STROT SDK

Build tools, agents, pipelines, and dashboards for your STROT instance — in Python.

Installation

pip install strot-sdk

# With CLI (login, init, deploy commands)
pip install strot-sdk[cli]

# With pandas support (query_df)
pip install strot-sdk[pandas]

# Everything
pip install strot-sdk[cli,pandas]

Quick Start

# 1. Authenticate
strot login

# 2. Scaffold a project
strot init tool my-calculator

# 3. Edit main.py (use Claude Code, Cursor, or any editor)
cd my-calculator

# 4. Test locally
strot test

# 5. Deploy to your STROT instance
strot deploy

SDK Reference

Tools (@function)

from strot_sdk import function, llm

@function(
    name='calculate_roi',
    description='Calculate return on investment',
    category='finance',
    parameters=[
        {'name': 'cost', 'type': 'number', 'description': 'Total cost'},
        {'name': 'revenue', 'type': 'number', 'description': 'Total revenue'},
    ],
    returns={'type': 'number', 'description': 'ROI percentage'}
)
class CalculateROI:
    def run(self, cost: float, revenue: float) -> float:
        return ((revenue - cost) / cost) * 100

Agents (@agent)

from strot_sdk import agent

@agent(
    name='sales_analyst',
    description='Analyzes sales data and provides insights',
    tools=['calculate_roi', 'top_n'],
    model='gpt-4o',
    temperature=0.1,
)
class SalesAnalyst:
    system_prompt = """You are a sales analyst.
    Analyze data and provide actionable recommendations."""

Cortex Pipelines (@cortex)

Build data pipelines that compile to JSON DSL and deploy to your STROT instance.

from strot_sdk import cortex
from strot_sdk.cortex import Flow

@cortex(name='daily_etl', description='Daily ETL pipeline')
class DailyETL:
    def build(self, flow: Flow):
        # Load data from a saved query
        data = flow.data_connector('load_sales', query_id=42)

        # Transform with LLM
        cleaned = flow.transform(data, prompt='Clean and normalize the data')

        # Route based on content
        router = flow.router(cleaned, routes=[
            {'name': 'high_value', 'description': 'Orders over $1000'},
            {'name': 'standard', 'description': 'Regular orders'},
        ], prompt='Classify by order value')

        # Publish results
        flow.publish(router, name='daily_report', destination='slack', channel='#data')

Available Flow methods:

Pages / Dashboards (@page)

Build dashboards that compile to JSON layout and deploy to your STROT instance.

from strot_sdk import page
from strot_sdk.pages import Dashboard, Row, KPI, Chart, Table

@page(name='sales_dashboard', description='Sales overview', type='dashboard')
class SalesDashboard:
    def layout(self):
        return Dashboard(
            Row(
                KPI(query_id=1, label='Revenue', format='currency'),
                KPI(query_id=2, label='Orders'),
                KPI(query_id=3, label='Customers'),
                KPI(query_id=4, label='Avg Order', format='currency'),
            ),
            Row(
                Chart(query_id=5, type='line', title='Revenue Trend', span=8),
                Chart(query_id=6, type='donut', title='By Region', span=4),
            ),
            Row(
                Table(query_id=7, title='Recent Orders', sortable=True),
            ),
        )

Block types: KPI, Chart, Table, Text, StatGrid, ProgressList

Chart types: line, bar, area, donut, scatter, stacked_bar, funnel

Grid: 12-column layout. Set span on any block (default varies by type).

LLM

All LLM calls go through your STROT instance — no API keys needed in your code.

from strot_sdk import llm

# Simple completion
result = llm.complete("Summarize this: " + text)
result = llm("Shorthand syntax works too")

# Chat
result = llm.chat([
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "What is 2+2?"},
])

# Classify
category = llm.classify("Great product!", ["positive", "negative", "neutral"])

# Extract structured data
data = llm.extract("John is 30 years old", {"name": "string", "age": "number"})

# Transform data
result = llm.transform(data, "Convert to French", output_format="json")

Data Access

from strot_sdk import strot, query, query_one, query_df

# Via registry (saved queries by name)
rows = strot.queries['monthly_sales'].execute()

# Via data source
rows = strot.dataSources['production'].query("SELECT * FROM users LIMIT 10")

# Direct SQL
rows = query("SELECT * FROM users", data_source_id=1)
row = query_one("SELECT * FROM users WHERE id = 1", data_source_id=1)

# Pandas DataFrame (requires strot-sdk[pandas])
df = query_df("SELECT * FROM orders", data_source_id=1)

Destinations

from strot_sdk import email, slack, webhook

email.send(to="team@example.com", subject="Report Ready", body="The daily report is ready.")
slack.send(channel="#alerts", message="New alert!")
webhook.post(url="https://api.example.com/hook", data={"event": "deploy"})

CLI Reference

Authentication

strot login                              # Interactive (opens browser)
strot login --token sk_live_abc123       # Direct API key
strot login -i https://app.strot.ai -o <org-uuid>
strot whoami                             # Show current user/org
strot logout                             # Clear credentials
strot logout --all                       # Clear all profiles

Project Scaffolding

strot init tool my-calculator            # Python tool
strot init agent my-analyst              # Python agent
strot init cortex my-pipeline            # Cortex pipeline
strot init page my-dashboard             # Page/dashboard
strot init tool my-tool -d "Description" -c finance

Resources

strot resources                          # List all resources
strot resources queries                  # List saved queries
strot resources datasources              # List data sources
strot resources tools                    # List deployed tools

Testing & Deployment

strot test                               # Run/compile locally
strot test -p input_text="Hello"         # Run with parameters
strot deploy                             # Deploy to STROT instance
strot deploy --dry-run                   # Validate without deploying

Configuration

Credentials are stored in ~/.strot/credentials (YAML, 0600 permissions):

version: 1
current_profile: default
profiles:
  default:
    url: https://app.strot.ai
    api_key: sk_live_abc123
    org: 98bf9a0a-c9cd-42a8-9ea4-4f7fee9a4535
    user_email: dev@example.com

Priority chain (highest to lowest):

1. Constructor arguments (StrotClient(url=..., api_key=...))
2. Environment variables (STROT_URL, STROT_API_KEY)
3. Credentials file (~/.strot/credentials)

Multiple profiles:

strot login --profile staging -i https://staging.strot.ai
strot login --profile production -i https://app.strot.ai

Development

git clone https://github.com/strot-ai/strot-sdk.git
cd strot-sdk
poetry install --with dev,cli

# Run tests
poetry run pytest -v

# Run with coverage
poetry run pytest --cov=strot_sdk

License

MIT