NexusflowAI Python API library

Welcome to the NexusflowAI API by Nexusflow.ai!

pip install nexusflowai

This package is based on and extends from the OpenAI Python Library. Cheers to the OpenAI team for an amazing API library and SDK!

Usage

Completions

from nexusflowai import NexusflowAI


nf = NexusflowAI(api_key="<api key>")


response = nf.completions.create(
    model="nexus-tool-use-20240816",
    prompt="""Function:
def get_weather(city_name: str):
\"\"\"
\"\"\"


User Query: i am in berkeley.<human_end>Call:""",
    stop=["<bot_end>"],
    max_tokens=10,
)
print(response)

ChatCompletions with Tools

from nexusflowai import NexusflowAI


nf = NexusflowAI(api_key="<api key>")


response = nf.chat.completions.create(
    model="nexus-tool-use-20240816",
    messages=[
        {
            "role": "user",
            "content": "i am in berkeley.",
        },
    ],
    tools=[
        {
            "type": "function",
            "function": {
                "name": "get_weather",
                "description": "",
                "parameters": {
                    "type": "object",
                    "properties": {
                        "city_name": {
                            "type": "string",
                            "description": "",
                        },
                    },
                    "required": ["city_name"],
                    "additionalProperties": False,
                }
            }
        }
    ],
)
print(response)

Multiturn ChatCompletions with Tools + Planning

For Multiturn Chat Completions with Planning, you must use the "nexusflowai_extras" message parameter returned with the chat completion in the previous response.

To do so, simply append the message output of the previous chat completion call to the conversation history.

Example:

from nexusflowai import NexusflowAI


nf = NexusflowAI(api_key="<api key>")

tools_list = [
    {
        "type": "function",
        "function": {
            "name": "get_weather",
            "description": "returns True if weather is nice, False otherwise",
            "parameters": {
                "type": "object",
                "properties": {},
                "required": [],
                "additionalProperties": False,
            }
        }
    },
    {
        "type": "function",
        "function": {
            "name": "call_taxi",
            "description": "",
            "parameters": {
                "type": "object",
                "properties": {},
                "required": [],
                "additionalProperties": False,
            }
        }
    }
]

# Setup Initial Message
messages_list = [
    {
        "role": "user",
        "content": "Get the weather, and then call a taxi.",
    },
]

response = nf.chat.completions.create(
    model="nexus-tool-use-20240816",
    messages=messages_list,
    tools=tools_list
)

print(response.model_dump_json(indent=4))
"""
Output contains `nexusflowai_extras` field in the chat.completion.choices.messages parameter.

output:
{
    "id": "44d060af-ae7c-4166-a3ab-2937231bb126",
    "choices": [
        {
            "finish_reason": "tool_calls",
            "index": 0,
            "message": {
                "content": null,
                "refusal": null,
                "role": "assistant",
                "tool_calls": [
                    {
                        "id": "call_de128c8884e24a0aba92ae88ea40852b",
                        "function": {
                            "arguments": "{}",
                            "name": "get_weather"
                        },
                        "type": "function",
                        "execution_result": null
                    },
                    {
                        "id": "call_e9cbc49b5bdd477da7c2db3deb0d7709",
                        "function": {
                            "arguments": "{}",
                            "name": "call_taxi"
                        },
                        "type": "function",
                        "execution_result": null
                    }
                ],
                "parsed": null,
                "nexusflowai_extras": "{\"original_plan\": \"get_weather(); call_taxi()\"}"
            }
        }
    ],
    "created": 1733280031,
    "model": "nexus-tool-use-20240816",
    "object": "chat.completion",
    "system_fingerprint": null,
    "usage": {
        "completion_tokens": 8,
        "prompt_tokens": 54,
        "total_tokens": 62,
        "latency": 0.35005688667297363,
        "time_to_first_token": null,
        "output_tokens_per_sec": null
    },
    "hints": null
}
"""

# Add previous assistant message with the `nexusflowai_extras` field to messages
messages_list.append(
    response.choices[0].message
)

response = nf.chat.completions.create(
    model="nexus-tool-use-20240816",
    messages=messages_list,
    tools=tools_list,
)

print(response.model_dump_json(indent=4))
"""
output:
{
    "id": "2bc46052-9951-475d-b0d6-6471207bdc90",
    "choices": [
        {
            "finish_reason": "tool_calls",
            "index": 0,
            "message": {
                "content": null,
                "refusal": null,
                "role": "assistant",
                "tool_calls": [
                    {
                        "id": "call_fc412fc4f399449a88a5cce69d0635ab",
                        "function": {
                            "arguments": "{}",
                            "name": "call_taxi"
                        },
                        "type": "function",
                        "execution_result": null
                    }
                ],
                "parsed": null,
                "nexusflowai_extras": "{\"original_plan\": \"call_taxi()\"}"
            }
        }
    ],
    "created": 1733279748,
    "model": "nexus-tool-use-20240816",
    "object": "chat.completion",
    "system_fingerprint": null,
    "usage": {
        "completion_tokens": 5,
        "prompt_tokens": 61,
        "total_tokens": 66,
        "latency": 0.24212288856506348,
        "time_to_first_token": null,
        "output_tokens_per_sec": null
    },
    "hints": null
}
"""

ChatCompletions with Structured Outputs

from typing import List, Dict, Tuple

from pydantic import BaseModel, Field

from nexusflowai import NexusflowAI


nf = NexusflowAI(api_key="<api key>")


class GasDistributionNetwork(BaseModel):
    networkID: str = Field(
        ...,
        description="The identifier for the gas distribution network.",
        title="Network ID",
    )
    pipelineValues: Dict[str, Tuple[int, int]] = Field(
        description="The mapping with key pipeline_1, pipeline_2, etc ... to tuple of (total length in kilometers, maximum amount of gas that can be distributed in cubic meters).",
        title="Pipeline Values",
    )
    maintenanceSchedules: List[str] = Field(
        ...,
        description="The schedule detailing when maintenance activities are to be performed.",
        title="Maintenance Schedule",
    )


response = nf.chat.completions.create(
    model="nexus-tool-use-20240816",
    messages=[
        {
            "role": "user",
            "content": """I am currently working on a project that involves mapping out a gas distribution network for a new residential area. The network is quite extensive and includes several pipelines that distribute natural gas to various sectors of the community. I need to create a JSON object that captures the essential details of this network. The information I have includes a unique identifier for the network, which is 'GDN-4521'. The total length of the pipeline_1 is 275 kilometers with a capacity 500,000 cubic meters. Pipeline 2 is 17 kilometers long and has a capacity of 12,000 cubic meters. Additionally, there is a detailed maintenance schedule, which includes quarterly inspections in January, April, July, and October.""",
        },
    ],
    response_format=GasDistributionNetwork,
)
print(response.raw_prompt)
print(response.choices[0].message.parsed)