Update

.github/workflows/ci.yml

@@ -52,4 +52,4 @@ jobs:
           IS_PAI_RAG_CI_TEST: true
           PAIRAG_RAG__embedding__source: "DashScope"
           PAIRAG_RAG__llm__source: "DashScope"
-          PAIRAG_RAG__llm__name: "qwen-turbo"
+          PAIRAG_RAG__llm__name: "qwen-max"

.github/workflows/main.yml

@@ -52,7 +52,7 @@ jobs:
           IS_PAI_RAG_CI_TEST: true
           PAIRAG_RAG__embedding__source: "DashScope"
           PAIRAG_RAG__llm__source: "DashScope"
-          PAIRAG_RAG__llm__name: "qwen-turbo"
+          PAIRAG_RAG__llm__model: "qwen-max"
 
       - name: Get Cover
         uses: orgoro/[email protected]

.github/workflows/main_gpu.yml

@@ -60,4 +60,4 @@ jobs:
           IS_PAI_RAG_CI_TEST: true
           PAIRAG_RAG__embedding__source: "DashScope"
           PAIRAG_RAG__llm__source: "DashScope"
-          PAIRAG_RAG__llm__name: "qwen-turbo"
+          PAIRAG_RAG__llm__model: "qwen-max"

example_data/function_tools/api-tool-with-intent-detection-for-travel-assistant/settings.toml

@@ -49,7 +49,7 @@ vector_store.type = "FAISS"
 # token = ""
 [rag.llm]
 source = "DashScope"
-model = "qwen-turbo"
+model = "qwen-max"
 
 [rag.llm.function_calling_llm]
 source = "DashScope"

src/pai_rag/app/api/query.py

@@ -145,14 +145,15 @@ def task_status(task_id: str):
 async def upload_data(
     files: List[UploadFile] = Body(None),
     oss_path: str = Form(None),
-    faiss_path: str = Form(None),
     index_name: str = Form(None),
     enable_raptor: bool = Form(False),
     enable_multimodal: bool = Form(False),
     background_tasks: BackgroundTasks = BackgroundTasks(),
 ):
     task_id = uuid.uuid4().hex
-
+    logger.info(
+        f"Upload data task_id: {task_id} index_name: {index_name} enable_multimodal: {enable_multimodal}"
+    )
     if oss_path:
         background_tasks.add_task(
             rag_service.add_knowledge,
@@ -161,7 +162,6 @@ async def upload_data(
             oss_path=oss_path,
             from_oss=True,
             index_name=index_name,
-            faiss_path=faiss_path,
             enable_raptor=enable_raptor,
             enable_multimodal=enable_multimodal,
         )
@@ -194,6 +194,7 @@ async def upload_data(
             oss_path=None,
             enable_raptor=enable_raptor,
             temp_file_dir=tmpdir,
+            enable_multimodal=enable_multimodal,
         )
 
