| 1 | //! `transcribe_selected` keeps a recording on the ASR route that was selected. |
| 2 | //! |
| 3 | //! The configured provider in every fixture below has a literal key and a |
| 4 | //! loopback wiremock endpoint, so a regression that falls back to it is |
| 5 | //! observable as a recorded request rather than inferred from a sleep. |
| 6 | |
| 7 | use wiremock::matchers::{method, path}; |
| 8 | use wiremock::{Mock, MockServer, ResponseTemplate}; |
| 9 | |
| 10 | use crate::config::Config; |
| 11 | use crate::test_support::{EnvVarGuard, lock_test_env}; |
| 12 | use crate::voice::{ASR_MODEL, DictateError, VOICE_CONTROL_MODEL, transcribe_selected}; |
| 13 | |
| 14 | const FIXTURE_KEY: &str = "fixture-provider-key"; |
| 15 | |
| 16 | async fn provider_fixture(content: &str) -> MockServer { |
| 17 | let server = MockServer::start().await; |
| 18 | Mock::given(method("POST")) |
| 19 | .and(path("/v1/chat/completions")) |
| 20 | .respond_with(ResponseTemplate::new(200).set_body_json(serde_json::json!({ |
| 21 | "choices": [{ "message": { "content": content } }] |
| 22 | }))) |
| 23 | .mount(&server) |
| 24 | .await; |
| 25 | server |
| 26 | } |
| 27 | |
| 28 | /// Fully explicit provider route: the key and endpoint come from the config |
| 29 | /// table, so nothing is read from real auth state or ambient endpoint env. |
| 30 | fn fixture_config(server: &MockServer) -> Config { |
| 31 | toml::from_str(&format!( |
| 32 | "provider = \"openai\"\n\n[providers.openai]\napi_key = \"{FIXTURE_KEY}\"\nbase_url = \"{}/v1\"\nmodel = \"fixture-model\"\n", |
| 33 | server.uri() |
| 34 | )) |
| 35 | .expect("fixture config parses") |
| 36 | } |
| 37 | |
| 38 | /// Hold every env mutation for the whole test: the returned guards live until |
| 39 | /// the caller drops them, which is after any `spawn_blocking` transcriber has |
| 40 | /// been awaited to completion. |
| 41 | struct IsolatedEnv { |
| 42 | // Field order is drop order: guards restore before the lock releases. |
| 43 | _guards: Vec<EnvVarGuard>, |
| 44 | _home: tempfile::TempDir, |
| 45 | _lock: crate::test_support::TestEnvLock, |
| 46 | } |
| 47 | |
| 48 | fn isolated_env(path_override: Option<&std::path::Path>) -> IsolatedEnv { |
| 49 | let lock = lock_test_env(); |
| 50 | let home = tempfile::tempdir().expect("isolated codewhale home"); |
| 51 | let mut guards = vec![ |
| 52 | EnvVarGuard::set("CODEWHALE_HOME", home.path()), |
| 53 | EnvVarGuard::remove("GROQ_API_KEY"), |
| 54 | EnvVarGuard::remove("CODEWHALE_BASE_URL"), |
| 55 | EnvVarGuard::remove("DEEPSEEK_BASE_URL"), |
| 56 | EnvVarGuard::remove("OPENAI_BASE_URL"), |
| 57 | EnvVarGuard::remove("OPENAI_API_KEY"), |
| 58 | EnvVarGuard::remove(codewhale_config::CLI_API_KEY_ENV), |
| 59 | EnvVarGuard::remove(codewhale_config::CLI_API_KEY_SOURCE_ENV), |
| 60 | EnvVarGuard::remove(codewhale_config::LEGACY_CLI_API_KEY_SOURCE_ENV), |
| 61 | ]; |
| 62 | if let Some(dir) = path_override { |
| 63 | guards.push(EnvVarGuard::set("PATH", dir)); |
| 64 | } |
| 65 | IsolatedEnv { |
| 66 | _guards: guards, |
| 67 | _home: home, |
| 68 | _lock: lock, |
| 69 | } |
| 70 | } |
| 71 | |
| 72 | #[tokio::test] |
| 73 | async fn failed_local_whisper_never_uploads_to_the_configured_provider() { |
| 74 | // No whisper binary resolves from an empty PATH; HOME is left alone. |
| 75 | let empty_path = tempfile::tempdir().expect("empty PATH dir"); |
