feat: task-based video generation provider

Cargo.toml

@@ -23,7 +23,7 @@ serde_json = "1"
 toml = "0.8"
 
 # HTTP
-reqwest = { version = "0.12", features = ["json", "multipart"] }
+reqwest = { version = "0.12", features = ["json", "multipart", "stream"] }
 axum = { version = "0.8", features = ["ws"] }
 axum-extra = { version = "0.10", features = ["typed-header"] }
 tower = "0.5"

mofa-engine-core/Cargo.toml

@@ -13,6 +13,8 @@ serde = { workspace = true }
 serde_json = { workspace = true }
 toml = { workspace = true }
 reqwest = { workspace = true }
+tokio-stream = { workspace = true }
+base64 = "0.22"
 thiserror = { workspace = true }
 anyhow = { workspace = true }
 tracing = { workspace = true }
@@ -22,3 +24,6 @@ sysinfo = { workspace = true }
 dashmap = { workspace = true }
 dirs = { workspace = true }
 tempfile = { workspace = true }
+
+[dev-dependencies]
+axum = { workspace = true }

mofa-engine-core/src/backends.rs

@@ -3,7 +3,9 @@
 pub mod local_tts;
 pub mod ollama;
 pub mod openai_compat;
+pub mod video_task;
 
 pub use local_tts::LocalTtsProvider;
 pub use ollama::OllamaProvider;
 pub use openai_compat::OpenAiCompatProvider;
+pub use video_task::VideoTaskProvider;

mofa-engine-core/src/backends/local_tts.rs

@@ -205,6 +205,8 @@ impl LocalTtsProvider {
             tokens_used: None,
             fallback_used: false,
             routing_reason: None,
+            reasoning: None,
+            files: Vec::new(),
         })
     }
 }
@@ -337,6 +339,7 @@ mod tests {
             messages: vec![Message {
                 role: "user".into(),
                 content: text.into(),
+                images: Vec::new(),
             }],
             input_file: None,
             params: serde_json::Value::Null,

mofa-engine-core/src/backends/ollama.rs

@@ -98,6 +98,9 @@ struct OllamaChatRequest {
 struct OllamaMessage {
     role: String,
     content: String,
+    /// Bare base64 image payloads (Ollama chat API format).
+    #[serde(default, skip_serializing_if = "Vec::is_empty")]
+    images: Vec<String>,
 }
 
 #[derive(Debug, Deserialize)]
@@ -211,10 +214,12 @@ impl Provider for OllamaProvider {
             messages: vec![OllamaMessage {
                 role: "user".into(),
                 content: " ".into(),
+                images: Vec::new(),
             }],
             stream: false,
             keep_alive: Some("5m".into()),
         };
+        // (health probe: no images)
         let url = format!("{}/api/chat", self.base_url);
         let resp = self
             .client
@@ -251,6 +256,7 @@ impl Provider for OllamaProvider {
             messages: vec![OllamaMessage {
                 role: "user".into(),
                 content: " ".into(),
+                images: Vec::new(),
             }],
             stream: false,
             keep_alive: Some("0".into()),
@@ -298,6 +304,16 @@ impl Provider for OllamaProvider {
             .map(|m| OllamaMessage {
                 role: m.role.clone(),
                 content: m.content.clone(),
+                // Ollama wants bare base64; kernel messages carry data URLs.
+                images: m
+                    .images
+                    .iter()
+                    .map(|url| {
+                        url.split_once(",")
+                            .map(|(_, payload)| payload.to_string())
+                            .unwrap_or_else(|| url.clone())
+                    })
+                    .collect(),
             })
             .collect();
 
@@ -354,6 +370,8 @@ impl Provider for OllamaProvider {
             tokens_used: chat_resp.eval_count,
             fallback_used: false,
             routing_reason: None,
+            reasoning: None,
+            files: Vec::new(),
         })
     }
 }