Quickstart

Deploy the ICE engine and execute your first context-augmented query.

0. Installation

Install the ICE binary for your preferred runtime.

Python

AMD64 (x86_64)

pip install https://dl.dopove.com/ice/v2.7.756/python/ice_community-2.7.756-cp312-cp312-manylinux2014_x86_64.whl

ARM64

pip install https://dl.dopove.com/ice/v2.7.756/python/ice_community-2.7.756-cp312-cp312-manylinux2014_aarch64.whl

Windows (Native)

pip install https://dl.dopove.com/ice/v2.7.756/python/ice_community-2.7.756-cp312-cp312-win_amd64.whl

Node.js

npm install https://dl.dopove.com/ice/v2.7.756/node/ice-node-community-2.7.756.tgz

1. Environment Configuration

Define the required environment variables for the ICE singleton.

Linux / macOS

export ICE_PORT=8000
export DATABASE_URL="postgresql://user:pass@localhost:5432/ice_db"
export REDIS_URL="redis://localhost:6379"
export ICE_LICENSE_JWT="your_license_here"

Windows (PowerShell)

$env:ICE_PORT="8000"
$env:DATABASE_URL="postgresql://user:pass@localhost:5432/ice_db"
$env:REDIS_URL="redis://localhost:6379"
$env:ICE_LICENSE_JWT="your_license_here"

2. Engine Initialization

Initialize the ICE kernel within your application.

Python

from ice.sdk import ICEClient

# Initialize singleton
ice = ICEClient(api_url="http://localhost:8000")

# Verify connectivity
if ice.ping():
    print("ICE Kernel Active")

Node.js

import { ICEClient } from '@dopove/ice-node';

const ice = new ICEClient({
  baseUrl: 'http://localhost:8000'
});

await ice.ping();

3. Data Ingestion

Ingest documents or raw text into the Semantic Ledger.

# Ingest document into specific session
ice.ingest(
    session_id="proj_001",
    file_path="./technical_specs.pdf",
    metadata={"category": "documentation"}
)

4. Contextual Query

Execute a query. ICE automatically retrieves relevant historical context and injects it into the prompt.

response = ice.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Summarize the memory limits."}],
    x_session_id="proj_001",   # Required — scopes memory to this session
    x_user_id="alice"           # Required — scopes data isolation to this user
)

print(response.choices[0].message.content)

---

Next Steps

• API Reference — full endpoint and header documentation
• Configuration — all environment variables
• Storage Architecture — how memory tiers and cold storage work
• Agentic Patterns — tool-call persistence and multi-step agent setup