Categories /

software-architecture

6 posts in category "software-architecture".

Pi Durable ♥ Actors: One Agent, One Owner, Any Worker

I have always thought of agents as actors with, well, agency. Pi Durable's one-owner rule makes that almost literal: give each agent its own harness and it is an actor. I spread scripted agents over two worker nodes, killed one, and watched where the work went.

• 11 mins

DeepSeek Harness on a Claude Subscription in 120 Lines

DeepSeek Harness has no way to use a Claude subscription, so I wrote a plugin for it. Its model seam is stateless like Pi's, which meant I could borrow pi-claude-code for the hard part and dsh's own pi-ai adapter for the rest.

• 10 mins

Two Harnesses Walk into a Claude Subscription

OpenCode and Pi (via pi-claude-code) can both use a local Claude Code login. OpenCode keeps a Claude Code session open across tool calls, while Pi runs a fresh worker for each step and replays the transcript.

• 15 mins

Code Mode That Survives a Crash: Durable Scripts in Pi Durable

Code mode lets the model write a script that calls tools. Pi Durable makes that script resume after a restart without rerunning finished calls. What code mode is, how durable replay works, and when to reach for it.

• 15 mins

Pi Durable: A Look at the Future of Agent Harnesses

Pi Durable's building blocks: persisted tasks, conversations, application state, and agents whose work can survive the process that started it.

• 10 mins

Jev at the Branches: The State Machine Is the Agent

A state-machine architecture in which deterministic code owns the plan and the facts, while Jev supplies bounded judgment only at decision branches.

• 12 mins