Huge refactoring

utils.py

@@ -1,62 +1,79 @@
 from dataclasses import dataclass
-from typing import Literal
+from typing import Literal, Any
+import subprocess
+
 
 @dataclass
-class TimelineEvent:
-    id: str
-    """ID of the event in format `asr_18`"""
+class Event:
+    """Event within the timeline"""
+
+    id: int
+    """Unique ID of the event (within single timeline)"""
+
+    type: Literal["voice", "vis", "ocr"]
+    """Type of the event"""
 
     timestamp: float
-    """When did the event happen"""
+    """When did the event start (or happen, if `duration` is 0)"""
 
     duration: float
-    """How long did the event last (zero if it does not make sence)"""
+    """How long did the event last (0 if can't apply)"""
 
-    payload: str
-    """Payload of the event (text for `asr`, description for `vis`, OCR result for `ocr`)"""
+    text: str
+    """Textual representation of the event (or data in `payload`, if it is not None)"""
 
-    custom: dict
-    """Custom data"""
-
-    links: list[str]
-    """ID of related timeline events, empty list for None"""
-
-    def get_ai_dict(self) -> dict:
-        """Returns dict that is sanitized for AI."""
-        d = {
-            "id": self.id,
-            "timestamp": self.timestamp,
-            "payload": self.payload
-        }
-        if self.duration:
-            d["duration"] = self.duration
-        if self.links:
-            d["links"] = self.links
-        return d
+    payload: str | None
+    """Payload of the event. `None` for `voice`, image path for `vis`, OCR result for `ocr`"""
 
 @dataclass
-class AnalysisWindow:
-    before_readonly: list[TimelineEvent]
-    """Timeline events for context (before current window)"""
+class Timeline:
+    """Timeline of events"""
 
-    modifiable: list[TimelineEvent]
-    """Timeline events that are to be analyzed"""
+    events: list[Event]
+    """Events of the timeline"""
 
-    after_readonly: list[TimelineEvent]
-    """Timeline events for context (after current window)"""
+def ffmpeg_split_video(input_path: str) -> tuple[str, str]:
+    """Extract audio and video into ./audio.mp3 and ./video.mp4"""
+    audio_path = "audio.mp3"
+    video_path = "video.mp4"
+    subprocess.run(
+        [
+            "ffmpeg",
+            "-y",
+            "-i", input_path,
 
-@dataclass
-class AgentMessage:
-    content: str
-    """Content of the message"""
+            # Audio
+            "-map", "0:a:0",
+            "-vn",
+            "-c:a", "libmp3lame",
+            "-q:a", "2",
+            audio_path,
 
-    role: Literal["system", "assistant", "user"]
-    """Who sent the message"""
+            # Video
+            "-map", "0:v:0",
+            "-an",
+            "-c:v", "libx264",
+            "-crf", "18",
+            "-preset", "fast",
+            video_path,
+        ],
+        check=True,
+    )
+    return audio_path, video_path
 
-@dataclass
-class TimelineProcessingResult:
-    events: list[TimelineEvent]
-    """List of events"""
-
-    desired_events: list[TimelineEvent]
-    """List of events desired for existance"""
+def ffmpeg_to_mp3(input_path: str) -> str:
+    """Convert `input_path` path to mp3 and save it at ./audio.mp3"""
+    output_path = "audio.mp3"
+    subprocess.run(
+        [
+            "ffmpeg",
+            "-y",
+            "-i", input_path,
+            "-vn",
+            "-c:a", "libmp3lame",
+            "-q:a", "2",
+            output_path,
+        ],
+        check=True,
+    )
+    return output_path