infrastructure-transcription

Provider-agnostic local speech-to-text (ASR): a Transcriber trait, a deterministic dev stub (zero native deps), and (feature-gated) an in-process whisper.cpp engine. Local-only by default — privileged audio never leaves the process.

What it is for

infrastructure-transcription — provider-agnostic local speech-to-text (ASR).

The surface is deliberately small: one Transcriber trait, plain input/output types (TranscriptionInput, Transcript), and a dependency-free StubTranscriber so the full ingest → transcribe → store → index pipeline builds and tests with zero native deps. The real in-process whisper.cpp impl lands behind the whisper-rs feature (WhisperRsTranscriber) and drops in behind the same trait; a faster-whisper subprocess escape hatch is reserved behind faster-whisper but is NOT the default.

Design + rationale: ADR 0019. Like the embedding surface, the default path performs NO external network egress — privileged audio stays local, which the legal consumer mandates (sending evidence audio to an external ASR API risks privilege waiver + confidentiality breach).

use infrastructure_transcription::{AudioSource, StubTranscriber, TranscribeOptions,
Transcriber, TranscriptionInput};
# async fn demo() -> Result<(), infrastructure_transcription::TranscriptionError> {
let t = StubTranscriber::new("es");
let input = TranscriptionInput::from_bytes(b"fake audio".to_vec());
let out = t.transcribe(&input, &TranscribeOptions::default()).await?;
assert!(!out.text.is_empty());
assert_eq!(out.language, "es");
# Ok(()) }

Capabilities

AudioSource

infrastructure-transcription — provider-agnostic local speech-to-text (ASR).

Segment

infrastructure-transcription — provider-agnostic local speech-to-text (ASR).

TranscribeOptions

infrastructure-transcription — provider-agnostic local speech-to-text (ASR).

Transcriber

infrastructure-transcription — provider-agnostic local speech-to-text (ASR).

Transcript

infrastructure-transcription — provider-agnostic local speech-to-text (ASR).

TranscriptionInput

infrastructure-transcription — provider-agnostic local speech-to-text (ASR).

How to use it

From this crate's own rustdoc:

use infrastructure_transcription::{AudioSource, StubTranscriber, TranscribeOptions,
    Transcriber, TranscriptionInput};
# async fn demo() -> Result<(), infrastructure_transcription::TranscriptionError> {
let t = StubTranscriber::new("es");
let input = TranscriptionInput::from_bytes(b"fake audio".to_vec());
let out = t.transcribe(&input, &TranscribeOptions::default()).await?;
assert!(!out.text.is_empty());
assert_eq!(out.language, "es");
# Ok(()) }

Module structure

No public modules: the crate root is its whole surface.

Public surface

`crate root`

Re-exports. Exported here, defined elsewhere.

Boundary

Depends on no other workspace tier.

Shares tier infrastructure with 82 other crates: infrastructure-acquire, infrastructure-adapters-google-calendar, infrastructure-adapters-google-gmail, infrastructure-adapters-google-places, infrastructure-adapters-google-trends, infrastructure-adapters-shodan, infrastructure-adapters-yelp, infrastructure-agent, … (82 total).

_What this crate deliberately does NOT own is a judgment. No committed registry records one for it, so none is stated here._

Where it sits

Dependencies

Runtime, from outside the workspace.

Development, from outside the workspace.

Build. None.

Depended on by. 1 workspace crate.

Signal flow — what reaches this crate, and what it reaches.

flowchart LR
  n_application_transcription["application-transcription"] -->|uses| SELF
  SELF["infrastructure-transcription"]
  classDef self fill:#1f883d,stroke:#1f883d,color:#fff;
  class SELF self;

Feature flags

flowchart LR
  n_cuda["cuda"] --> n_whisper_rs["whisper-rs"]
  n_cuda["cuda"] --> n_whisper_rs_cuda["whisper-rs/cuda"]
  n_default["default"]
  n_faster_whisper["faster-whisper"]
  n_whisper_rs["whisper-rs"] --> n_dep_whisper_rs["dep:whisper-rs"]
  n_whisper_rs["whisper-rs"] --> n_dep_ogg["dep:ogg"]
  n_whisper_rs["whisper-rs"] --> n_dep_opus["dep:opus"]
  n_whisper_rs["whisper-rs"] --> n_dep_hound["dep:hound"]
  n_whisper_rs["whisper-rs"] --> n_dep_tokio["dep:tokio"]

Targets

Error model

No public error type was detected: no public item declares a type named *Error, and no public signature returns one.

Operational characteristics

No unsafe block, unsafe fn, unsafe impl or unsafe trait was found by the parser anywhere in this crate's source.

Configuration

No environment variable is read with a literal name anywhere in this crate. A variable whose key is computed at run time cannot be listed here, and is not claimed to be absent.

Related capabilities

1 workspace crate depends on this one: application-transcription.

Verification

Evidence by module. How often each public module is named by something executable.

What the tests establish, by name:

• real_clip — tests/whisper_smoke.rs
• audio_shorter_than_target_is_one_unforced_chunk — src/chunk.rs
• empty_input_yields_no_chunks — src/chunk.rs
• every_sample_is_covered_by_at_least_one_chunk — src/chunk.rs
• forced_splits_overlap_so_a_boundary_word_is_not_dropped — src/chunk.rs
• forces_a_split_at_the_hard_cap_when_no_silence_gap_exists — src/chunk.rs
• splits_at_a_real_silence_gap_after_the_target_length — src/chunk.rs
• default_language_and_forced_language — src/dev.rs
• deterministic_same_audio — src/dev.rs
• different_audio_differs — src/dev.rs
• empty_audio_rejected — src/dev.rs
• over_cap_rejected — src/dev.rs
• path_input_hashes_path_deterministically — src/dev.rs
• translate_sets_english — src/dev.rs
• a_sentence_occurring_a_few_times_non_pathologically_is_not_flagged — src/repetition.rs
• below_threshold_consecutive_repeats_are_not_flagged — src/repetition.rs
• case_and_whitespace_differences_still_count_as_the_same_sentence — src/repetition.rs
• detects_the_exact_reproduced_failure_shape — src/repetition.rs
• empty_text_is_not_flagged — src/repetition.rs
• normal_varied_text_is_not_flagged — src/repetition.rs

Documentation coverage

pie showData
    title Public items with rustdoc
    "Documented" : 9
    "No rustdoc detected" : 1

Metrics

pie showData
    title Public API by kind
    "enum" : 1
    "method" : 4
    "struct" : 4
    "trait" : 1

pie showData
    title Rust source composition
    "Code" : 867
    "Blank or comment" : 339

Generation

Rendered by tools-corpus corpus readme from repository evidence alone, renderer schema 2, lexicon current. No model, network service or database was consulted. Regenerate with tools-corpus corpus readme --write; verify with --check.