[KDD'2026] "VideoRAG: Chat with Your Videos"
Python large-language-modelsllmslong-video-understandingmulti-modal-llmsragretrieval-augmented-generation
Fix analyzing vidoes. Changes should be all done
Gerald Hewes7f2989c4d64b
diff --git a/longervideos/videorag_experiment.py b/longervideos/videorag_experiment.pyindex ea04101..1c7067a 100644--- a/longervideos/videorag_experiment.py+++ b/longervideos/videorag_experiment.py@@ -38,8 +38,7 @@ if __name__ == '__main__':with open(f'longervideos/dataset.json', 'r') as f:longervideos = json.load(f)- #videorag = VideoRAG(cheap_model_func=gpt_4o_mini_complete, best_model_func=gpt_4o_mini_complete, working_dir=f"./videorag-workdir/{sub_category}")- videorag = VideoRAG(cheap_model_func=ollama_mini_complete, best_model_func=ollama_complete, working_dir=f"./videorag-workdir/{sub_category}")+ videorag = VideoRAG(llm=ollama_config, working_dir=f"./videorag-workdir/{sub_category}")videorag.load_caption_model(debug=False)answer_folder = f'./videorag-answers/{sub_category}'diff --git a/videorag/_op.py b/videorag/_op.pyindex 05da8eb..57cb8b0 100755--- a/videorag/_op.py+++ b/videorag/_op.py@@ -183,8 +183,8 @@ async def _handle_entity_relation_summary(description: str,global_config: dict,) -> str:- use_llm_func: callable = global_config["cheap_model_func"]- llm_max_tokens = global_config["cheap_model_max_token_size"]+ use_llm_func: callable = global_config["llm"]["cheap_model_func"]+ llm_max_tokens = global_config["llm"]["cheap_model_max_token_size"]tiktoken_model_name = global_config["tiktoken_model_name"]summary_max_tokens = global_config["entity_summary_to_max_tokens"]@@ -359,7 +359,7 @@ async def extract_entities(entity_vdb: BaseVectorStorage,global_config: dict,) -> Union[BaseGraphStorage, None]:- use_llm_func: callable = global_config["best_model_func"]+ use_llm_func: callable = global_config["llm"]["best_model_func"]entity_extract_max_gleaning = global_config["entity_extract_max_gleaning"]ordered_chunks = list(chunks.items())