i just recorded a full tutorial on pi durable 1:48 Checkpoints + crash recovery 8:38 Harness vs. app 13:46 App architecture + storage 19:30 Six setup steps 23:11 Local Qwen + llama.cpp demo + restart recovery

@nicopreme Alejandro is the 🐐

@nicopreme you are

The harness runs conversations, model calls, tools, and recovery. Your application supplies the interface, user management, and host supervision. Pi Durable shares Pi AI model integrations.

The demo runs Qwen3.5-0.8B with llama.cpp, reads a file using tools, then resumes interrupted work from SQLite. Durability doesn't make arbitrary side effects safe to repeat. Tool replay safety matters.

Full tutorial on YouTube:

i know it :D i was waiting for your video and i got a feeling that you will make an explaining like you did with pi and Tau 🤗 now gonna watch and learn some staff thank you so much for this <3

it's a pleasure!

<3

any chance you'd make a tutorial building a hermes-like agent using pi-durable?

i'm working towards it!

thank you!

How's the harness abstraction handle state during a crash?

crash recovery that resumes from SQLite is the part. durable AI just means a crash doesn't lose the thread.

Durable execution is the missing primitive for always-on agents — checkpoints and crash recovery turn a demo into something you can actually trust in production. The local Qwen + llama.cpp angle is the interesting part: durable doesn't have to mean cloud-dependent.

Checkpointing and SQLite recovery look solid, harness abstraction keeps state consistent across crashes

@nicopreme Great work, thanks!