Aleph Alpha Python Client

11.5.1 verified Thu Apr 16 auth: no python

The `aleph-alpha-client` is the official Python client for interacting with Aleph Alpha's API endpoints. It provides synchronous and asynchronous interfaces to access various AI capabilities, including completion, embedding, evaluation, and tool calling with large language and multimodal models. The library is actively maintained, with frequent releases, and is currently at version 11.5.1.

pip install aleph-alpha-client

Common errors

error ValueError: ALEPH_ALPHA_API_TOKEN environment variable not set or is default. ↓

cause The API client requires an authentication token, which is typically read from the `ALEPH_ALPHA_API_TOKEN` environment variable, but it's either missing or set to a placeholder.

fix

Set the ALEPH_ALPHA_API_TOKEN environment variable to your actual Aleph Alpha API key. For example: export ALEPH_ALPHA_API_TOKEN="your_secret_token" in your shell, or pass it directly to the Client constructor.

error aleph_alpha_client.aleph_alpha_client.errors.ModelError: Could not find model with name 'your_model_name' on hosting 'cloud' ↓

cause The specified model name is incorrect, mistyped, or not available on the default ('cloud') or specified hosting for your account.

fix

Verify the exact name of the model you intend to use. You can use methods like client.available_models() to check for available models and their hostings.

error aleph_alpha_client.aleph_alpha_client.errors.ValidationError: prompt is too long: X tokens > Y maximum ↓

cause The combined length of your prompt exceeds the maximum token limit allowed by the chosen Aleph Alpha model's context window.

fix

Reduce the length of your input prompt. Consider summarizing parts of the prompt, breaking it into multiple requests, or choosing a model with a larger context window if available.

Warnings

breaking Breaking changes in v3.0.0 removed `AlephAlphaClient` and `AlephAlphaModel`. The class `ImagePrompt` was also removed. ↓

fix Migrate to `Client` or `AsyncClient` for API interaction. Use `Image` instead of `ImagePrompt` for image-based prompts.

breaking The parameter order for `client.semantic_embed` changed, swapping `hosting` and `request`. ↓

fix Ensure you are using keyword arguments (e.g., `client.semantic_embed(request=my_request, hosting='cloud')`) or update your positional argument order.

gotcha The `maximum_tokens` parameter limits the *generated output* length, not the total context window (input + output). Setting it too low can result in truncated responses. ↓

fix Adjust `maximum_tokens` to allow for the desired response length. Be aware of the model's total context limit when combining a long prompt with `maximum_tokens`. If the prompt alone is too long, it will result in an API error.

Imports

Client
```
from aleph_alpha_client import Client
```
Main synchronous client class for API interaction.
AsyncClient
```
from aleph_alpha_client import AsyncClient
```
Main asynchronous client class for API interaction.
CompletionRequest
```
from aleph_alpha_client import CompletionRequest
```
Used to specify parameters for text completion requests.
Prompt
```
from aleph_alpha_client import Prompt
```
Base class for constructing prompts, including from text or multimodal items.
Image
wrong
```
from aleph_alpha_client import ImagePrompt
```
correct
```
from aleph_alpha_client import Image
```
As of v3.0.0, `ImagePrompt` was removed; use `Image` for image prompt items instead.
AlephAlphaClient
wrong
```
from aleph_alpha_client import AlephAlphaClient
```
correct
```
from aleph_alpha_client import Client
```
`AlephAlphaClient` was deprecated and removed in v3.0.0; use `Client` or `AsyncClient` instead.

Quickstart

This quickstart demonstrates how to initialize the synchronous client, construct a basic text completion request, and print the model's response. It expects the Aleph Alpha API token to be set as an environment variable `ALEPH_ALPHA_API_TOKEN`.

import os
from aleph_alpha_client import Client, CompletionRequest, Prompt

# Ensure ALEPH_ALPHA_API_TOKEN is set in your environment variables
api_token = os.environ.get('ALEPH_ALPHA_API_TOKEN', 'YOUR_API_TOKEN')
if not api_token or api_token == 'YOUR_API_TOKEN':
    raise ValueError("ALEPH_ALPHA_API_TOKEN environment variable not set or is default. Please set it to your actual API token.")

# Instantiate the synchronous client
client = Client(token=api_token)

# Define the prompt and completion request
request = CompletionRequest(
    prompt=Prompt.from_text("Provide a short description of AI:"),
    maximum_tokens=64,
    model="luminous-base" # Or another available model, e.g., "pharia-1-llm-7b-control"
)

try:
    # Send the completion request
    response = client.complete(request=request, model=request.model)
    print(response.completions[0].completion)
except Exception as e:
    print(f"An error occurred: {e}")