Maya Labs : Program machines using natural language.
The Maya Labs Python SDK provides easy async access to our PAC-1 program synthesis engine, with a CLI to instruct and generate programs from instructions in natural language.
Installation
pip install --upgrade mayalabs
Preview
demo.webm
Usage
Get the Maya Labs API key from the Settings > Developer section.
export MAYA_API_KEY = "mayakey-..."
Or set mayalabs.api_key to its value:
import mayalabs
mayalabs.api_key = "mayakey-..."
script = """
1. trigger on receive
2. research {{term}} on wikipedia
3. extract 'title' and 'summary' from tabular data
4. send response back
"""
function = mayalabs.Function(name="Scrape2")
# Creating new worker...
function.update(script=script)
# Generating program graph...
# Starting worker...
# Installing dependencies...
# Deployed!
output = function.call({"term": "Dr. Vikram Sarabhai"})
print(output)
# finds and outputs title and summary from wikipedia results
PAC-1 takes in steps written in English, writes & assembles a discrete program graph, and deploys ready-to-use software on our compute infrastructure, that you can call within your code. You can set up missing dependencies / visualize the flow of logic by following the link to flow-based editor it provides after deployment.
Check our documentation for more usage and tutorials.
Use Cases
- If...Then Conditionals and Looping
- Custom Functions
- Web Scrapers
- Repeating Workflows
- Custom Dashboards
- Data Transformation
- Platform Bots
- Business Processes
- Long-term memory
Command Line Usage
Today, script need to be manually written step-by-step, but we are offering early WIP preview of iterative script generation via our CLI. Try running:
$ mayalabs instruct -c 'fetch Name and Email from gsheet, write function to merge all columns, and return data'
And then use the generated script as function.
vid.mp4
Requirements
- Python 3.7.1+
Issues
All feedback and bug reports welcome! Report in the issues section, or mail us at humans@mayalabs.io.
Roadmap
- Cleaner logs and exception handling
- Dependency configuration UX improvement
- Import skill repositories for reusing skills created by user or present in store
- Create and call downstream auto-deployed functions
- Allow multi-profile and team profile access