Skip to main content

Async SDK Usage

Not yet implemented

Async SDK methods are not yet implemented. There are currently no acomplete, aask, aingest, aask_stream, or async session methods on crp.SDKClient.

The sections below describe the planned async surface once it is wired.

---

Planned async client

import asyncio
import crp

async def main():
    client = crp.SDKClient()

    r = await client.acomplete("Summarise the EU AI Act")
    print(r.text)
    print(r.crp.risk)

asyncio.run(main())

---

Planned async ask() and streaming

async def ask_question():
    client = crp.SDKClient()
    a = await client.aask("Write a deployment guide")
    print(a.text)
    print(a.sources)

async def stream_question():
    client = crp.SDKClient()
    async for chunk in client.aask_stream("Write a deployment guide"):
        print(chunk.text, end="")

---

Planned async ingestion

async def ingest_docs():
    client = crp.SDKClient()
    await client.aingest("./docs/")
    print(client.storage.fact_count())

---

Planned async sessions

async def chat():
    client = crp.SDKClient()
    # Planned: client.asession() or session async methods
    r1 = await client.aask("What is the EU AI Act?")
    r2 = await client.aask("Which articles apply to high-risk systems?")

---

Concurrency tips

• Create one crp.SDKClient per process and share it across tasks. The client holds one session handle internally.
• For multi-tenant services, instantiate separate clients per tenant or wait for explicit session_id= support.