CtrlK
BlogDocsLog inGet started
Tessl Logo

tessl/pypi-aleph-alpha-client

Python client to interact with Aleph Alpha API endpoints

Pending

Quality

Pending

Does it follow best practices?

Impact

Pending

No eval scenarios have been run

Overview
Eval results
Files

steering.mddocs/

Steering & Content Control

Create and use steering concepts to guide model behavior and output style. Steering allows fine-grained control over how models generate content by defining positive and negative examples.

Capabilities

Steering Paired Examples

Define positive and negative examples to teach the model desired behavior patterns.

class SteeringPairedExample:
    negative: str
    positive: str
    
    def __init__(self, negative: str, positive: str):
        """
        Paired example showing negative and positive versions of content.
        
        Parameters:
        - negative: Example of undesired style or content
        - positive: Example of desired style or content
        """

    def to_json(self) -> Mapping[str, Any]:
        """Convert example to JSON format."""

Steering Concept Creation

Create steering concepts from paired examples that can be applied to completion requests.

class SteeringConceptCreationRequest:
    examples: List[SteeringPairedExample]
    
    def __init__(self, examples: List[SteeringPairedExample]):
        """
        Creates a new steering concept that can be used in completion requests.
        
        Parameters:
        - examples: A list of SteeringPairedExample objects showing desired transformations
        """

    def to_json(self) -> Mapping[str, Any]:
        """Convert request to JSON format."""

Steering Concept Response

Response containing the created steering concept identifier.

class SteeringConceptCreationResponse:
    id: str
    
    def __init__(self, id: str):
        """
        Response from steering concept creation.
        
        Attributes:
        - id: Unique identifier for the created steering concept
        """

    @staticmethod
    def from_json(json: Dict[str, Any]) -> SteeringConceptCreationResponse:
        """Create response from JSON data."""

Steering Methods

Create and manage steering concepts using synchronous and asynchronous clients.

def create_steering_concept(
    self, 
    request: SteeringConceptCreationRequest
) -> SteeringConceptCreationResponse:
    """
    Create a new steering concept.
    
    Parameters:
    - request: Steering concept configuration with examples
    
    Returns:
    SteeringConceptCreationResponse with concept ID
    """

async def create_steering_concept(
    self, 
    request: SteeringConceptCreationRequest
) -> SteeringConceptCreationResponse:
    """
    Create a new steering concept (async).
    
    Parameters:
    - request: Steering concept configuration with examples
    
    Returns:
    SteeringConceptCreationResponse with concept ID
    """

Usage Examples

Creating and using steering concepts for content control:

from aleph_alpha_client import (
    Client, 
    SteeringPairedExample, 
    SteeringConceptCreationRequest,
    CompletionRequest,
    Prompt
)

client = Client(token="your-api-token")

# Create steering examples for casual vs formal language
examples = [
    SteeringPairedExample(
        negative="I appreciate your valuable feedback on this matter.",
        positive="Thanks for the real talk, fam."
    ),
    SteeringPairedExample(
        negative="The financial projections indicate significant growth potential.",
        positive="Yo, these numbers are looking mad stacked!"
    ),
    SteeringPairedExample(
        negative="Please consider attending our upcoming meeting.",
        positive="Come hang out at our meetup if you're free!"
    ),
    SteeringPairedExample(
        negative="We should proceed with caution regarding this decision.",
        positive="Let's not rush into this, just saying."
    )
]

# Create steering concept
steering_request = SteeringConceptCreationRequest(examples=examples)
steering_response = client.create_steering_concept(steering_request)

print(f"Created steering concept: {steering_response.id}")

# Use steering concept in completion
completion_request = CompletionRequest(
    prompt=Prompt.from_text("Write a professional email about project delays"),
    maximum_tokens=150,
    temperature=0.7,
    steering_concepts=[steering_response.id]  # Apply casual style steering
)

response = client.complete(completion_request, model="luminous-extended")
print("Steered completion:")
print(response.completions[0].completion)

# Create steering for technical explanation style
tech_examples = [
    SteeringPairedExample(
        negative="Machine learning is when computers learn things automatically.",
        positive="Machine learning algorithms optimize parameters through iterative training on labeled datasets to minimize loss functions."
    ),
    SteeringPairedExample(
        negative="AI helps solve problems by being smart.",
        positive="Artificial intelligence systems utilize pattern recognition, statistical inference, and computational optimization to process complex data."
    ),
    SteeringPairedExample(
        negative="The program runs faster now.",
        positive="The optimized algorithm demonstrates improved computational complexity from O(n²) to O(n log n)."
    )
]

tech_steering_request = SteeringConceptCreationRequest(examples=tech_examples)
tech_steering_response = client.create_steering_concept(tech_steering_request)

# Apply technical steering
tech_completion = CompletionRequest(
    prompt=Prompt.from_text("Explain how neural networks work"),
    maximum_tokens=200,
    temperature=0.3,
    steering_concepts=[tech_steering_response.id]
)

tech_response = client.complete(tech_completion, model="luminous-extended")
print("Technical explanation:")
print(tech_response.completions[0].completion)

# Combine multiple steering concepts
mixed_completion = CompletionRequest(
    prompt=Prompt.from_text("Describe the benefits of cloud computing for small businesses"),
    maximum_tokens=180,
    steering_concepts=[
        steering_response.id,      # Casual tone
        tech_steering_response.id  # Technical depth
    ]
)

mixed_response = client.complete(mixed_completion, model="luminous-extended")
print("Multi-steered completion (casual + technical):")
print(mixed_response.completions[0].completion)

# Async steering concept creation
import asyncio

async def create_async_steering():
    async with AsyncClient(token="your-api-token") as client:
        examples = [
            SteeringPairedExample(
                negative="This is a good solution to the problem.",
                positive="This solution elegantly addresses the core requirements while maintaining scalability."
            )
        ]
        
        request = SteeringConceptCreationRequest(examples=examples)
        response = await client.create_steering_concept(request)
        return response.id

concept_id = asyncio.run(create_async_steering())
print(f"Async created concept: {concept_id}")

Install with Tessl CLI

npx tessl i tessl/pypi-aleph-alpha-client

docs

chat-interface.md

client-management.md

document-prompt-template.md

embeddings.md

evaluation.md

explanations.md

index.md

prompt-construction.md

steering.md

structured-output.md

text-completion.md

tokenization.md

translation.md

utilities.md

tile.json