askM

Aviation professionals spend significant time searching through manuals for specific information. askM replaces that search with a conversation — entirely on your own device.

How it works

You import the technical documents you provide (ZIP packages or local storage). askM parses every document unit, chunks it, and builds a local hybrid search index — semantic vectors fused with full-text keyword search. Ask a question in plain language — “Can we dispatch with one pack inoperative?” — and an on-device language model answers from the retrieved passages, with numbered source citations you can tap to read the original section.

If the answer is not in your documents, askM says so. And since v1.2, the citations exactly match what the model actually read — every listed source was in the model’s context, no more and no less.

v1.2.1 repairs importing a manual directly from a file: depending on how the archive was packaged, none of its documents reached the library, and the app reported success anyway. Imports now land exactly like manuals downloaded from a server, a revised manual replaces the one it supersedes, and an archive that contains no manual is refused with a clear message.

Enhanced Search (v1.2)

The optional Enhanced Search model (about 350 MB, downloaded once from Settings) upgrades the semantic half of retrieval with an instruction-tuned embedding model running on-device via MLX. In our internal evaluation on a real MMEL — 22 graded dispatch questions — the correct passage reached the top five results 91% of the time with Enhanced Search versus 68% without it. Follow-up questions also carry the context of your previous question into the search, so “and for the other engine?” finds what you meant.

Built for aviation

• Aircraft fleet management with registration-based manual filtering — and a guard that never silently answers from another aircraft’s documentation
• Supports FCOM, FCTM, MMEL, CDL, QRH, AFM, APD, WBM and similar structured manuals
• Built-in document viewer with dark mode for cockpit use
• Visible indexing stages — watch the app clear, load the embedding model, parse-chunk-embed and finalise your library

Private by architecture

The language model runs on-device. Your documents, your questions and the answers never leave your iPad or Mac. There is no account, no cloud service and no telemetry — privacy is not a policy, it is the architecture.

askM is a personal study aid. It is not certified or intended for operational, maintenance, dispatch or any safety-critical use. You provide the documents and are responsible for having the rights to use them.