open-standards/eoc/eoc-rs/crates/eoc-multimodal/tests/http_mocks.rs
Transaction Science 3fe109d58f ingest batch 4: 9 crates · table-stakes completers across all three workspaces
OpenPay (3 crates, ~13k LOC, 175 tests)
  op-connect           sub-merchant onboarding + KYB + split-payments + payouts + 1099-K
  op-3ds2              3DS2.3 SCA processor (AReq/ARes/CReq/CRes/RReq/RRes + exemptions)
  op-tax               US/EU/UK/AU/CA/IN tax with Avalara/Stripe/CCH/Vertex/TaxJar adapters

Smart Byte (3 crates, ~7k LOC, 57 tests)
  smart-byte-libp2p    libp2p transport (kad/gossipsub/mdns/autonat/relay/dcutr/identify/ping/req_resp)
  smart-byte-crdt      LWW/GCounter/PnCounter/OrSet/LwwMap/RgaList/2P-Set + HLC + vector clock + Automerge/Yjs interop
  smart-byte-keri-witness  KERI key-rotation cycle + witness layer + watchers + recovery + delegation

EOC (3 crates, ~8k LOC, 108 tests)
  eoc-tools            function-calling across Anthropic/OpenAI/Google/Mistral/Cohere + parallel exec + sandbox
  eoc-multimodal       vision (CLIP/SigLIP/ColPali) + audio (Whisper/MMS) + video sampling + modality router
  eoc-spec-decode      vanilla + SpS-with-temp + lookahead speculative decoding + EAGLE/Medusa stubs

All three workspaces compile cleanly. Forbidden-unsafe, no unwrap outside tests.

Co-Authored-By: Claude Opus 4.7 <[email protected]>
2026-05-24 04:00:04 -04:00

92 lines
3.3 KiB
Rust

//! End-to-end HTTP behaviour tests using `wiremock`.
//!
//! Covers Whisper transcription and OpenAI TTS — the two vendor calls
//! that don't share the chat-completions body schema and therefore need
//! their own fixtures.
use eoc_multimodal::audio::tts::{OpenAiTtsBackend, Synthesizer, VoiceSpec};
use eoc_multimodal::audio::whisper_api::{Transcriber, WhisperApiBackend};
use eoc_multimodal::modality::AudioRef;
use serde_json::json;
use wiremock::matchers::{header, header_exists, method, path};
use wiremock::{Mock, MockServer, ResponseTemplate};
#[tokio::test]
async fn whisper_known_bytes_returns_expected_text() {
let server = MockServer::start().await;
let body = json!({
"text": "hello world",
"language": "en",
"duration": 1.5,
"segments": [
{"start": 0.0, "end": 1.5, "text": "hello world"}
]
});
Mock::given(method("POST"))
.and(path("/"))
.and(header("Authorization", "Bearer sk-test"))
.and(header_exists("Content-Type"))
.respond_with(ResponseTemplate::new(200).set_body_json(body))
.mount(&server)
.await;
let backend = WhisperApiBackend::new("sk-test", "whisper-1").with_endpoint(server.uri());
let audio = AudioRef::Bytes {
content_type: "audio/wav".to_string(),
bytes: vec![0x00, 0x01, 0x02, 0x03],
};
let result = backend.transcribe(&audio).await.expect("ok");
assert_eq!(result.text, "hello world");
assert_eq!(result.language.as_deref(), Some("en"));
assert_eq!(result.segments.len(), 1);
// 1.5 s * 5 J/s = 7.5 J = 7_500_000 µJ.
assert_eq!(result.joule_cost.microjoules, 7_500_000);
}
#[tokio::test]
async fn whisper_401_returns_invalid_api_key() {
let server = MockServer::start().await;
Mock::given(method("POST"))
.and(path("/"))
.respond_with(ResponseTemplate::new(401).set_body_string("denied"))
.mount(&server)
.await;
let backend = WhisperApiBackend::new("sk-bad", "whisper-1").with_endpoint(server.uri());
let audio = AudioRef::Bytes {
content_type: "audio/wav".to_string(),
bytes: vec![0x00],
};
let err = backend.transcribe(&audio).await.expect_err("should fail");
assert!(matches!(
err,
eoc_multimodal::MultimodalError::InvalidApiKey
));
}
#[tokio::test]
async fn tts_returns_audio_bytes_with_content_type() {
let server = MockServer::start().await;
let synthetic_wav: Vec<u8> = b"RIFF....WAVEfmt ".to_vec();
Mock::given(method("POST"))
.and(path("/"))
.and(header("Authorization", "Bearer sk-test"))
.respond_with(
ResponseTemplate::new(200)
.insert_header("Content-Type", "audio/wav")
.set_body_bytes(synthetic_wav.clone()),
)
.mount(&server)
.await;
let backend = OpenAiTtsBackend::new("sk-test", "tts-1").with_endpoint(server.uri());
let out = backend
.synthesize("hello world", VoiceSpec::default_alloy())
.await
.expect("ok");
assert_eq!(out.content_type, "audio/wav");
assert_eq!(out.bytes, synthetic_wav);
// "hello world" is 11 chars / 150 cps = 0.0733 s * 1.5 J/s ≈ 110_000 µJ.
assert!(out.joule_cost.microjoules > 0);
assert!(out.joule_cost.microjoules < 1_000_000);
}