code changes stemming from schema changes

crates/_types/enums/src/by_table/generic_inference_jobs/inference_input_source_token_type.rs

@@ -0,0 +1,79 @@
+use std::collections::BTreeSet;
+
+#[cfg(test)]
+use strum::EnumCount;
+#[cfg(test)]
+use strum::EnumIter;
+
+
+/// Used in the `generic_inference_jobs` table in `VARCHAR(32)` field `maybe_input_source_token`.
+///
+/// DO NOT CHANGE VALUES WITHOUT A MIGRATION STRATEGY.
+#[cfg_attr(test, derive(EnumIter, EnumCount))]
+#[derive(Clone, Copy, Eq, PartialEq, Hash, Ord, PartialOrd, Deserialize, Serialize)]
+pub enum InferenceInputSourceTokenType {
+  #[serde(rename = "media_upload")]
+  MediaUpload,
+}
+
+// TODO(bt, 2022-12-21): This desperately needs MySQL integration tests!
+impl_enum_display_and_debug_using_to_str!(InferenceInputSourceTokenType);
+impl_mysql_enum_coders!(InferenceInputSourceTokenType);
+
+/// NB: Legacy API for older code.
+impl InferenceInputSourceTokenType {
+  pub fn to_str(&self) -> &'static str {
+    match self {
+      Self::MediaUpload => "media_upload",
+    }
+  }
+
+  pub fn from_str(value: &str) -> Result<Self, String> {
+    match value {
+      "media_upload" => Ok(Self::MediaUpload),
+      _ => Err(format!("invalid value: {:?}", value)),
+    }
+  }
+
+  pub fn all_variants() -> BTreeSet<InferenceInputSourceTokenType> {
+    // NB: BTreeSet is sorted
+    // NB: BTreeSet::from() isn't const, but not worth using LazyStatic, etc.
+    BTreeSet::from([
+      InferenceInputSourceTokenType::MediaUpload,
+    ])
+  }
+}
+
+#[cfg(test)]
+mod tests {
+  use crate::by_table::generic_inference_jobs::inference_input_source_token_type::InferenceInputSourceTokenType;
+  use crate::test_helpers::assert_serialization;
+
+  #[test]
+  fn test_serialization() {
+    assert_serialization(InferenceInputSourceTokenType::MediaUpload, "media_upload");
+  }
+
+  #[test]
+  fn to_str() {
+    assert_eq!(InferenceInputSourceTokenType::MediaUpload.to_str(), "media_upload");
+  }
+
+  #[test]
+  fn from_str() {
+    assert_eq!(InferenceInputSourceTokenType::from_str("media_upload").unwrap(), InferenceInputSourceTokenType::MediaUpload);
+  }
+
+  #[test]
+  fn all_variants() {
+    // Static check
+    let mut variants = InferenceInputSourceTokenType::all_variants();
+    assert_eq!(variants.len(), 1);
+    assert_eq!(variants.pop_first(), Some(InferenceInputSourceTokenType::MediaUpload));
+    assert_eq!(variants.pop_first(), None);
+
+    // Generated check
+    use strum::IntoEnumIterator;
+    assert_eq!(InferenceInputSourceTokenType::all_variants().len(), InferenceInputSourceTokenType::iter().len());
+  }
+}

crates/_types/enums/src/by_table/generic_inference_jobs/inference_model_type.rs

@@ -18,6 +18,10 @@ use strum::EnumIter;
 #[cfg_attr(test, derive(EnumIter, EnumCount))]
 #[derive(Clone, Copy, Eq, PartialEq, Hash, Ord, PartialOrd, Deserialize, Serialize)]
 pub enum InferenceModelType {
+  #[serde(rename = "so_vits_svc")]
+  SoVitsSvc,
+
+  // TODO: Does this need to be "legacy_tacotron2" ?
   #[serde(rename = "tacotron2")]
   Tacotron2,
 
