Skip to content
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
44 changes: 26 additions & 18 deletions VISION.md
Original file line number Diff line number Diff line change
@@ -1,34 +1,42 @@
# MailPlus Intelligence Vision

Version: 0.1
Version: 0.2

MailPlus Intelligence is a privacy-aware analysis layer for Synology MailPlus data, focused on useful local intelligence rather than cloud-dependent mailbox automation.
MailPlus Intelligence is the structured recall and semantic intelligence layer for email. Synology MailPlus remains the canonical raw mail archive; this repo indexes metadata, reconstructs threads, classifies lanes, suppresses noise, and promotes selected reviewed insights into durable memory surfaces.

It should help the operator understand mail patterns, threads, references, and governance-relevant signals while keeping fixture-bound tests and private data boundaries explicit.
The product is not a mailbox warehouse and not cloud mailbox automation. Its value is controlled extraction: keep raw messages and attachments in MailPlus, then derive enough metadata, summaries, obligations, decisions, events, and correspondence context to support search, planning, memory, and approved follow-up tasks.

## Who It Serves

- The operator managing Synology-hosted mail.
- Agents repairing parsing, threading, and governance workflows.
- Future users who need local mail intelligence without handing over mailbox contents.
- The operator who wants useful recall across MailPlus without duplicating raw mail into long-term memory.
- Agents that need fixture-backed thread, classification, and semantic-output contracts before acting on email-derived information.
- Future integration surfaces such as wiki pages, `memory/`, `MEMORY.md`, and reminders, where only reviewed derived intelligence should land.

## Current Product Boundary

- v0.1 is fixture-mode: metadata fixture sync, SQLite schema bootstrap, deterministic thread reconstruction, lane classification, noise suppression, deterministic semantic extraction, optional LLM extraction, promotion queue review, dry-run exporters, scheduler locks, CLI inspection, and `mpi doctor`.
- Live MailPlus/IMAP ingestion is documented but not connected.
- Production export to wiki, memory, or reminders is dry-run only until explicit live export work lands.
- Raw mail, attachments, and retention history remain in MailPlus.

## Product Principles

- Fixture-bound validation is the default proof path.
- Malformed references and edge cases should become tests.
- Public readiness requires careful separation from live private mail.
- Threading behavior must be explainable and inspectable.
- Governance docs should match the code's actual privacy boundary.
- Reference raw mail; do not copy it into durable memory surfaces.
- Metadata and derived semantic outputs must be bounded, reviewable, and redacted according to the privacy boundary.
- Fixture-bound validation is the default proof path, especially for malformed references, threading edge cases, classification, and extraction.
- Promotion is an explicit decision. High-value summaries and tasks should queue for review before becoming durable memory.
- Live adapters must preserve the canonical-store boundary and be testable without private mailbox content.

## Near-Term Direction

- Harden threading and malformed-reference handling.
- Keep offline fixtures representative and safe.
- Separate governance checkout work from live MailPlus integration work.
- Improve public-readiness docs and validation gates.
- Keep Phase 2 focused: live MailPlus adapter, live export contract, and promotion workflow without weakening raw-mail boundaries.
- Harden thread reconstruction, selected-message text cache policy, noise suppression, and semantic output schema.
- Improve `mpi doctor`, scheduler locks, and dry-run exporter evidence so operators can trust what would be promoted.
- Keep public docs, fixtures, and tests representative without exposing real mailbox contents.

## Non-Goals

- Do not depend on live private mail for routine tests.
- Do not expose mailbox content in docs, logs, or generated artifacts.
- Do not turn local intelligence into cloud mailbox surveillance.
- Do not become the canonical mail archive or attachment store.
- Do not require live private mail for routine tests.
- Do not auto-export obligations, summaries, or reminders without explicit approval.
- Do not leak raw message bodies, attachment contents, credentials, or private correspondents into docs, logs, fixtures, or PR text.
Loading