit `thread_id`.
- Otherwise use `message_id` as new `thread_id`.
## Design
### Current behavior (`imap.go:333-369`)
1. Parse MIME body headers, extract `In-Reply-To`
2. If found, lookup parent by message ID → inherit `thread_id`
### New behavior
1. Parse MIME body headers, extract both `In-Reply-To` and `References`
2. Try `In-Reply-To` first (unchanged)
3. If no parent found via `In-Reply-To`, parse `References` header (space-separated `<msg-id>` list)
4. Iterate `References` from **last to first** (most recent ancestor first)
5. First match in DB wins → inherit `thread_id`
### Why last-to-first
`References` lists message IDs chronologically: `<oldest> ... <newest>`. The last entry is the most recent message in the chain, most likely already processed by the bridge.
### Scope
- Single file change: `internal/mail/imap.go` (threading block in `processMessage`, ~10 lines)
- New tests: `internal/mail/imap_test.go` (References parsing, fallback, multi-ID)
- No changes to DB, config, API, or other packages
### Test cases
1. `In-Reply-To` matches → existing behavior unchanged
2. `In-Reply-To` absent, `References` has match → inherits thread_id from most recent match
3. `References` has multiple IDs, only older one in DB → still matches
4. Both headers absent → thread_id = message_id (unchanged)
5. `References` present but no match in DB → thread_id = message_id (new thread)