update

.gitignore

@@ -1 +1,7 @@
-local
+local
+local-new
+data
+examples
+README_CN.md
+**/__pycache__
+*.pyc

src/config.yaml

@@ -16,5 +16,5 @@ agent:
       extraction_agent: extract_information_with_case
       reflection_agent: reflect_with_case
     customized:
-      schema_agent:  get_retrieved_schema
+      schema_agent: get_retrieved_schema
       extraction_agent: extract_information_direct

src/modules/knowledge_base/schema_repository.py

@@ -2,18 +2,18 @@ from typing import List, Optional
 from pydantic import BaseModel, Field
 from langchain_core.output_parsers import JsonOutputParser
 
-# ==================================================================== #  
-#                                NER TASK                              #  
-# ==================================================================== #  
+# ==================================================================== #
+#                                NER TASK                              #
+# ==================================================================== #
 class Entity(BaseModel):
     name : str = Field(description="The specific name of the entity. ")
     type : str = Field(description="The type or category that the entity belongs to.")
 class EntityList(BaseModel):
     entity_list : List[Entity] = Field(description="Named entities appearing in the text.")
-    
-# ==================================================================== #  
-#                               RE TASK                                #  
-# ==================================================================== #  
+
+# ==================================================================== #
+#                               RE TASK                                #
+# ==================================================================== #
 class Relation(BaseModel):
     head : str = Field(description="The starting entity in the relationship.")
     tail : str = Field(description="The ending entity in the relationship.")
@@ -22,9 +22,9 @@ class Relation(BaseModel):
 class RelationList(BaseModel):
     relation_list : List[Relation] = Field(description="The collection of relationships between various entities.")
 
-# ==================================================================== #  
-#                               EE TASK                                #  
-# ==================================================================== #  
+# ==================================================================== #
+#                               EE TASK                                #
+# ==================================================================== #
 class Event(BaseModel):
     event_type : str = Field(description="The type of the event.")
     event_trigger : str = Field(description="A specific word or phrase that indicates the occurrence of the event.")
@@ -33,38 +33,38 @@ class Event(BaseModel):
 class EventList(BaseModel):
     event_list : List[Event] = Field(description="The events presented in the text.")
 
-# ==================================================================== #  
-#                          TEXT DESCRIPTION                            #  
-# ==================================================================== #  
+# ==================================================================== #
+#                          TEXT DESCRIPTION                            #
+# ==================================================================== #
 class TextDescription(BaseModel):
     field: str = Field(description="The field of the given text, such as 'Science', 'Literature', 'Business', 'Medicine', 'Entertainment', etc.")
     genre: str = Field(description="The genre of the given text, such as 'Article', 'Novel', 'Dialog', 'Blog', 'Manual','Expository', 'News Report', 'Research Paper', etc.")
-    
-# ==================================================================== #  
-#                        USER DEFINED SCHEMA                           #  
-# ==================================================================== #  
+
+# ==================================================================== #
+#                        USER DEFINED SCHEMA                           #
+# ==================================================================== #
 
 # --------------------------- Research Paper ----------------------- #
 class MetaData(BaseModel):
     title : str = Field(description="The title of the article")
     authors : List[str] = Field(description="The list of the article's authors")
-    abstract: str = Field(description="The article's abstract") 
+    abstract: str = Field(description="The article's abstract")
     key_words: List[str] = Field(description="The key words associated with the article")
-    
+
 class Baseline(BaseModel):
     method_name : str = Field(description="The name of the baseline method")
     proposed_solution : str = Field(description="the proposed solution in details")
     performance_metrics : str = Field(description="The performance metrics of the method and comparative analysis")
-    
+
 class ExtractionTarget(BaseModel):
-    
+
     key_contributions: List[str] = Field(description="The key contributions of the article")
     limitation_of_sota : str=Field(description="the summary limitation of the existing work")
     proposed_solution : str = Field(description="the proposed solution in details")
     baselines : List[Baseline] = Field(description="The list of baseline methods and their details")
     performance_metrics : str = Field(description="The performance metrics of the method and comparative analysis")
     paper_limitations : str=Field(description="The limitations of the proposed solution of the paper")
-    
+
 # --------------------------- News ----------------------- #
 class Person(BaseModel):
     name: str = Field(description="The name of the person")
@@ -79,7 +79,7 @@ class Event(BaseModel):
     process: Optional[str] = Field(description="Details of the event process")
     result: Optional[str] = Field(default=None, description="Result or outcome of the event")
 
