aops 0.1.1

LangChain integration library for AOps

aops

LangChain integration library for AOps — load agent prompt configurations from the AOps backend and use them directly in LangChain chains.

Requirements

• Python 3.12+
• AOps backend running (self-hosted)
• API key issued from the AOps UI (Agent detail page → New API Key)

Installation

pip install aops

Quick Start

from dotenv import load_dotenv
load_dotenv()

import aops
aops.init(api_key="aops_...")

from aops.langchain import pull
from langchain_core.prompts import ChatPromptTemplate, HumanMessagePromptTemplate
from langchain_core.output_parsers import StrOutputParser
from langchain_openai import ChatOpenAI

system_prompt = pull("my-agent/my-chain")

chain = (
    ChatPromptTemplate.from_messages([
        system_prompt,
        HumanMessagePromptTemplate.from_template("{user_input}"),
    ])
    | ChatOpenAI(model="gpt-4o-mini")
    | StrOutputParser()
)

result = chain.invoke({"user_input": "Hello!"})

Configuration

API Key

AOps API keys embed the server host, so no separate base_url is needed.

aops_{base64(host)}_{token}

Issue a key from the AOps UI: Agent detail page → API Keys → New API Key

aops.init()

Call init() once before using any aops functions:

import aops

aops.init(api_key="aops_...")

Or use environment variables — init() is optional when env vars are set:

# .env
AGENTOPS_API_KEY=aops_...
OPENAI_API_KEY=sk-...

Environment Variable	Default	Description
AGENTOPS_API_KEY	—	API key (host is parsed from it)
AGENTOPS_BASE_URL	parsed from key	Override the host embedded in the key
AGENTOPS_API_PREFIX	/api/v1	API path prefix
AGENTOPS_CACHE_TTL	300	Prompt cache TTL in seconds (0 = no cache)

API

pull(ref, *, version=None)

Fetch a chain from AOps and return a SystemMessagePromptTemplate.

from aops.langchain import pull

prompt = pull("my-agent/my-chain")          # latest
prompt = pull("my-agent/my-chain", version=2)  # pinned version

The chain's persona and content are merged into a single system message:

# Persona
{persona}

# Content
{content}

content may contain LangChain template variables (e.g. {language}). persona is treated as a fixed string — its braces are escaped automatically.

---

@chain_prompt(agent_name, chain_name, *, version=None)

Decorator that fetches the prompt and injects it as the first argument.

Function decorator

from aops.langchain import chain_prompt
from langchain_core.prompts import SystemMessagePromptTemplate

@chain_prompt("my-agent", "my-chain")
def answer(prompt: SystemMessagePromptTemplate, user_input: str) -> str:
    return (
        ChatPromptTemplate.from_messages([
            prompt,
            HumanMessagePromptTemplate.from_template("{user_input}"),
        ])
        | ChatOpenAI(model="gpt-4o-mini")
        | StrOutputParser()
    ).invoke({"user_input": user_input})

result = answer(user_input="What is AOps?")

Class decorator

The prompt is injected into __init__ as the first argument after self. Build the chain once and reuse it:

@chain_prompt("my-agent", "my-chain")
class MyAgent:
    def __init__(self, prompt: SystemMessagePromptTemplate) -> None:
        self.chain = (
            ChatPromptTemplate.from_messages([
                prompt,
                HumanMessagePromptTemplate.from_template("{user_input}"),
            ])
            | ChatOpenAI(model="gpt-4o-mini")
            | StrOutputParser()
        )

    def run(self, user_input: str) -> str:
        return self.chain.invoke({"user_input": user_input})

agent = MyAgent()
result = agent.run(user_input="Hello!")

License

MIT