@@ -33,6 +37,7 @@ impl_mysql_enum_coders!(InferenceModelType);
 impl InferenceModelType {
   pub fn to_str(&self) -> &'static str {
     match self {
+      Self::SoVitsSvc => "so_vits_svc",
       Self::Tacotron2 => "tacotron2",
       Self::Vits => "vits",
     }
@@ -40,6 +45,7 @@ impl InferenceModelType {
 
   pub fn from_str(value: &str) -> Result<Self, String> {
     match value {
+      "so_vits_svc" => Ok(Self::SoVitsSvc),
       "tacotron2" => Ok(Self::Tacotron2),
       "vits" => Ok(Self::Vits),
       _ => Err(format!("invalid value: {:?}", value)),
@@ -50,6 +56,7 @@ impl InferenceModelType {
     // NB: BTreeSet is sorted
     // NB: BTreeSet::from() isn't const, but not worth using LazyStatic, etc.
     BTreeSet::from([
+      InferenceModelType::SoVitsSvc,
       InferenceModelType::Tacotron2,
       InferenceModelType::Vits,
     ])
@@ -63,18 +70,21 @@ mod tests {
 
   #[test]
   fn test_serialization() {
+    assert_serialization(InferenceModelType::SoVitsSvc, "so_vits_svc");
     assert_serialization(InferenceModelType::Tacotron2, "tacotron2");
     assert_serialization(InferenceModelType::Vits, "vits");
   }
 
   #[test]
   fn to_str() {
+    assert_eq!(InferenceModelType::SoVitsSvc.to_str(), "so_vits_svc");
     assert_eq!(InferenceModelType::Tacotron2.to_str(), "tacotron2");
     assert_eq!(InferenceModelType::Vits.to_str(), "vits");
   }
 
   #[test]
   fn from_str() {
+    assert_eq!(InferenceModelType::from_str("so_vits_svc").unwrap(), InferenceModelType::SoVitsSvc);
     assert_eq!(InferenceModelType::from_str("tacotron2").unwrap(), InferenceModelType::Tacotron2);
     assert_eq!(InferenceModelType::from_str("vits").unwrap(), InferenceModelType::Vits);
   }
@@ -83,7 +93,8 @@ mod tests {
   fn all_variants() {
     // Static check
     let mut variants = InferenceModelType::all_variants();
-    assert_eq!(variants.len(), 2);
+    assert_eq!(variants.len(), 3);
+    assert_eq!(variants.pop_first(), Some(InferenceModelType::SoVitsSvc));
     assert_eq!(variants.pop_first(), Some(InferenceModelType::Tacotron2));
     assert_eq!(variants.pop_first(), Some(InferenceModelType::Vits));
     assert_eq!(variants.pop_first(), None);

crates/_types/enums/src/by_table/generic_inference_jobs/mod.rs

@@ -1,3 +1,4 @@
 pub mod inference_category;
+pub mod inference_input_source_token_type;
 pub mod inference_model_type;
 pub mod inference_result_type;

crates/lib/mysql_queries/src/queries/generic_inference/web/insert_generic_inference_job.rs

@@ -1,6 +1,8 @@
 use anyhow::anyhow;
 use crate::payloads::generic_inference_args::GenericInferenceArgs;
 use enums::by_table::generic_inference_jobs::inference_category::InferenceCategory;
+use enums::by_table::generic_inference_jobs::inference_input_source_token_type::InferenceInputSourceTokenType;
+use enums::by_table::generic_inference_jobs::inference_model_type::InferenceModelType;
 use enums::common::visibility::Visibility;
 use errors::AnyhowResult;
 use sqlx::MySqlPool;
@@ -11,12 +13,16 @@ pub struct InsertGenericInferenceArgs<'a> {
   pub uuid_idempotency_token: &'a str,
 
   pub inference_category: InferenceCategory,
-  pub maybe_model_type: Option<&'a str>,
+  pub maybe_model_type: Option<InferenceModelType>,
   pub maybe_model_token: Option<&'a str>,
 
-  pub maybe_inference_args: Option<GenericInferenceArgs>,
+  pub maybe_input_source_token: Option<&'a str>,
+  pub maybe_input_source_token_type: Option<InferenceInputSourceTokenType>,
+
   pub maybe_raw_inference_text: Option<&'a str>,
 
+  pub maybe_inference_args: Option<GenericInferenceArgs>,
+
   pub maybe_creator_user_token: Option<&'a UserToken>,
   pub creator_ip_address: &'a str,
   pub creator_set_visibility: Visibility,
@@ -44,9 +50,13 @@ SET
   maybe_model_type = ?,
   maybe_model_token = ?,
 
-  maybe_inference_args = ?,
+  maybe_input_source_token = ?,
+  maybe_input_source_token_type = ?,
+
   maybe_raw_inference_text = ?,
 
