Guides¶

Practical, hands-on guides for using CRP in real-world scenarios. Each guide focuses on a concrete outcome - running the demo, building multi-turn retrieval sessions, or streaming tokens to a UI.

Deployment status

All guides work with the self-hosted SDK today. Managed SaaS endpoints for Gateway and Comply are on the waitlist at *.crprotocol.io.

Getting Started¶

Guide	What You'll Learn
Demo App	Run the interactive demo to see CRP's safety, retrieval, and continuation features in action
Streaming	Stream tokens and continuation-window events with `client.stream()`
Multi-Turn	Build knowledge across multiple conversations with `client.ingest()` and `client.ask()`

Data & Persistence¶

Guide	What You'll Learn
Ingestion	Feed external data into CRP without an LLM call
Session Persistence	Resume sessions across restarts with CKF cold storage

Deployment¶

Guide	What You'll Learn
HTTP Sidecar	Run CRP as a REST API for cross-process integration

Prerequisites¶

All guides assume you have CRP installed:

pip install crprotocol

And a provider configured (see Providers or Local Models & LM Studio).

The SDK in one page¶

import crp

client = crp.SDKClient()            # auto-detects provider

# Single-turn generation
r = client.complete("Explain CRP.")
print(r.text, r.crp.risk, r.crp.grounded)

# Retrieval over ingested documents
client.ingest("./docs/")
a = client.ask("What does CRP govern?", depth="standard")
print(a.text, a.sources, a.crp.risk)

# Session inspection
s = client.session()
print(s.id, s.status(), s.fact_count, s.window_count)