AI Completions

Experience Fusionbase AI: Seamlessly Interact with Every Entity in the Acclaimed Fusionbase Data Hub Through Advanced Chat-Based Contextual Access.

The Fusionbase AI API is not yet publicly accessible for all accounts. If you are interested in getting early access, drop us a message via [email protected]

The /api/v2/chat/completions endpoint of the Fusionbase API is designed to generate chat completions based on a given context and a series of messages. This endpoint is particularly useful for obtaining context-aware responses in a conversational format.

Request Headers

• Content-Type: application/json

• X-API-KEY: Your Fusionbase API key.

Request Body

• model (required): String. Specifies the model to be used for generating completions. For example, "fusion-one".

• linked_context (optional): Array of Objects. Each object represents a context entity linked to the chat.

  • type: String. The type of the linked context entity (e.g., "ORGANIZATION", "PERSON").

  • id: String. A unique identifier for the context entity.

• messages (required): Array of Objects. Each object represents a message in the conversation.

  • role: String. The role of the message sender ("system" or "user").

  • content: String. The content of the message.

Request Samples

Python

import requests

url = "https://api.fusionbase.com/api/v2/chat/completions"
headers = {
    "Content-Type": "application/json",
    "X-API-KEY": "<YOUR API KEY>"
}
data = {
    "model": "fusion-one",
    "linked_context": [
        {"type": "ORGANIZATION", "id": "35b263b5408bc85899b5a4fab6be5d75"},
        {"type": "PERSON", "id": "bc6a357dc563ad838e213ff06e0c1c91"}
    ],
    "messages": [
        {"role": "system", "content": "You are an insurance and underwriting expert."},
        {"role": "user", "content": "What are potential insurance risks of the given company?"}
    ]
}

response = requests.post(url, json=data, headers=headers)
print(response.json())

NodeJS

Java

Go

cURL

Response Structure

• id: A unique 32-character alphanumeric hash identifying the completion request.

• object: String, indicating the type of the object, here "chat.completion".

• created: Unix timestamp indicating the creation time of the response.

• model: String, indicating the model used, revised to match the request's model.

• usage: Object detailing the token usage.

  • prompt_tokens: Number of tokens used in the prompt.

  • completion_tokens: Number of tokens used in the completion.

  • total_tokens: Total number of tokens used.

• choices: Array of objects, each representing a generated response.

  • message: Object containing the role and content of the response.

  • logprobs: Null or containing log probabilities for the response.

  • finish_reason: Reason for the completion's end, e.g., "stop".

  • index: The index of the choice.

Response Sample

{
    "id": "1234abcd5678efgh9012ijkl3456mnop",
    "object": "chat.completion",
    "created": 1677858242,
    "model": "fusion-one",
    "usage": {
        "prompt_tokens": 13,
        "completion_tokens": 7,
        "total_tokens": 20
    },
    "choices": [
        {
            "message": {
                "role": "assistant",
                "content": "The potential insurance risks for the given company include..."
            },
            "logprobs": null,
            "finish_reason": "stop",
            "index": 0
        }
    ]
}

Error Handling

The API uses standard HTTP response codes. Errors are indicated with 4xx (client errors) and 5xx (server errors), each accompanied by a message detailing the issue.

Notes

• Keep the API key confidential.

• Context and message relevance is crucial for accurate results.