Mantis SDK

Python SDK for interacting with the Mantis visualization platform. Provides programmatic control over space creation, data visualization, and analysis.

Installation

python -m venv venv

# Windows
.\venv\Scripts\activate

# macOS/Linux
source venv/bin/activate

python -m pip install -r ./src/requirements.txt

python -m playwright install chromium

Note: If the Playwright installation fails, try running:

python -m playwright install-deps

before proceeding.

Authentication

The MantisClient object requires the parameter cookie to be passed in. This can be obtained by logging into the Mantis webpage in your browser, opening devtools, and finding the session cookies that are passed to the requests. The following keys in cookies are required: session-id, next-auth.callback-url, next-auth.csrf-token, next-auth.session-token. Use the cookie string as the cookie param.

Quick Start

Note: If you are using a local backend, you must run it with docker, or the space creation will NOT work. To do so, cd docker from the backend root, then docker compose up -d --build. After you run the build once, you can re-run it simply with docker compose up. If things dno't work, check the logs and make sure they are not empty.

from client import MantisClient, SpacePrivacy, DataType, ReducerModels 
from render_args import RenderArgs 
import pandas as pd

mantis = MantisClient("/api/proxy/", cookie)

# Create DF (Real data will need more points)
df = pd.DataFrame({
    "Symbol": ["AAPL", "GOOGL", "MSFT", "AMZN", "FB"],
    "Market Cap": [2000, 1500, 1300, 1200, 800],
    "Description": ["Apple Inc.", "Alphabet Inc.", "Microsoft Corporation", "Amazon.com Inc.", "Facebook Inc."]
})

# Set types
data_types = {"Symbol": DataType.Title,
              "Market Cap": DataType.Numeric,
              "Description": DataType.Semantic}

# Make space
new_space_id = mantis.create_space("Stock data", 
                                   data=df, 
                                   data_types=data_types,
                                   reducer=ReducerModels.UMAP,
                                   privacy_level=SpacePrivacy.Private)["space_id"]

# Open space
space = await mantis.open_space(space_id)

# Interact with space
await space.select_points(100) 
plot = await space.render_plot("Market Cap", "embed_y")

# Close when done
await space.close()

Key Features

Create spaces from DataFrames or CSV files Control visualization layout and rendering Execute code within spaces Generate plots and capture screenshots Manage panels and space configuration Support for async operations

Data Types

DataType.Title # Title/ID field 
DataType.Semantic # Text for semantic analysis
DataType.Numeric # Numerical values 
DataType.Categoric # Categorical values 
DataType.Date # Date/time values 
DataType.Links # URL/link fields 

Space Privacy Levels

SpacePrivacy.PUBLIC # Visible to all users 
SpacePrivacy.PRIVATE # Only visible to owner 
SpacePrivacy.SHARED # Visible to specific users 

Dimension Reduction

ReducerModels.UMAP # UMAP reduction 
ReducerModels.PCA # PCA + UMAP 
ReducerModels.TSNE # t-SNE 

AI Providers

AIProvider.OpenAI
AIProvider.HuggingFace

Examples

Create Space from CSV

# Load Mantis
mantis = MantisClient("/api/proxy/", cookie)

# Set data path + types
data_path = "./StockData.csv"

data_types = {"Symbol": DataType.Title,
              "Market Cap": DataType.Numeric,
              "Description": DataType.Semantic}

# Create space
new_space_id = mantis.create_space("Stock data,", 
                                   data=data_path, 
                                   data_types=data_types,
                                   reducer=ReducerModels.UMAP,
                                   privacy_level=SpacePrivacy.PRIVATE)["space_id"]

Capture Visualization

# Select 100 points
await space.select_points (100)

# Plot embed dimensions
plot = await space.render_plot ("embed_x", "embed_y")

Run Analysis

code = """

computation = 6**4
print ('Hello from SDK, :P -> ' + str(computation))

"""

# Wait for code execution to finish (returns output)
await space.run_code (code)

Manage Panels

await space.close_panel ("bags")
await space.close_panel ("quicksheet")
await space.close_panel ("userlogs")