Added reference resolver WIP (codex)

main.py

@@ -14,7 +14,8 @@ import torch
 from asr import Asr, AsrRawResult
 from asr_filter import AsrFilter, AsrFilterResult
 from asr_eventizer import AsrEventizer
-from video_references import VideoReferenceBuilder
+from video_references import UnresolvedReferences, VideoReferenceBuilder
+from reference_resolver import ReferenceResolver
 from structure_builder import Structure, StructureBuilder
 from structure_refiner import StructureRefiner
 from markdown_builder import MarkdownBuilder
@@ -100,6 +101,31 @@ def setup_arguments() -> argparse.Namespace:
         type=str,
         default=ai_api_key
     )
+    parser.add_argument(
+        "--resolver-ai-model",
+        type=str,
+        default="google/gemini-3.5-flash"
+    )
+    parser.add_argument(
+        "--resolver-ai-base-url",
+        type=str,
+        default="https://api.proxyapi.ru/v1"
+    )
+    parser.add_argument(
+        "--resolver-ai-api-key",
+        type=str,
+        default=ai_api_key
+    )
+    parser.add_argument(
+        "--resolver-frame-interval",
+        type=float,
+        default=1.0
+    )
+    parser.add_argument(
+        "--resolver-max-frames",
+        type=int,
+        default=9
+    )
     parser.add_argument(
         "--structure-ai-model",
         type=str,
@@ -253,10 +279,27 @@ def on_reference_resolver(current_step: Step, input_data: dict | None) -> tuple[
             return (Step.STRUCTURE_BUILDER, json.load(f))
     if not os.path.isfile("video.mp4"):
         logging.info("No video, skipping")
-        return (Step.STRUCTURE_BUILDER, input_data)
-    # NOT IMPLEMENTED
-    logging.warning("Reference resolver is not implemented yet")
-    return (None, None)
+        with open("audio_events.json", "rb") as f:
+            return (Step.STRUCTURE_BUILDER, json.load(f))
+    if input_data is None:
+        logging.error("Can't resolve video references without input_data")
+        return (None, None)
+    with open("audio_events.json", "rb") as f:
+        timeline = Timeline(**json.load(f))
+    logging.info("Creating the agent")
+    agent = Agent(
+        model=ARGS.resolver_ai_model,
+        base_url=ARGS.resolver_ai_base_url,
+        api_key=ARGS.resolver_ai_api_key
+    )
+    logging.info("Resolving video references...")
+    resolver = ReferenceResolver(
+        agent,
+        frame_interval=ARGS.resolver_frame_interval,
+        max_frames=ARGS.resolver_max_frames,
+    )
+    result = resolver.resolve(timeline, UnresolvedReferences(**input_data))
+    return (Step.STRUCTURE_BUILDER, result.model_dump(mode="json"))
 
 def on_structure_builder(current_step: Step, input_data: dict | None) -> tuple[Step | None, dict | None]:
     # don't if done