| 76 | let _env = isolated_env(Some(empty_path.path())); |
| 77 | let server = provider_fixture("{\"text\":\"must not be requested\"}").await; |
| 78 | let config = fixture_config(&server); |
| 79 | |
| 80 | let result = transcribe_selected(&config, "local-whisper", &[0; 16], None).await; |
| 81 | |
| 82 | assert!( |
| 83 | matches!(&result, Err(DictateError::Transcription(msg)) if msg.contains("local whisper")), |
| 84 | "{result:?}" |
| 85 | ); |
| 86 | let requests = server.received_requests().await.expect("request log"); |
| 87 | assert_eq!( |
| 88 | requests.len(), |
| 89 | 0, |
| 90 | "audio reached the provider: {requests:?}" |
| 91 | ); |
| 92 | } |
| 93 | |
| 94 | #[tokio::test] |
| 95 | async fn failed_groq_without_a_key_never_falls_back_to_the_configured_provider() { |
| 96 | let _env = isolated_env(None); |
| 97 | let server = provider_fixture("{\"text\":\"must not be requested\"}").await; |
| 98 | let config = fixture_config(&server); |
| 99 | |
| 100 | let result = transcribe_selected(&config, "groq", &[0; 16], Some("existing text")).await; |
| 101 | |
| 102 | assert!( |
| 103 | matches!(&result, Err(DictateError::Transcription(msg)) if msg.contains("GROQ_API_KEY")), |
| 104 | "{result:?}" |
| 105 | ); |
| 106 | let requests = server.received_requests().await.expect("request log"); |
| 107 | assert_eq!( |
| 108 | requests.len(), |
| 109 | 0, |
| 110 | "audio reached the provider: {requests:?}" |
| 111 | ); |
| 112 | } |
| 113 | |
| 114 | #[tokio::test] |
| 115 | async fn selected_provider_asr_uses_plain_or_composer_aware_request() { |
| 116 | let _env = isolated_env(None); |
| 117 | let server = provider_fixture("{\"text\":\"hello from provider\"}").await; |
| 118 | let config = fixture_config(&server); |
| 119 | |
| 120 | // Plain ASR returns the raw message content. |
| 121 | let plain = transcribe_selected(&config, "provider", &[0; 16], None) |
| 122 | .await |
| 123 | .expect("plain provider ASR"); |
| 124 | assert_eq!(plain, "{\"text\":\"hello from provider\"}"); |
| 125 | // Voice control parses the JSON reply and sends the composer text along. |
| 126 | let assisted = transcribe_selected(&config, "provider", &[0; 16], Some("draft so far")) |
| 127 | .await |
| 128 | .expect("voice-control provider ASR"); |
| 129 | assert_eq!(assisted, "hello from provider"); |
| 130 | |
| 131 | let requests = server.received_requests().await.expect("request log"); |
| 132 | assert_eq!(requests.len(), 2, "{requests:?}"); |
| 133 | for request in &requests { |
| 134 | assert_eq!( |
| 135 | request |
| 136 | .headers |
| 137 | .get("authorization") |
| 138 | .and_then(|value| value.to_str().ok()), |
| 139 | Some(format!("Bearer {FIXTURE_KEY}").as_str()) |
| 140 | ); |
| 141 | } |
| 142 | let plain_body: serde_json::Value = serde_json::from_slice(&requests[0].body).unwrap(); |
| 143 | assert_eq!(plain_body["model"], ASR_MODEL); |
| 144 | assert!(!plain_body.to_string().contains("draft so far")); |
| 145 | |
| 146 | let control_body: serde_json::Value = serde_json::from_slice(&requests[1].body).unwrap(); |
| 147 | assert_eq!(control_body["model"], VOICE_CONTROL_MODEL); |
| 148 | let context = control_body["messages"][1]["content"][0]["text"] |
| 149 | .as_str() |
| 150 | .expect("composer context block"); |
| 151 | let context: serde_json::Value = serde_json::from_str(context).unwrap(); |
| 152 | assert_eq!(context["current_text"], "draft so far"); |
| 153 | } |
| 154 |