+  maybe_inference_args = ?,
+
   maybe_creator_user_token = ?,
   creator_ip_address = ?,
   creator_set_visibility = ?,
@@ -60,12 +70,17 @@ SET
         args.uuid_idempotency_token,
 
         args.inference_category.to_str(),
-        args.maybe_model_type,
+
+        args.maybe_model_type.map(|t| t.to_str()),
         args.maybe_model_token,
 
-        serialized_args_payload,
+        args.maybe_input_source_token,
+        args.maybe_input_source_token_type,
+
         args.maybe_raw_inference_text,
 
+        serialized_args_payload,
+
         args.maybe_creator_user_token.map(|t| t.to_string()),
         args.creator_ip_address,
         args.creator_set_visibility.to_str(),

crates/lib/mysql_queries/src/queries/media_uploads/get_media_upload_for_inference.rs

@@ -59,7 +59,7 @@ SELECT
     mu.public_bucket_directory_hash,
     mu.original_file_size_bytes,
     mu.original_duration_millis,
-    mu.extra_file_modification_info as maybe_extra_file_modification_info,
+    mu.maybe_extra_file_modification_info,
     mu.creator_set_visibility as `creator_set_visibility: enums::common::visibility::Visibility`,
     mu.created_at,
     mu.updated_at

crates/lib/mysql_queries/src/queries/media_uploads/insert_media_upload.rs

@@ -29,15 +29,13 @@ pub struct Args <'a> {
   pub checksum_sha2: &'a str,
 
   pub public_upload_path: &'a MediaUploadOriginalFilePath,
-  pub extra_file_modification_info: MediaUploadModificationDetails,
+  pub maybe_extra_file_modification_info: Option<MediaUploadModificationDetails>,
 
   pub maybe_creator_user_token: Option<&'a UserToken>,
   pub maybe_creator_anonymous_visitor_token: Option<&'a str>,
   pub creator_ip_address: &'a str,
   pub creator_set_visibility: Visibility,
 
-  pub maybe_creator_synthetic_id: Option<u64>,
-
   pub mysql_pool: &'a MySqlPool,
 }
 
@@ -103,6 +101,11 @@ LIMIT 1
 
   let media_token = MediaUploadToken::generate();
 
+  let maybe_extra_file_modification_info = args.maybe_extra_file_modification_info
+      .clone() // FIXME
+      .map(|info| info.to_json())
+      .transpose()?;
+
   let query = sqlx::query!(
         r#"
 INSERT INTO media_uploads
@@ -122,8 +125,8 @@ SET
   checksum_sha2 = ?,
 
   public_bucket_directory_hash = ?,
-  public_bucket_directory_full_path = ?,
-  extra_file_modification_info = ?,
+
+  maybe_extra_file_modification_info = ?,
 
   maybe_creator_user_token = ?,
   maybe_creator_synthetic_id = ?,
@@ -147,8 +150,8 @@ SET
         args.checksum_sha2,
 
         args.public_upload_path.get_object_hash(),
-        args.public_upload_path.get_full_object_path_str(),
-        "", // TODO: args.extra_file_modification_info,
+
+        maybe_extra_file_modification_info,
 
         args.maybe_creator_user_token,
         maybe_creator_synthetic_id,

crates/lib/mysql_queries/src/queries/voice_conversion/results/insert_voice_conversion_result.rs

@@ -102,7 +102,7 @@ SET
   model_token = ?,
 
   media_token = ?,
-  media_token_type= ?,
+  media_token_type = ?,
 
   maybe_creator_user_token = ?,
   maybe_creator_synthetic_id = ?,

crates/service/web/storyteller_web/src/http_server/endpoints/media_uploads/upload_media.rs

@@ -273,12 +273,11 @@ pub async fn upload_media_handler(
     maybe_original_frame_height: None,
     checksum_sha2: &hash,
     public_upload_path: &public_upload_path,
-    extra_file_modification_info: MediaUploadModificationDetails {}, // TODO
+    maybe_extra_file_modification_info: None,
     maybe_creator_user_token: maybe_user_token.as_ref(),
     maybe_creator_anonymous_visitor_token: None,
     creator_ip_address: &ip_address,
     creator_set_visibility,
-    maybe_creator_synthetic_id: None, // TODO: Don't forget about this.
     mysql_pool: &server_state.mysql_pool,
   })
       .await

crates/service/web/storyteller_web/src/http_server/endpoints/tts/enqueue_infer_tts_handler/enqueue_infer_tts_handler.rs

