Documentation

Tutorials, how-to guides, reference material, and design notes for semantic jq: fuzzy JSON filters, bounded classification, deterministic jq execution, and explicit model-backed operators.

Everything you need to understand, evaluate, and use ajq.

ajq is usable today. The deterministic jq spine, semantic planning and execution, local inference (--backend local/mock), Ollama/OpenAI-compatible/Anthropic backends, cost controls, persistent judgement cache, local asset provisioning, model management, and release archives with a checksum-verifying install script are shipped. Scale-out/windowing remains a roadmap item. For the full picture, see Project status.

Start here

🎓 Tutorials

Run your first ajq pipeline and inspect how semantic JSON filters are planned.

🛠️ How-to guides

Install ajq, process JSON/NDJSON streams, write fuzzy filters, classify records, use cloud or local models, control costs, manage the cache, and configure defaults.

📚 Reference

CLI flags, configuration, backends, exit codes, I/O modes, semantic functions, and --explain fields.

💡 Explanation

Split execution, the determinism contract, the executor, architecture, and project status.

Not sure where to look?

“How do I get it running?” → Tutorial: Your first ajq pipeline
“How do I connect a provider?” → Use cloud backends
“How do I avoid surprise spend?” → Control semantic costs
“What does flag --foo do?” → Reference
“Why is it designed this way?” → Explanation

Feedback

Was this page helpful?

Glad to hear it! Please tell us how we can improve.

Sorry to hear that. Please tell us how we can improve.

Last modified July 10, 2026: docs(TP-057): align discovery positioning language (7c0ac62)