reminix-openai 0.0.20

Reminix agents for OpenAI - serve AI agents as REST APIs

reminix-openai

Reminix agents for the OpenAI API. Serve OpenAI models as a REST API with chat, task, and thread agents.

Ready to go live? Deploy to Reminix Cloud for zero-config hosting, or self-host on your own infrastructure.

Installation

pip install reminix-openai

This will also install reminix-runtime as a dependency.

Quick Start

Chat Agent

The chat agent follows the chat type and supports streaming responses.

from openai import AsyncOpenAI
from reminix_openai import OpenAIChatAgent
from reminix_runtime import serve

client = AsyncOpenAI()
agent = OpenAIChatAgent(client, name="my-chatbot", model="gpt-4o")
serve(agents=[agent])

Task Agent

The task agent follows the task type and returns structured output. Streaming is not supported.

from openai import AsyncOpenAI
from reminix_openai import OpenAITaskAgent
from reminix_runtime import serve

summary_schema = {
    "type": "object",
    "properties": {
        "title": {"type": "string"},
        "bullet_points": {"type": "array", "items": {"type": "string"}},
    },
    "required": ["title", "bullet_points"],
}

client = AsyncOpenAI()
agent = OpenAITaskAgent(client, output_schema=summary_schema, name="summarizer", model="gpt-4o")
serve(agents=[agent])

Thread Agent

The thread agent follows the thread type and supports tool use over multiple turns. Streaming is not supported.

from openai import AsyncOpenAI
from reminix_openai import OpenAIThreadAgent
from reminix_runtime import serve

def get_weather(location: str) -> str:
    return f"The weather in {location} is sunny."

client = AsyncOpenAI()
agent = OpenAIThreadAgent(
    client,
    tools=[get_weather],
    name="assistant",
    model="gpt-4o",
    max_turns=10,
)
serve(agents=[agent])

Your agents are now available at:

• POST /agents/{name}/invoke - Execute the agent

API Reference

OpenAIChatAgent(client, *, name, model, description, instructions)

Create an OpenAI chat agent. Follows the chat type and supports streaming.

Parameter	Type	Default	Description
client	AsyncOpenAI	required	An OpenAI async client
name	str	"openai-agent"	Name for the agent (used in URL path)
model	str	"gpt-4o-mini"	Model to use for completions
description	str	"openai chat agent"	Description shown in agent metadata
instructions	str	None	System instructions prepended to messages
tags	list[str]	None	Tags for categorizing/filtering agents
metadata	dict	None	Custom metadata merged into agent info

Returns: OpenAIChatAgent - A Reminix chat agent instance

OpenAITaskAgent(client, *, output_schema, name, model, description, instructions, tags, metadata)

Create an OpenAI task agent. Follows the task type and returns structured output. Streaming is not supported.

Parameter	Type	Default	Description
client	AsyncOpenAI	required	An OpenAI async client
output_schema	dict	required	A JSON Schema dict defining the structured output
name	str	"openai-task-agent"	Name for the agent (used in URL path)
model	str	"gpt-4o-mini"	Model to use for completions
description	str	"openai task agent"	Description shown in agent metadata
instructions	str	None	System instructions prepended to messages
tags	list[str]	None	Tags for categorizing/filtering agents
metadata	dict	None	Custom metadata merged into agent info

Returns: OpenAITaskAgent - A Reminix task agent instance

OpenAIThreadAgent(client, *, tools, name, model, max_turns, description, instructions, tags, metadata)

Create an OpenAI thread agent. Follows the thread type and supports tool use over multiple turns. Streaming is not supported.

Parameter	Type	Default	Description
client	AsyncOpenAI	required	An OpenAI async client
tools	list	required	A list of tool functions the agent can call
name	str	"openai-thread-agent"	Name for the agent (used in URL path)
model	str	"gpt-4o-mini"	Model to use for completions
max_turns	int	10	Maximum number of tool-use turns before stopping
description	str	"openai thread agent"	Description shown in agent metadata
instructions	str	None	System instructions prepended to messages
tags	list[str]	None	Tags for categorizing/filtering agents
metadata	dict	None	Custom metadata merged into agent info

Returns: OpenAIThreadAgent - A Reminix thread agent instance

Example with Custom Configuration

from openai import AsyncOpenAI
from reminix_openai import OpenAIChatAgent
from reminix_runtime import serve

client = AsyncOpenAI(
    api_key="your-api-key",
    base_url="https://your-proxy.com/v1"  # Optional: custom endpoint
)

agent = OpenAIChatAgent(
    client,
    name="gpt4-agent",
    model="gpt-4o"
)

serve(agents=[agent])

Endpoint Input/Output Formats

POST /agents/{name}/invoke

Execute the agent with a prompt or messages.

Request with prompt:

{
  "input": {
    "prompt": "Summarize this text: ..."
  }
}

Request with messages:

{
  "input": {
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "Hello!"}
    ]
  }
}

Response:

{
  "output": "Hello! How can I help you today?"
}

Streaming

For streaming responses, set stream: true in the request (chat agent only):

{
  "input": {
    "prompt": "Tell me a story"
  },
  "stream": true
}

The response will be sent as Server-Sent Events (SSE).

Runtime Documentation

For information about the server, endpoints, request/response formats, and more, see the reminix-runtime package.

Deployment

Ready to go live?

• Deploy to Reminix Cloud - Zero-config cloud hosting
• Self-host - Run on your own infrastructure

Links

• GitHub Repository
• OpenAI Documentation

License

Apache-2.0