https://github.com/dlt-hub/metabase_data_api
Metabase data api python wrapper for notebooks and automation
https://github.com/dlt-hub/metabase_data_api
colab-notebook colaboratory metabase notebook-jupyter pandas python
Last synced: 3 months ago
JSON representation
Metabase data api python wrapper for notebooks and automation
- Host: GitHub
- URL: https://github.com/dlt-hub/metabase_data_api
- Owner: dlt-hub
- License: mit
- Created: 2022-03-02T13:24:02.000Z (over 4 years ago)
- Default Branch: master
- Last Pushed: 2022-05-31T11:52:57.000Z (about 4 years ago)
- Last Synced: 2025-06-08T19:06:00.602Z (about 1 year ago)
- Topics: colab-notebook, colaboratory, metabase, notebook-jupyter, pandas, python
- Language: Python
- Homepage:
- Size: 273 KB
- Stars: 2
- Watchers: 3
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE.txt
Awesome Lists containing this project
README
# Why?
## Python notebook access
This package was created to enable reading data from Metabase into a python notebook for further analysis.
It's a way to provide business users with access to data in complex analyses.
For example, a product manager user may refresh an experiment evaluation or a sankey diagram by using only their metabase credentials.
Or, a marketer could use a notebook that their data team made, to get user segmentation data and push it to Mailchimp or Facebook audiences.
Possible uses:
* Adding notebook-based analyses such as a/b testing significance calculators or data based attribution models
* Adding complex visualisations such as sankey diagrams
* Programmatic access to data for other parties (via metabase access management) means you can easily manage access in metabase rather than database.
## Credential management, easy authentication and querying
The advantage over going directly to the db are :
* Unified query method: You can query in the same way a card that reads data from Google Analytics or a database
* Unified access control: Access to data can be managed in metabase for sources or even individual questions. If metabase reads from multiple source systems, this also reduces #nr of credentials.
Typical use cases are:
* Simple authentication and querying: Metabase is likely to be publicly accessible (as opposed to data warehouses that might be VPC only). Authentication is handled and does not require any complex flows.
* Possiblity to consume "drag and drop" queries created in Metabase: A non tech user could create and maintain a data feed.
# Examples:
`pip install metabase-data-api`
see the examples folder
code snippet:
```
import pandas as pd
from metabase_data_api import MetabaseApi as M
import json
session_params = dict(user='name@company.ai',
password='demopass1',
url='https://yourteam.metabaseapp.com/'
)
mb_api = M(**session_params)
query = 'SELECT 12 as col'
#get raw file data via export
d = mb_api.export_from_query(query, database_id=4)
json.loads(d.decode("utf-8"))
df = pd.DataFrame.from_records(d)
print(df)
# col
#0 12
```
# Limits
The export method uses the file download functionality which is ***limited to 1m rows***.
The get_ methods are calling the same endpoints that are used by metabase for its own charting and are ***limited to 2k rows of results***