Snippet Bridge

The problem

I keep about a hundred text snippets in Beeftext, the expander I type with. I also dictate with Wispr Flow, which has snippets of its own. Keeping two copies in step by hand is the kind of job that quietly stops happening.

What it does

Beeftext stays the master copy. One command converts the snippets into the JSON that Wispr Flow imports, then compares it with what Flow already has and reports what is new, what has changed and what exists only in Flow.

What it does, feature by feature

• One-way conversion. Each eligible Beeftext snippet becomes a Flow snippet with a spoken trigger; disabled, empty, oversized and duplicate ones are skipped and reported.
• Drift report. New, changed and Flow-only snippets, read from Flow without writing to it.
• Limits enforced. Triggers over 60 characters and texts over 4,000 are skipped and listed, never cut short.
• Safe import. Changes land through Flow's own bulk import, not a database edit.

Why it is built this way

Flow syncs its local database with its cloud and the last edit wins, so writing into that database directly risks being silently overwritten, or overwriting something. The script only ever reads it, and every change goes in through Flow's own import button.

What was hard

Triggers that fire by accident. Flow expands a snippet when its phrase turns up anywhere in a sentence, not only when you say it on its own, so a trigger like 'seo prompt' would paste a whole block into the middle of normal dictation. Triggers now start with 'beef' by default ('beef thank you'), which keeps them out of ordinary sentences, and the prefix can be changed. Flow's limits shape the rest: 60 characters per trigger and 4,000 per snippet, so anything longer is skipped and reported. And because the importer skips phrases it already has, an edited snippet is listed as changed so it can be deleted and imported again.

What it ships with

• One command
• Read-only on Flow
• Oversized snippets flagged

Stack