@@ -12,6 +12,7 @@ use crate::http_server::endpoints::tts::enqueue_infer_tts_handler::get_model_wit
 use crate::http_server::web_utils::response_error_helpers::to_simple_json_error;
 use crate::server_state::ServerState;
 use enums::by_table::generic_inference_jobs::inference_category::InferenceCategory;
+use enums::by_table::generic_inference_jobs::inference_model_type::InferenceModelType;
 use enums::by_table::tts_models::tts_model_type::TtsModelType;
 use enums::common::visibility::Visibility;
 use http_server_common::request::get_request_api_token::get_request_api_token;
@@ -353,6 +354,11 @@ pub async fn enqueue_infer_tts_handler(
   let job_token;
   let job_token_type;
 
+  let model_type = match tts_model.tts_model_type {
+    TtsModelType::Tacotron2 => InferenceModelType::Tacotron2,
+    TtsModelType::Vits => InferenceModelType::Vits,
+  };
+
   if use_new_job_system {
     // This branch uses the `generic-inference-job` service and tables.
     info!("Creating tts inference job record (new generic job system)...");
@@ -364,13 +370,15 @@ pub async fn enqueue_infer_tts_handler(
     let query_result = insert_generic_inference_job(InsertGenericInferenceArgs {
       uuid_idempotency_token: &request.uuid_idempotency_token,
       inference_category: InferenceCategory::TextToSpeech,
-      maybe_model_type: None, // TODO(bt, 2023-04-08): Add this
+      maybe_model_type: Some(model_type),
       maybe_model_token: Some(request.tts_model_token.as_str()),
+      maybe_input_source_token: None, // NB: TTS doesn't have input media
+      maybe_input_source_token_type: None, // NB: TTS doesn't have input media
+      maybe_raw_inference_text: Some(inference_text.as_str()),
       maybe_inference_args: Some(GenericInferenceArgs {
         inference_category: Some(InferenceCategoryAbbreviated::TextToSpeech),
         args: None, // NB: We don't need to encode any args yet.
       }),
-      maybe_raw_inference_text: Some(inference_text.as_str()),
       maybe_creator_user_token: maybe_creator_user_token_typed.as_ref(),
       creator_ip_address: &ip_address,
       creator_set_visibility: set_visibility,

crates/service/web/storyteller_web/src/http_server/endpoints/voice_conversion/inference/enqueue_voice_conversion_inference.rs

@@ -21,6 +21,8 @@ use r2d2_redis::redis::Commands;
 use redis_common::redis_keys::RedisKeys;
 use std::fmt;
 use std::sync::Arc;
+use enums::by_table::generic_inference_jobs::inference_input_source_token_type::InferenceInputSourceTokenType;
+use enums::by_table::generic_inference_jobs::inference_model_type::InferenceModelType;
 use tokens::files::media_upload::MediaUploadToken;
 use tokens::jobs::inference::InferenceJobToken;
 use tokens::users::user::UserToken;
@@ -224,15 +226,17 @@ pub async fn enqueue_voice_conversion_inference_handler(
   let query_result = insert_generic_inference_job(InsertGenericInferenceArgs {
     uuid_idempotency_token: &request.uuid_idempotency_token,
     inference_category: InferenceCategory::VoiceConversion,
-    maybe_model_type: None, // TODO(bt, 2023-04-08): Add this.
+    maybe_model_type: Some(InferenceModelType::SoVitsSvc),
     maybe_model_token: Some(model_token.as_str()),
+    maybe_input_source_token: Some(media_token.as_str()),
+    maybe_input_source_token_type: Some(InferenceInputSourceTokenType::MediaUpload),
+    maybe_raw_inference_text: None, // NB: Voice conversion isn't TTS, so there's no text.
     maybe_inference_args: Some(GenericInferenceArgs {
       inference_category: Some(InferenceCategoryAbbreviated::VoiceConversion),
       args: Some(PolymorphicInferenceArgs::VoiceConversionInferenceArgs {
-        maybe_media_token: Some(media_token),
+        maybe_media_token: Some(media_token.clone()),
       }),
     }),
-    maybe_raw_inference_text: None,
     maybe_creator_user_token: maybe_user_token.as_ref(),
     creator_ip_address: &ip_address,
     creator_set_visibility: set_visibility,