-class NewsReport(BaseModel):  
+class NewsReport(BaseModel):
     title: str = Field(description="The title or headline of the news report")
     summary: str = Field(description="A brief summary of the news report")
     publication_date: Optional[str] = Field(description="The publication date of the report")
@@ -88,4 +88,13 @@ class NewsReport(BaseModel):
     quotes: Optional[dict] = Field(default=None, description="Quotes related to the news, with keys as the citation sources and values as the quoted content. ")
     viewpoints: Optional[List[str]] = Field(default=None, description="Different viewpoints regarding the news")
 
-# --------- You can customize new extraction schemas below -------- #
+# --------- You can customize new extraction schemas below -------- #
+class ChemicalSubstance(BaseModel):
+    name: str = Field(description="Name of the chemical substance")
+    formula: str = Field(description="Molecular formula")
+    appearance: str = Field(description="Physical appearance")
+    uses: List[str] = Field(description="Primary uses")
+    hazards: str = Field(description="Hazard classification")
+
+class ChemicalList(BaseModel):
+  chemicals: List[ChemicalSubstance] = Field(description="List of chemicals")

src/pipeline.py

@@ -48,6 +48,7 @@ class Pipeline:
 
     # main entry
     def get_extract_result(self,
+                           three_agents,
                            task: TaskType,
                            instruction: str = "",
                            text: str = "",
@@ -58,10 +59,9 @@ class Pipeline:
                            truth: str = "",
                            mode: str = "quick",
                            update_case: bool = False,
-                           show_trajectory: bool = False
+                           show_trajectory: bool = False,
+                           isgui: bool = False,
                            ):
-        print(f" task: {task},\n instruction: {instruction},\n text: {text},\n output_schema: {output_schema},\n constraint: {constraint},\n use_file: {use_file},\n file_path: {file_path},\n truth: {truth},\n mode: {mode},\n update_case: {update_case}")
-
         # Check Consistancy
         mode, update_case = self.__check_consistancy(self.llm, task, mode, update_case)
 
@@ -72,6 +72,11 @@ class Pipeline:
             process_method = config['agent']['mode'][mode].copy()
         else:
             process_method = mode
+
+        if isgui and mode == "customized":
+            process_method = three_agents
+            # print("Customized Agents: ", three_agents)
+
         sorted_process_method = self.__init_method(data, process_method)
         print("Process Method: ", sorted_process_method)
 

src/webui.py

@@ -9,44 +9,54 @@ from models import *
 examples = [
     {
         "task": "NER",
+        "mode": "quick",
         "use_file": False,
         "text": "Finally, every other year , ELRA organizes a major conference LREC , the International Language Resources and Evaluation Conference .",
         "instruction": "",
         "constraint": """["nationality", "country capital", "place of death", "children", "location contains", "place of birth", "place lived", "administrative division of country", "country of administrative divisions", "company", "neighborhood of", "company founders"]""",
         "file_path": None,
+        "update_case": False,
     },
     {
         "task": "RE",
+        "mode": "quick",
         "use_file": False,
         "text": "The aid group Doctors Without Borders said that since Saturday , more than 275 wounded people had been admitted and treated at Donka Hospital in the capital of Guinea , Conakry .",
         "instruction": "",
         "constraint": """["nationality", "country capital", "place of death", "children", "location contains", "place of birth", "place lived", "administrative division of country", "country of administrative divisions", "company", "neighborhood of", "company founders"]""",
         "file_path": None,
+        "update_case": False,
     },
-        {
+    {
         "task": "EE",
+        "mode": "quick",
         "use_file": False,
         "text": "The file suggested to the user contains no software related to video streaming and simply carries the malicious payload that later compromises victim \u2019s account and sends out the deceptive messages to all victim \u2019s contacts .",
         "instruction": "",
         "constraint": """{"phishing": ["damage amount", "attack pattern", "tool", "victim", "place", "attacker", "purpose", "trusted entity", "time"], "data breach": ["damage amount", "attack pattern", "number of data", "number of victim", "tool", "compromised data", "victim", "place", "attacker", "purpose", "time"], "ransom": ["damage amount", "attack pattern", "payment method", "tool", "victim", "place", "attacker", "price", "time"], "discover vulnerability": ["vulnerable system", "vulnerability", "vulnerable system owner", "vulnerable system version", "supported platform", "common vulnerabilities and exposures", "capabilities", "time", "discoverer"], "patch vulnerability": ["vulnerable system", "vulnerability", "issues addressed", "vulnerable system version", "releaser", "supported platform", "common vulnerabilities and exposures", "patch number", "time", "patch"]}""",
         "file_path": None,
+        "update_case": False,
+    },
+    {
+        "task": "Base",
+        "mode": "quick",
+        "use_file": True,
+        "file_path": "data/Harry_Potter_Chapter1.pdf",
+        "instruction": "Extract main characters and the background setting from this chapter.",
+        "constraint": "",
+        "text": "",
+        "update_case": False,
+    },
+    {
+        "task": "Base",
+        "mode": "quick",
+        "use_file": True,
+        "file_path": "data/Tulsi_Gabbard_News.html",
+        "instruction": "Extract key information from the given text.",
+        "constraint": "",
+        "text": "",
+        "update_case": False,
     },
-    # {
-    #     "task": "Base",
-    #     "use_file": True,
-    #     "file_path": "data/Harry_Potter_Chapter_1.pdf",
-    #     "instruction": "Extract main characters and the background setting from this chapter.",
-    #     "constraint": "",
-    #     "text": "",
-    # },
-    # {
-    #     "task": "Base",
-    #     "use_file": True,
-    #     "file_path": "data/Tulsi_Gabbard_News.html",
-    #     "instruction": "Extract key information from the given text.",
-    #     "constraint": "",
-    #     "text": "",
-    # },
 ]
 
