Python SDK

Drop-in proxy for OpenAI. Save 40-60% on every call.

Installation

pip install agentready-sdk openai

Method 1: Drop-in Proxy (Recommended)

Just swap your base_url. Zero code changes to your existing OpenAI calls:

from openai import OpenAI

client = OpenAI(
    base_url="https://agentready.cloud/v1",   # ← only change
    api_key="ak_...",                          # your AgentReady key
    default_headers={
        "X-Upstream-API-Key": "sk-...",        # your OpenAI key
    },
)

# Everything works exactly like before — but 40-60% cheaper
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": long_prompt}]
)
# → Prompt compressed automatically, saving ~50% tokens

One-liner Helper

import agentready

client = agentready.openai("ak_...", upstream_key="sk-...")
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello!"}],
)

# Async client
client = agentready.create_client("ak_...", upstream_key="sk-...", async_client=True)
response = await client.chat.completions.create(...)

Method 2: patch_openai()

Monkey-patch existing code with one function call:

from agentready import patch_openai
patch_openai(api_key="ak_...")

# All existing OpenAI code now compresses automatically
from openai import OpenAI
client = OpenAI()
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": long_prompt}]
)

Method 3: Manual Compression

import agentready
agentready.api_key = "ak_..."

result = agentready.compress(
    "Your long prompt here...",
    level="medium",        # light, medium, aggressive
    preserve_code=True,
    target_model="gpt-4",
)

print(result.text)              # compressed output
print(result.tokens_saved)      # 1,247
print(result.reduction_percent) # 52.3
print(result.savings_usd)       # 0.0374

LangChain Integration

from agentready.integrations.langchain import TokenCutCallbackHandler
from langchain_openai import ChatOpenAI

handler = TokenCutCallbackHandler(api_key="ak_...")
llm = ChatOpenAI(model="gpt-4o", callbacks=[handler])

# All prompts compressed before sending
response = llm.invoke("Your long context here...")

LlamaIndex Integration

from agentready.integrations.llamaindex import TokenCutPostprocessor

postprocessor = TokenCutPostprocessor(api_key="ak_...")
query_engine = index.as_query_engine(
    node_postprocessors=[postprocessor]
)
response = query_engine.query("What is the system architecture?")

CrewAI Integration

from agentready.integrations.crewai import create_crewai_llm
from crewai import Agent, Task, Crew

llm = create_crewai_llm(
    agentready_key="ak_...",
    upstream_key="sk-...",
    model="gpt-4o",
)

agent = Agent(
    role="Researcher",
    goal="Research AI trends",
    backstory="Expert AI researcher.",
    llm=llm,
)

task = Task(description="Research and summarize AI trends.", agent=agent)
crew = Crew(agents=[agent], tasks=[task])
result = crew.kickoff()

Streaming

Full streaming support — works exactly like OpenAI:

client = agentready.openai("ak_...", upstream_key="sk-...")

stream = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": long_prompt}],
    stream=True,
)

for chunk in stream:
    print(chunk.choices[0].delta.content or "", end="")

Pricing

Beta — Free unlimited usage. After beta: pay-per-token, ~60% less than direct API costs.