GenAI service API

GenAI service API#

The class below is a mixin class which encapsulates the functionality of the GenAI service API. It is not a comprehensive list of all the methods available in the GenAI service API, as the methods may vary depending on a specific deployment configuration.

The source of truth for the GenAI service API should be the API documentation available under the /docs endpoint of the GenAI service. In a traditional deployment, the URL would be https://<squirro-cluster>/service/genai/docs.

All the methods of these classes are made available in the SquirroClient class.

Example Usage#

Attention

To interact with the genai service via SquirroClient, ensure the service is accessible from the host where the SquirroClient is running. This is usually done by adjusting the Nginx configuration to allow external access to the /service/genai endpoint.

The following example demonstrates how to interact with the GenAI service API to get the answer to a simple prompt using the SquirroClient.

from squirro_client import SquirroClient

# Set these variables accordingly:
REFRESH_TOKEN = "v4.public.xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
CLUSTER="https://my-squirro-cluster.squirro.cloud" # Ensure `CLUSTER` **does not** have a trailing slash
GENAI_API_URL = CLUSTER + "/service/genai" # `GENAI_API_URL` may be different depending on where GenAI is deployed
PROJECT_ID = "your-project-id"
OPENAI_API_KEY = "sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"

client = SquirroClient(None, None, cluster=CLUSTER, genai_api_url=GENAI_API_URL)
client.authenticate(refresh_token=REFRESH_TOKEN)

json = {
    "instruction": "Tell me about France.",
    "runtime_config": {
        "llm_api_key": OPENAI_API_KEY, # may be required if the `genai` service has not been provided with one, **even** if the platform has one
        "squirro_refresh_token":  REFRESH_TOKEN, # currently, we need to provide the Squirro refresh token to allow the GenAI service to access project data. In the future, it'll be inferred from the access token created by the SquirroClient
}

res = client.genai_request(
        "post",
        f"/v0/projects/{PROJECT_ID}/streaming_chat",
        json=json,
    )

for event in res.iter_lines():
    print(event)

For more complex usage, which includes retrieving documents from Squirro to provide context to the LLM, we can use agents:

agents = client.get_agents(PROJECT_ID)

# The "All Data" agent is available in every project by default and attempts to retrieve documents related to the prompt from all data sources connected to the project
agent = next((a for a in agents if a["name"] == "All Data"), None)
agent_id = agent["id"]

json = {
    "instruction": "Who was Albert Einstein?",
    "agent_id": agent_id,
    "runtime_config": {
        "llm_api_key": OPENAI_API_KEY,
        "squirro_refresh_token":  REFRESH_TOKEN,
    }
}

res = client.genai_request(
        "post",
        f"/v0/projects/{PROJECT_ID}/streaming_chat",
        json=json,
    )

for event in res.iter_lines():
    print(event)

Classes#

Any(*args, **kwargs)

Special type indicating an unconstrained type.

GenAIApiMixin()

Contents