     return {"task_id": task_id}

src/pai_rag/app/web/tabs/settings_tab.py

@@ -191,14 +191,16 @@ def create_setting_tab() -> Dict[str, Any]:
                     elem_id="use_mllm",
                     container=False,
                 )
-                with gr.Row(visible=False) as use_mllm_col:
+                with gr.Row(visible=False, elem_id="use_mllm_col") as use_mllm_col:
                     mllm = gr.Radio(
                         ["paieas", "dashscope"],
                         label="LLM Model Source",
                         elem_id="mllm",
                         interactive=DEFAULT_IS_INTERACTIVE.lower() != "false",
                     )
-                    with gr.Row(visible=(mllm == "paieas")) as m_eas_col:
+                    with gr.Row(
+                        visible=(mllm == "paieas"), elem_id="m_eas_col"
+                    ) as m_eas_col:
                         mllm_eas_url = gr.Textbox(
                             label="EAS Url",
                             elem_id="mllm_eas_url",
@@ -216,7 +218,9 @@ def create_setting_tab() -> Dict[str, Any]:
                             elem_id="mllm_eas_model_name",
                             interactive=True,
                         )
-                    with gr.Row(visible=(mllm == "dashscope")) as api_mllm_col:
+                    with gr.Row(
+                        visible=(mllm == "dashscope"), elem_id="api_mllm_col"
+                    ) as api_mllm_col:
                         mllm_api_model_name = gr.Dropdown(
                             label="LLM Model Name",
                             elem_id="mllm_api_model_name",
@@ -231,7 +235,11 @@ def create_setting_tab() -> Dict[str, Any]:
                     elem_id="use_oss",
                     container=False,
                 )
-                with gr.Row(visible=False) as use_oss_col:
+                with gr.Row(visible=False, elem_id="use_oss_col") as use_oss_col:
+                    oss_bucket = gr.Textbox(
+                        label="OSS Bucket",
+                        elem_id="oss_bucket",
+                    )
                     oss_ak = gr.Textbox(
                         label="Access Key",
                         elem_id="oss_ak",
@@ -245,10 +253,7 @@ def create_setting_tab() -> Dict[str, Any]:
                     oss_endpoint = gr.Textbox(
                         label="OSS Endpoint",
                         elem_id="oss_endpoint",
-                    )
-                    oss_bucket = gr.Textbox(
-                        label="OSS Bucket",
-                        elem_id="oss_bucket",
+                        default="oss-cn-hangzhou.aliyuncs.com",
                     )
                 use_oss.input(
                     fn=ev_listeners.change_use_oss,
@@ -312,8 +317,10 @@ def create_setting_tab() -> Dict[str, Any]:
     elems.update(vector_db_components)
     elems.update(
         {
-            "use_oss_col": use_oss_col,
-            "use_mllm_col": use_mllm_col,
+            m_eas_col.elem_id: m_eas_col,
+            api_mllm_col.elem_id: api_mllm_col,
+            use_oss_col.elem_id: use_oss_col,
+            use_mllm_col.elem_id: use_mllm_col,
         }
     )
     return elems

src/pai_rag/app/web/tabs/upload_tab.py

@@ -109,6 +109,7 @@ def upload_knowledge(
             input_files=[file.name for file in upload_files],
             enable_raptor=enable_raptor,
             index_name=index_name,
+            enable_multimodal=enable_multimodal,
         )
         for file in upload_files:
             base_name = os.path.basename(file.name)

src/pai_rag/app/web/view_model.py

@@ -263,7 +263,7 @@ def to_app_config(self):
 
         config["retriever"]["image_similarity_top_k"] = self.image_similarity_top_k
 
-        config["retriever"]["need_image"] = self.need_image
+        config["retriever"]["search_image"] = self.need_image
         if self.retrieval_mode == "Hybrid":
             config["retriever"]["vector_store_query_mode"] = VectorStoreQueryMode.HYBRID
         elif self.retrieval_mode == "Embedding Only":
@@ -432,6 +432,8 @@ def to_component_settings(self) -> Dict[str, Dict[str, Any]]:
             "choices": MLLM_MODEL_KEY_DICT.get(self.mllm, []),
             "visible": self.mllm.lower() != "paieas",
         }
+        settings["m_eas_col"] = {"visible": self.mllm == "paieas"}
+        settings["api_mllm_col"] = {"visible": self.mllm == "dashscope"}
 
         settings["use_oss"] = {"value": self.use_oss}
         settings["use_oss_col"] = {"visible": self.use_oss}

src/pai_rag/app/web/webui.py

@@ -8,7 +8,6 @@
 from pai_rag.app.web.tabs.settings_tab import create_setting_tab
 from pai_rag.app.web.tabs.upload_tab import create_upload_tab
 from pai_rag.app.web.tabs.chat_tab import create_chat_tab
-from pai_rag.app.web.tabs.agent_tab import create_agent_tab
 from pai_rag.app.web.tabs.data_analysis_tab import create_data_analysis_tab
 from pai_rag.app.web.index_utils import index_related_component_keys
 
@@ -84,9 +83,11 @@ def make_homepage():
         with gr.Tab("\N{fire} Chat"):
             chat_elements = create_chat_tab()
             elem_manager.add_elems(chat_elements)
+        """ hide agent tab
         with gr.Tab("\N{rocket} Agent"):
             agent_elements = create_agent_tab()
             elem_manager.add_elems(agent_elements)
+        """
         with gr.Tab("\N{bar chart} Data Analysis"):
             analysis_elements = create_data_analysis_tab()
             elem_manager.add_elems(analysis_elements)

src/pai_rag/config/settings.toml

@@ -44,7 +44,7 @@ vector_store.type = "FAISS"
 # token = ""
 [rag.llm]
 source = "DashScope"
-model = "qwen-turbo"
+model = "qwen-max"
 
