notion-sdk（Notion SDK 搜索）

Notion SDK Skill

Control Notion programmatically using the official notion-client Python SDK. See PyPI for current version.

Self-Evolving Skill: This skill improves through use. If instructions are wrong, parameters drifted, or a workaround was needed — fix this file immediately, don't defer. Only update for real, reproducible issues.

When to Use This Skill

Use this skill when:

• Creating pages or databases in Notion via API
• Querying Notion databases programmatically
• Adding blocks (text, code, headings) to Notion pages
• Automating Notion workflows with Python
• Integrating external data sources with Notion

Preflight: Token Collection

Before any Notion API operation, collect the integration token:

AskUserQuestion(questions=[{
    "question": "Please provide your Notion Integration Token (starts with ntn_ or secret_)",
    "header": "Notion Token",
    "options": [
        {"label": "I have a token ready", "description": "Token from notion.so/my-integrations"},
        {"label": "Need to create one", "description": "Go to notion.so/my-integrations → New integration"}
    ],
    "multiSelect": false
}])

After user provides token:

1. Validate format (must start with ntn_ or secret_)
2. Test with validate_token() from scripts/notion_wrapper.py
3. Remind user: Each page/database must be shared with the integration

Quick Start

1. Create a Page in Database

from notion_client import Client
from scripts.create_page import (
    create_database_page,
    title_property,
    status_property,
    date_property,
)

client = Client(auth="ntn_...")
page = create_database_page(
    client,
    data_source_id="abc123...",  # Database ID
    properties={
        "Name": title_property("My New Task"),
        "Status": status_property("In Progress"),
        "Due Date": date_property("2025-12-31"),
    }
)
print(f"Created: {page['url']}")

2. Add Content Blocks

from scripts.add_blocks import (
    append_blocks,
    heading,
    paragraph,
    bullet,
    code_block,
    callout,
)

blocks = [
    heading("Overview", level=2),
    paragraph("This page was created via the Notion API."),
    callout("Remember to share the page with your integration!", emoji="⚠️"),
    heading("Tasks", level=3),
    bullet("First task"),
    bullet("Second task"),
    code_block("print('Hello, Notion!')", language="python"),
]
append_blocks(client, page["id"], blocks)

3. Query Database

from scripts.query_database import (
    query_data_source,
    checkbox_filter,
    status_filter,
    and_filter,
    sort_by_property,
)

# Find incomplete high-priority items
results = query_data_source(
    client,
    data_source_id="abc123...",
    filter_obj=and_filter(
        checkbox_filter("Done", False),
        status_filter("Priority", "High")
    ),
    sorts=[sort_by_property("Due Date", "ascending")]
)
for page in results:
    title = page["properties"]["Name"]["title"][0]["plain_text"]
    print(f"- {title}")

Available Scripts

References

• Property Types - All 24 property types with examples
• Block Types - All block types with structures
• Rich Text - Formatting, links, mentions
• Pagination - Handling large result sets

Important Constraints

Rate Limits

• 3 requests/second average (burst tolerated briefly)
• Use api_call_with_retry() for automatic rate limit handling
• 429 responses include Retry-After header

Authentication Model

• Page-level sharing required (not workspace-wide)
• User must explicitly add integration to each page/database:

- Page →... menu → Connections → Add connection → Select integration

API Version (v2.6.0+)

• Uses data_source_id instead of database_id for multi-source databases
• Legacy database_id still works for simple databases
• Scripts handle both patterns automatically

Operations NOT Supported

• Workspace settings modification
• User permissions management
• Template creation/management
• Billing/subscription access

API Behavior Patterns

Insights discovered through integration testing (test citations for verification).

Rate Limiting & Retry Logic

api_call_with_retry() handles transient failures automatically:

*Citation: test_client.py::TestRetryLogic (lines 146-193)*

Read-After-Write Consistency

Newly created blocks may not be immediately queryable. Add 0.5s minimum delay:

append_blocks(client, page_id, blocks)
time.sleep(0.5)  # Eventual consistency delay
children = client.blocks.children.list(page_id)

*Citation: test_integration.py::TestBlockAppend::test_retrieve_appended_blocks (line 298)*

v2.6.0 API Migration

*Citation: test_integration.py::TestDatabaseQuery (line 110)*

Archive-Only Deletion

Pages cannot be permanently deleted via API - only archived (moved to trash):

client.pages.update(page_id, archived=True)  # Trash, not delete

*Citation: test_integration.py cleanup fixture (lines 72-76)*

Edge Cases & Validation

Property Builder Edge Cases

*Citation: test_property_builders.py::TestPropertyBuildersEdgeCases (lines 302-341)*

Input Validation Responsibility

Builders are intentionally permissive - validation happens at API level:

Best Practice: Validate in your application before building properties.

*Citation: test_property_builders.py::TestPropertyBuildersInvalidInputs (lines 347-376)*

Token Validation

• Case-sensitive: Only lowercase ntn_ and secret_ valid
• Format check happens before API call (saves unnecessary requests)
• Empty/whitespace tokens rejected immediately

*Citation: test_client.py::TestClientEdgeCases (lines 196-224)*

Query & Filter Patterns

Compound Filter Composition

# Empty compound (matches all)
and_filter()  # {"and": []}

# Deep nesting supported
and_filter(
    or_filter(filter_a, filter_b),
    and_filter(filter_c, filter_d)
)

*Citation: test_filter_builders.py::TestFilterEdgeCases (lines 323-360)*

Filter Limitations

Filters don't exclude NULL properties - check in Python:

if row["properties"]["Rating"]["number"] is not None:
    # Process non-null values

*Citation: test_integration.py::TestDatabaseQuery::test_query_database_with_filter (lines 120-135)*

Pagination Invariants

Always check has_more before using next_cursor.

*Citation: test_integration.py::TestDatabaseQuery::test_query_database_with_pagination (lines 137-151)*

Error Handling

from notion_client import APIResponseError, APIErrorCode

try:
    result = client.pages.create(...)
except APIResponseError as e:
    if e.code == APIErrorCode.ObjectNotFound:
        print("Page/database not found or not shared with integration")
    elif e.code == APIErrorCode.Unauthorized:
        print("Token invalid or expired")
    elif e.code == APIErrorCode.RateLimited:
        print(f"Rate limited. Retry after {e.additional_data.get('retry_after')}s")
    else:
        raise

Installation

uv pip install notion-client  # v2.6+ required for data_source support

Or use PEP 723 inline dependencies (scripts include them).

Troubleshooting

Post-Execution Reflection

After this skill completes, check before closing:

1. Did the command succeed? — If not, fix the instruction or error table that caused the failure.
2. Did parameters or output change? — If the underlying tool's interface drifted, update Usage examples and Parameters table to match.
3. Was a workaround needed? — If you had to improvise (different flags, extra steps), update this SKILL.md so the next invocation doesn't need the same workaround.

Only update if the issue is real and reproducible — not speculative.