Examples

Python + OpenAI

Add persistent memory to an OpenAI-powered Python agent in two lines.

1. Install

pip install "coorda[openai]"

2. Wrap your OpenAI client

One import, one wrap line. Every completion call now reads and writes through Coorda.

from openai import OpenAI
from coorda import Coorda

coorda = Coorda(api_key="ck_live_...")
client = coorda.wrap(OpenAI(), agent_id="triage-bot")

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Customer asking about billing"}],
    coorda={"entity_id": "customer_123"},
)

What happens on every call

Before the call, Coorda searches for prior context tagged to customer_123 and prepends it as a system message. After the call, the input/output pair is stored in a background thread so the next agent (even a different one) can retrieve it.

Before and after

Without the SDK, the same integration requires manual HTTP calls around every LLM invocation. With wrap(), Coorda becomes invisible.

Without the SDK

import requests
from openai import OpenAI

COORDA = "https://coorda-app.vercel.app"
H = {"X-API-Key": "ck_live_...", "Content-Type": "application/json"}

def handle(entity_id, user_msg):
    prior = requests.get(
        f"{COORDA}/api/context/inject",
        params={"query": user_msg, "entityId": entity_id},
        headers=H,
    ).json()
    messages = [{"role": "user", "content": user_msg}]
    if prior.get("hasContext"):
        messages = [{"role": "system", "content": prior["context"]}] + messages
    resp = OpenAI().chat.completions.create(
        model="gpt-4o", messages=messages,
    )
    output = resp.choices[0].message.content
    requests.post(f"{COORDA}/api/context", headers=H, json={
        "taskId": "t1", "entityId": entity_id, "agentId": "triage",
        "input": user_msg, "output": output,
    })
    return output

With wrap()

from openai import OpenAI
from coorda import Coorda

coorda = Coorda(api_key="ck_live_...")
client = coorda.wrap(OpenAI(), agent_id="triage")

def handle(entity_id, user_msg):
    resp = client.chat.completions.create(
        model="gpt-4o",
        messages=[{"role": "user", "content": user_msg}],
        coorda={"entity_id": entity_id},
    )
    return resp.choices[0].message.content

Per-call options

client.chat.completions.create(
    model="gpt-4o",
    messages=[...],
    coorda={
        "entity_id": "customer_123",
        "skip_inject": True,        # don't pull prior context for this call
        "skip_store": True,         # don't save the output
        "min_similarity": 0.5,      # stricter relevance threshold
        "limit": 3,                 # fewer pieces of prior context
        "metadata": {"priority": "high"},
    },
)

# Pass coorda=False to fully bypass Coorda for one call.
client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "internal-only query"}],
    coorda=False,
)

Safe by default

If Coorda is unreachable or returns an error, your agent still runs. All Coorda errors are swallowed — the underlying OpenAI call always executes, and storage happens asynchronously.