 [rag.multimodal_embedding]
 source = "cnclip"

src/pai_rag/core/rag_application.py

@@ -32,6 +32,7 @@
 
 DEFAULT_EMPTY_RESPONSE_GEN = "Empty Response"
 DEFAULT_RAG_INDEX_FILE = "localdata/default_rag_indexes.json"
+logger = logging.getLogger(__name__)
 
 
 def uuid_generator() -> str:
@@ -89,11 +90,20 @@ def load_knowledge(
         enable_raptor=False,
         enable_multimodal=False,
     ):
+        logger.info(
+            f"""Loading data:
+            input_files: {input_files}
+            index_name: {index_name}
+            enable_multimodal: {enable_multimodal}
+            enable_raptor: {enable_raptor}"""
+        )
+
         session_config = self.config.model_copy()
         index_entry = index_manager.get_index_by_name(index_name)
         session_config.embedding = index_entry.embedding_config
         session_config.index.vector_store = index_entry.vector_store_config
         session_config.node_parser.enable_multimodal = enable_multimodal
+
         data_loader = resolve_data_loader(session_config)
         data_loader.load_data(
             file_path_or_directory=input_files,

src/pai_rag/core/rag_config.py

@@ -22,7 +22,11 @@
     PaiBaseEmbeddingConfig,
 )
 from pai_rag.integrations.index.pai.vector_store_config import PaiVectorIndexConfig
-from pai_rag.integrations.llms.pai.llm_config import PaiBaseLlmConfig
+from pai_rag.integrations.llms.pai.llm_config import (
+    DashScopeMultiModalLlmConfig,
+    PaiBaseLlmConfig,
+    PaiEasLlmConfig,
+)
 from pai_rag.integrations.nodeparsers.pai.pai_node_parser import NodeParserConfig
 from pai_rag.integrations.postprocessor.pai.pai_postprocessor import (
     RerankModelPostProcessorConfig,
@@ -75,7 +79,7 @@ class RagConfig(BaseModel):
         BeforeValidator(validate_case_insensitive),
     ]
     multimodal_llm: Annotated[
-        Union[PaiBaseLlmConfig.get_subclasses()],
+        Union[DashScopeMultiModalLlmConfig, PaiEasLlmConfig],
         Field(discriminator="source"),
         BeforeValidator(validate_case_insensitive),
     ] | None = None

src/pai_rag/core/rag_module.py

@@ -40,8 +40,6 @@
 def resolve(cls: Any, **kwargs):
     cls_key = kwargs.__repr__()
     if cls_key not in cls_cache:
-        print("New cls_key", cls_key)
-
         cls_cache[cls_key] = cls(**kwargs)
     return cls_cache[cls_key]
 

src/pai_rag/core/rag_service.py

@@ -55,6 +55,7 @@ def get_config(self):
     def reload(self, new_config: Dict):
         self.rag_configuration.update(new_config)
         self.rag.refresh(self.rag_configuration.get_value())
+        self.rag_configuration.persist()
 
     def add_knowledge(
         self,

src/pai_rag/integrations/index/pai/vector_store_config.py

@@ -115,7 +115,7 @@ class PaiVectorIndexConfig(BaseModel):
     vector_store: Annotated[
         Union[BaseVectorStoreConfig.get_subclasses()], Field(discriminator="type")
     ]
-    enable_multimodal: bool = False
+    enable_multimodal: bool = True  # default enable multimodal
     persist_path: str = DEFAULT_LOCAL_STORAGE_PATH
 
 

src/pai_rag/integrations/llms/pai/llm_config.py

@@ -144,7 +144,7 @@ class DashScopeLlmConfig(PaiBaseLlmConfig):
     source: Literal[SupportedLlmType.dashscope] = SupportedLlmType.dashscope
     api_key: str | None = None
     base_url: str = "https://dashscope.aliyuncs.com/compatible-mode/v1"
-    model: str = "qwen-turbo"
+    model: str = "qwen-max"
 
 
 class OpenAILlmConfig(PaiBaseLlmConfig):
@@ -160,6 +160,10 @@ class PaiEasLlmConfig(PaiBaseLlmConfig):
     model: str = "default"
 
 
+class DashScopeMultiModalLlmConfig(DashScopeLlmConfig):
+    model: str = "qwen-vl-max"
+
+
 SupporttedLlmClsMap = {cls.get_type(): cls for cls in PaiBaseLlmConfig.get_subclasses()}