AI Assistants

Coding assistants know TanStack Query far better than Qora. Left alone, they write TanStack code with Dart syntax: queryFn instead of fetcher, cacheTime instead of gcTime, state.when(...), which does not exist. Two files fix most of it: a short rules file the assistant loads in every session, and the documentation in a format it can read.

1. Add the Qora rules to your project

qora-rules.md is a one-page summary written for assistants. It contains the setup, how to render states, the offline behaviour, mutations and invalidation, and a table of the mistakes assistants make most often, each with the correct Qora form.

Download it to the root of your project:

curl -o qora-rules.md https://qora.meragix.dev/qora-rules.md

Then reference it from the instruction file your assistant reads:

For Cursor, the file starts with:

---
description: Qora server-state library rules
globs: "**/*.dart"
alwaysApply: false
---

Re-download the file when you upgrade Qora: it states the version it describes.

2. Point the assistant at the documentation

The documentation is published in formats meant for language models:

For a focused task, pass the one or two relevant pages rather than the full file: the assistant follows a short, precise context better than a long one.

3. Check what it wrote

• Run dart analyze. An invented method or parameter is a compile error, which is the quickest way to catch a TanStack habit.
• Look for a switch on QoraState. Code that checks state.isLoading before reading data usually drops the previousData kept during refetches and errors.
• Test offline. Code generated from other libraries tends to expect a Failure or Initial when the network is off. Qora pauses instead: Loading with FetchStatus.paused.

When in doubt, compare with the Quick Start, which runs as is, or ask the assistant to read Coming from TanStack Query, which maps every option.