Notion工作区优化工具

Notion Workspace Optimizer

Local page-metadata scoring, recommendation labeling, and title-prefix grouping.

This repository is an independent portfolio project. It is not affiliated with Notion and it does not connect to, read from, mutate, archive, restructure, or otherwise control a live Notion workspace.

Implemented mechanisms

Page-policy scoring

src/optimizer.py accepts caller-supplied Page metadata and computes a bounded local policy score from:

• days stale;
• nesting depth;
• outgoing-link count; and
• child-page count.

The score is a deterministic heuristic used to rank items for review. It is not a probability, confidence estimate, semantic similarity score, measured workspace-health metric, or learned model output.

Recommendation labels

The optimizer maps ranked page metadata to local suggestion labels:

• ARCHIVE_CANDIDATE for sufficiently stale leaf pages;
• COLLAPSE_NESTING for deeply nested pages;
• SPLIT_HUB for highly linked pages; and
• REVIEW otherwise.

These labels do not perform any external action. They are recommendations returned in a Python data structure with operational_authority=false.

Title-prefix grouping

src/cluster.py groups caller-supplied page references by the lower-cased text before the first - in the title. This is literal deterministic prefix grouping. It is not semantic hashing, embedding similarity, natural-language clustering, or duplicate detection.

Input boundary

The scoring and grouping functions reject malformed control parameters and impossible negative page metrics. A caller may also supply a timezone-aware observation timestamp to reproduce the complete optimizer envelope deterministically.

What this repository does not establish

• No Notion affiliation, endorsement, employment, or proprietary access.
• No live Notion API, database, page, block, user, or workspace access.
• No automatic archiving, relation rewrites, database-index tuning, or workspace mutation.
• No semantic hashing, embeddings, duplicate-page detection, or learned clustering.
• No live MCP, APEX, Mastermind, or provider-mesh integration.
• No production operational authority, scale, reliability, latency, throughput, or security guarantee.

Verification

Repository CI preserves the existing repository-owned verifier and adds a bounded Python contract matrix. The relevant local checks are:

python -m pytest -q
python scripts/verify_public_truth.py

The public-truth gate fails closed if live-provider claims, semantic/duplicate-detection claims, stale hyper-scaling metadata, or stale promotion authority reappear.

Example

from datetime import datetime, timezone
from src.optimizer import Page, optimize

pages = [
    Page("old", "old-note", depth=3, links=1, days_stale=120, children=0),
    Page("hub", "ops-hub", depth=1, links=30, days_stale=10, children=12),
]

result = optimize(
    pages,
    top_k=2,
    observed_at=datetime(2026, 8, 14, tzinfo=timezone.utc),
)
print(result["actions"])

The example uses only local caller-supplied metadata and emits recommendation labels, not workspace mutations.