 
@@ -73,18 +83,29 @@ def create_interface():
 
         with gr.Row():
             with gr.Column():
-                model_gr = gr.Dropdown(choices=["deepseek-chat", "deepseek-reasoner", "gpt-3.5-turbo", "gpt-4o", "gpt-4o-mini"], label="🤖 Select your Model")
+                model_gr = gr.Dropdown(choices=["deepseek-chat", "deepseek-reasoner", "gpt-3.5-turbo", "gpt-4o", "gpt-4o-mini"], label="🪄 Select your Model")
                 api_key_gr = gr.Textbox(label="🔑 Enter your API-Key")
                 base_url_gr = gr.Textbox(label="🔗 Enter your Base-URL", value="DEFAULT")
             with gr.Column():
                 task_gr = gr.Dropdown(choices=["Base", "NER", "RE", "EE"], label="🎯 Select your Task")
-                use_file_gr = gr.Checkbox(label="📂 Use File", value=True)
+                mode_gr = gr.Dropdown(choices=["quick", "standard", "customized"], label="🧭 Select your Mode")
+                schema_agent_gr = gr.Dropdown(choices=["NOT REQUIRED", "get_default_schema", "get_retrieved_schema", "get_deduced_schema"], label="🤖 Select your Schema-Agent", visible=False)
+                extraction_Agent_gr = gr.Dropdown(choices=["NOT REQUIRED", "extract_information_direct", "extract_information_with_case"], label="🤖 Select your Extraction-Agent", visible=False)
+                reflection_agent_gr = gr.Dropdown(choices=["NOT REQUIRED", "reflect_with_case"], label="🤖 Select your Reflection-Agent", visible=False)
+
+        with gr.Row():
+            use_file_gr = gr.Checkbox(label="📂 Use File", value=True)
+            update_case_gr = gr.Checkbox(label="💰 Update Case", value=False)
 
         file_path_gr = gr.File(label="📖 Upload a File", visible=True)
         text_gr = gr.Textbox(label="📖 Text", placeholder="Enter your Text", visible=False)
         instruction_gr = gr.Textbox(label="🕹️ Instruction", visible=True)
         constraint_gr = gr.Textbox(label="🕹️ Constraint", visible=False)
 
+        def customized_mode(mode):
+            if mode == "customized":
+                return gr.update(visible=True), gr.update(visible=True), gr.update(visible=True)
+
         def update_fields(task):
             if task == "Base":
                 return gr.update(visible=True, label="🕹️ Instruction", placeholder="Enter your Instruction"), gr.update(visible=False)
@@ -106,14 +127,19 @@ def create_interface():
             example = examples[example_index]
             return (
                 gr.update(value=example["task"]),
+                gr.update(value=example["mode"]),
                 gr.update(value=example["use_file"]),
                 gr.update(value=example["file_path"], visible=example["use_file"]),
                 gr.update(value=example["text"], visible=not example["use_file"]),
                 gr.update(value=example["instruction"], visible=example["task"] == "Base"),
                 gr.update(value=example["constraint"], visible=example["task"] in ["NER", "RE", "EE"]),
+                gr.update(value=example["update_case"]),
+                gr.update(value="NOT REQUIRED", visible=False),
+                gr.update(value="NOT REQUIRED", visible=False),
+                gr.update(value="NOT REQUIRED", visible=False),
             )
 
