- Qiskit Code Assistant's OpenAI API compatibility is in preview release status and is subject to change.
- If you have feedback or want to contact the developer team, use the Qiskit Slack Workspace channel or the related public GitHub repositories.
Qiskit Code Assistant - OpenAI API compatibility
Qiskit Code Assistant offers compatibility with a subset of the OpenAI API specification, specifically with the completions API endpoints. The goal of this compatibility is to allow third-party packages to connect to Qiskit Code Assistant seamlessly by using well-known AI-related libraries and methods such as OpenAI, LiteLLM, or others.
OpenAI API endpoints supported
| Method | Path | Comment |
|---|---|---|
| GET | /v1/models | List all models |
| GET | /v1/model/{model} | Get model detail |
| POST | /v1/completions | Send prompt to model for completion |
The /v1/completions endpoint fails with a 403 error if the model disclaimer has been accepted. See the following for how to view and accept the model disclaimer.
Additional endpoints (not part of OpenAI schema, provided for convenience) include:
| Method | Path | Comment |
|---|---|---|
| GET | /v1/model/{model}/disclaimer | Get model's disclaimer |
| POST | /v1/model/{model}/disclaimer | Accept model's disclaimer |
| POST | /v1/completions/accept | Accept or reject completion |
To retrieve/view the model disclaimer, make a GET request to the disclaimer endpoint. For example:
curl -X 'GET' \
'https://qiskit-code-assistant.quantum.ibm.com/v1/model/granite-8b-qiskit/disclaimer' \
-H 'accept: application/json' \
-H 'Authorization: Bearer <IBM Quantum API token>'
If you agree with the model disclaimer, to accept it, POST to the disclaimer endpoint providing the disclaimer's ID and whether it is accepted or rejected. For example:
curl -X 'POST' \
'https://qiskit-code-assistant.quantum.ibm.com/v1/model/granite-8b-qiskit/disclaimer' \
-H 'accept: application/json' \
-H 'Authorization: Bearer <IBM Quantum API token>' \
-H 'Content-Type: application/json' \
-d '{
"disclaimer": "<DISCLAIMER_ID>",
"accepted": true
}'
Examples
Use the official OpenAI Python library
The OpenAI Python library provides convenient access to the OpenAI REST API (such as the one provided by Qiskit Code Assistant) from any Python 3.8+ application. See more details in the Installation section of the OpenAI Python API library Readme.
from openai import OpenAI
# Initialize the client with your API key
client = OpenAI(api_key=auth_token_iqp, base_url=assistant_url)
# Make a request to the completions API
try:
response = client.completions.create(
model=model, # or another compatible model
prompt="#Transpile a random circuit using the Qiskit Transpiler Service",
)
# Print the generated text
print(response.choices[0].text)
except Exception as e:
print(f"An error occurred: {e}")
Use LiteLLM
LiteLLM is a convenient Python library to access multiple LLM APIs using the OpenAI format (Bedrock, Huggingface, VertexAI, TogetherAI, Azure, OpenAI, Groq, and so on). See the LiteLLM docs for more details.
from litellm import completion
response = completion(
model=f"text-completion-openai/{model}",
base_url= assistant_url,
messages=[
{"role": "user", "content": "#Transpile a random circuit using the Qiskit Transpiler Service"}
],
api_key=<IBM Quantum API token>
)
completion_response = response.json()
print(completion_response)