Files
deepseek-harness/packages/experimental/api-speech-to-text
Turtle e7def469e1 feat(i18n): key translation pairing records by heading section (#5036)
* feat(i18n): key translation pairing records by heading section

Replace the whole-file blob hashes in *.i18n.yaml with one entry per
heading section that still holds language-specific content, keyed by the
English heading-slug path. Blocks identical across the pair (generated
table rows, code blocks, shared regions) stay out of the hash, so edits to
different sections and identical regenerations merge with Git's default
text merge, including on GitHub.

Remove the dsh-translation-pairing merge driver, its resolver command and
installer wiring; gen-translation-brief recovers last-confirmed text from
commit history. Migrate all active records, archive the superseded Agent
Note, and record the replayed conflict reduction in the new one.

* test(i18n): refresh translation prompt snapshot for the terminology note

* feat(i18n): exclude only code blocks and generated regions from pairing records

Replace byte-equality masking with an explicit rule: a record hash skips
fenced code blocks and BEGIN/END GENERATED regions, the two kinds of
content the pairing gate already requires to be identical on both sides.

Move generated data into regions written into both pages:
gen-config-catalog renders both languages and the record with per-package
regions labelled by code identifiers; gen-module-graph keeps its table in a
region; gen-doc-graphs splices the event matrix regions into the Chinese
page. Add a generic region splice beside the region grammar.
2026-09-23 22:32:06 +08:00
..
2026-09-23 20:47:45 +08:00

description, kind
description kind
Expose bounded transient transcription through the authenticated Web Remote. package-reference

@deepseek-ai/dsh-experimental-api-speech-to-text

English | 中文

Summary

The speech Remote connects browser recordings to ctx.speechToText. It exposes provider discovery and one complete-recording transcription call.

Table of Contents


Use this package

Compose with the speech Service Definition and Typert. maxAudioBytes and maxDurationSeconds limit accepted recordings. The browser UI mounts this package’s generated /remote contribution when enabled.

prepare(providerId, { downloadSource }) forwards one advertised source to the provider; omission keeps its configured policy. The catalog carries downloadSources for the picker. The provider rejects unavailable choices and changes to an active task’s source.


Understand the implementation

Maintainer details — click to expand

catalog() exposes provider identities, the default selection and recording limits. transcribe() validates canonical base64 and 16 kHz mono PCM16 WAV before resolving the selected provider. The existing gateway owns authentication and cancellation transport. Audio is transient, never a Session event or attachment; only the user’s later ordinary submission records recognized text. No runtime invariant companion is published because validation is stateless and preparation belongs to the provider.

follow() streams complete catalogs, including preparation states and current preferences. prepare() starts or joins a Host task; cancelPreparation() explicitly cancels it. configure() persists the supplied preference fields. A disconnected observer does not cancel preparation.


Further Exploration

Voice input subsystem


Model Experience

None, as recordings and preparation remain outside model requests; ordinary user submission owns any later text.

KV Cache effect

No direct effect; ordinary submission owns the message content.

Known Limitations and Deferred Work

  • No file upload, persistent transcript history or streaming protocol is exposed. Base64 adds transport overhead; recordings remain bounded by the advertised limits.

Dev Note

Maintainer details — click to expand

None.