-        def submit(model, api_key, base_url, task, instruction, constraint, text, use_file, file_path):
+        def submit(model, api_key, base_url, task, mode, instruction, constraint, text, use_file, file_path, update_case, schema_agent, extraction_Agent, reflection_agent):
             try:
                 if base_url == "DEFAULT":
                     if model in ["gpt-3.5-turbo", "gpt-4o", "gpt-4o-mini"]:
@@ -139,14 +165,29 @@ def create_interface():
                     text = text
                     file_path = None
 
+                agent3 = {}
+                if mode == "customized":
+                    if schema_agent not in ["", "NOT REQUIRED"]:
+                        agent3["schema_agent"] = schema_agent
+                    if extraction_Agent not in ["", "NOT REQUIRED"]:
+                        agent3["extraction_agent"] = extraction_Agent
+                    if reflection_agent not in ["", "NOT REQUIRED"]:
+                        agent3["reflection_agent"] = reflection_agent
+
                 # 调用 Pipeline
                 _, _, ger_frontend_schema, ger_frontend_res = pipeline.get_extract_result(
                     task=task,
-                    instruction=instruction,
-                    constraint=constraint,
+                    text=text,
                     use_file=use_file,
                     file_path=file_path,
-                    text=text,
+                    instruction=instruction,
+                    constraint=constraint,
+                    mode=mode,
+                    three_agents=agent3,
+                    isgui=True,
+                    update_case=update_case,
+                    output_schema="",
+                    truth="",
                     show_trajectory=False,
                 )
 
@@ -160,12 +201,17 @@ def create_interface():
 
         def clear_all():
             return (
+                gr.update(value="", visible=False),  # sechema_agent
+                gr.update(value="", visible=False),  # extraction_Agent
+                gr.update(value="", visible=False),  # reflection_agent
                 gr.update(value=""),  # task
+                gr.update(value=""),  # mode
                 gr.update(value="", visible=False),  # instruction
                 gr.update(value="", visible=False),  # constraint
                 gr.update(value=True),  # use_file
                 gr.update(value="", visible=False),  # text
                 gr.update(value=None, visible=True),  # file_path
+                gr.update(value=False),  # update_case
                 gr.update(value=""),
                 gr.update(value=""),
                 gr.update(value="", visible=False),  # error_output
@@ -189,6 +235,7 @@ def create_interface():
                 json_output_gr = gr.Code(label="😉 Final Answer", language="json", lines=10, interactive=False)
 
         task_gr.change(fn=update_fields, inputs=task_gr, outputs=[instruction_gr, constraint_gr])
+        mode_gr.change(fn=customized_mode, inputs=mode_gr, outputs=[schema_agent_gr, extraction_Agent_gr, reflection_agent_gr])
         use_file_gr.change(fn=update_input_fields, inputs=use_file_gr, outputs=[text_gr, file_path_gr])
 
         example_button_gr.click(
@@ -196,11 +243,16 @@ def create_interface():
             inputs=[],
             outputs=[
                 task_gr,
+                mode_gr,
                 use_file_gr,
                 file_path_gr,
                 text_gr,
                 instruction_gr,
                 constraint_gr,
+                update_case_gr,
+                schema_agent_gr,
+                extraction_Agent_gr,
+                reflection_agent_gr,
             ],
         )
         submit_button_gr.click(
@@ -210,11 +262,16 @@ def create_interface():
                 api_key_gr,
                 base_url_gr,
                 task_gr,
+                mode_gr,
                 instruction_gr,
                 constraint_gr,
                 text_gr,
                 use_file_gr,
                 file_path_gr,
+                update_case_gr,
+                schema_agent_gr,
+                extraction_Agent_gr,
+                reflection_agent_gr,
             ],
             outputs=[py_output_gr, json_output_gr, error_output_gr],
             show_progress=True,
@@ -222,12 +279,17 @@ def create_interface():
         clear_button.click(
             fn=clear_all,
             outputs=[
+                schema_agent_gr,
+                extraction_Agent_gr,
+                reflection_agent_gr,
                 task_gr,
+                mode_gr,
                 instruction_gr,
                 constraint_gr,
                 use_file_gr,
                 text_gr,
                 file_path_gr,
+                update_case_gr,
                 py_output_gr,
                 json_output_gr,
                 error_output_gr,