CHAPTER 1·MEMORY
ICE
Indexless Context Engine. Turn-granular memory for AI agents.
Our first product
Explore ChapterIntroducing ICE V3
Most agent memory forgets between conversations. Ours doesn't.
Built with systems thinking, in the product and in the company.
Where stateful agents break down.
- Context that grows but does not change how the agent behaves.
- Memory that is gone between conversations.
- Retrieval that returns text that is semantically close, but wrong for the moment.
Dopove is a systems thinking company. We build useful memory for AI agents, and we apply the same thinking to how we run our own marketing.
ICE, the Indexless Context Engine, is turn-granular memory for AI agents that makes the context you already have useful.
How ICE applies systems thinking to agent memory.
Turn-granular storage
Each conversational turn is stored as its own frame, like a frame in a film, so the order of events survives. Fixed-window chunking loses it.
Multi-lane retrieval
Retrieval is organised like the sections of a library: episodic for what happened, semantic for what it means, procedural for how to do it, and working for what the agent is focused on now.
Cross-session persistence
Memory carries across sessions, so each interaction can build on the last, like a feedback loop.
While others chase longer contexts, we ask what makes the information in the window useful. ICE works on the signal-to-noise ratio of the context an agent already has.
Developer FirstOne import.
One import.
Persistent Memory.
ICE wraps standard LLM client libraries to automatically handle embedding, database storage, and context retrieval. No database setup, chunking logic, or index management needed.
Transparent LLM Client Wrapper
Zero-Conf Local Storage
Multi-Platform Native Core C-Extension
main.py
import ice
import asyncio
import os
# 1. Set your signed license key
os.environ["ICE_LICENSE_JWT"] = "your_signed_jwt_here"
async def main():
# 2. Ingest documents into memory
await ice.ingest("s3://data-lake/architecture.pdf")
# 3. Query with automatic memory retrieval
response = await ice.query(
prompt="Analyze this.",
model="gpt-4o"
)
print(response)
asyncio.run(main())
The Cast
Three products, one way of thinking.
GoingMerry is the next product we are building after ICE.
We are experimenting with an autonomous approach to our own marketing, to reduce founder workload. This site reflects what we are learning, including what is still unfinished.