xgae 0.2.1__tar.gz → 0.2.3__tar.gz

{xgae-0.2.1 → xgae-0.2.3}/CHANGELOG.md

@@ -1,10 +1,15 @@
-# Release Changelog
+## [0.2.3] - 2025-9-19
+### Modified
+- CustomPromptRag: remove FastEmbedEmbeddings, use 'text-embedding-v3' model for chinese, avoid download 'bge-small-zh-v1.5'
+
+
 ## [0.2.1] - 2025-9-17
 ### Added
-- Example Langgraph ReactAgent: add Chromadb for custom prompt RAG 
+- Example ReactAgent: add CustomPromptRag, use FastEmbedEmbeddings and 'BAAI/bge-small-zh-v1.5' model
 ### Modified
 - pyproject.toml: add [project.optional-dependencies] 'examples'
 
+
 ## [0.2.0] - 2025-9-10
 ### Added
 - Agent Engine release 0.2

{xgae-0.2.1 → xgae-0.2.3}/PKG-INFO

@@ -1,16 +1,14 @@
 Metadata-Version: 2.4
 Name: xgae
-Version: 0.2.1
+Version: 0.2.3
 Summary: Extreme General Agent Engine
 Requires-Python: >=3.13
 Requires-Dist: colorlog==6.9.0
 Requires-Dist: langchain-mcp-adapters==0.1.9
-Requires-Dist: langchain==0.3.27
 Requires-Dist: langfuse==2.60.9
 Requires-Dist: litellm==1.74.15
 Requires-Dist: mcp==1.13.0
 Provides-Extra: examples
 Requires-Dist: chromadb==1.1.0; extra == 'examples'
-Requires-Dist: fastembed==0.7.3; extra == 'examples'
 Requires-Dist: langchain-community==0.3.29; extra == 'examples'
 Requires-Dist: langgraph==0.6.5; extra == 'examples'

{xgae-0.2.1 → xgae-0.2.3}/pyproject.toml

@@ -1,33 +1,35 @@
 [project]
 name = "xgae"
-version = "0.2.1"
+version = "0.2.3"
 description = "Extreme General Agent Engine"
 readme = "README.md"
 requires-python = ">=3.13"
 dependencies = [
     "colorlog==6.9.0",
-    "langchain-mcp-adapters==0.1.9",
     "litellm==1.74.15",
     "mcp==1.13.0",
     "langfuse==2.60.9",
-    "langchain==0.3.27",
+    "langchain-mcp-adapters==0.1.9",
 ]
 
+
 [project.optional-dependencies]
 examples = [
     "langgraph==0.6.5",
     "langchain-community==0.3.29",
-    "fastembed==0.7.3",
     "chromadb==1.1.0",
 ]
 
+
 [build-system]
 requires = ["hatchling"]
 build-backend = "hatchling.build"
 
+
 [tool.hatch.build]
 exclude = ["log/*", ".idea/*"]
 
+
 [project.scripts]
 xgae = "xgae.engine_cli_app:main"
 xgae-tools = "xgae.tools.without_general_tools_app:main"

xgae-0.2.3/src/examples/agent/langgraph/react/custom_prompt_rag.py

@@ -0,0 +1,107 @@
+import logging
+import os
+from typing import override, List
+
+from langchain_core.documents import Document
+from langchain_core.embeddings import Embeddings
+from langchain_core.vectorstores import VectorStore
+
+from langchain_community.vectorstores import Chroma
+from chromadb.utils.embedding_functions import OpenAIEmbeddingFunction
+
+class ChromaEmbedding(Embeddings):
+    embedding_model_name = "text-embedding-v3"
+
+    def __init__(self):
+        api_key = os.getenv('LLM_API_KEY')
+        api_base = os.getenv('LLM_API_BASE', "https://dashscope.aliyuncs.com/compatible-mode/v1")
+
+        self.embedding_function = OpenAIEmbeddingFunction(
+            api_key = api_key,
+            api_base = api_base,
+            model_name = self.embedding_model_name,
+        )
+
+
+    @override
+    def embed_documents(self, texts: List[str]) -> List[List[float]]:
+        return self.embedding_function(texts)
+
+
+    @override
+    def embed_query(self, text: str) -> List[float]:
+        return self.embed_documents([text])[0]
+
+
+class CustomPromptRag:
+    PROMPT_SIMILAR_SCORE = 0.85   # need tune-up score based on different embedding model
+
+    def __init__(self):
+        prompt_docs = self._load_prompts()
+        self.vector_store = self._init_vector_store(prompt_docs)
+
+
+    # should read from DB, load all custom prompt or COT
+    def _load_prompts(self) -> List[Document]:
+        prompt_docs = []
+        prompt_docs.append(self._create_prompt_doc(
+            prompt_summary="Fault location and analysis of fault causes",
+            prompt_path="templates/example/fault_user_prompt.txt"
+        ))
+        return prompt_docs
+
+
+    def _create_prompt_doc(self, prompt_summary: str, prompt_path: str)-> Document:
+        return Document(
+            page_content=prompt_summary,
+            metadata={
+                "source": prompt_path,
+            }
+        )
+
+
+    def _init_vector_store(self, docs: List[Document]) -> VectorStore:
+        embeddings = ChromaEmbedding()
+        return Chroma.from_documents(
+            documents=docs,
+            embedding=embeddings,
+            persist_directory=None
+        )
+
+
+    def search_prompt(self, query:str)-> str:
+        prompt_path = None
+        results = self.vector_store.similarity_search_with_score(query=query, k=1)
+        if results and len(results) > 0:
+            doc, score = results[0]
+            if score > self.PROMPT_SIMILAR_SCORE:
+                logging.info(f"CustomPromptRag search: SIMILAR_SCORE: {score} > {self.PROMPT_SIMILAR_SCORE}, "
+                             f"\nquery: '{query}' \nprompt_summary: '{doc.page_content}'\n")
+            else:
+                prompt_path = doc.metadata['source']
+                logging.info(f"CustomPromptRag search: SIMILAR_SCORE: {score}, prompt_path: '{prompt_path}'")
+
+        return prompt_path
+
+
+if __name__ == "__main__":
+    from xgae.utils.setup_env import setup_logging
+
+    setup_logging()
+
+    custom_prompt_rag = CustomPromptRag()
+
+    querys = ["locate 10.2.3.4 fault and solution", # 0.79
+              "定位 10.2.3.4 故障,并给出解决方案",      # 0.81
+              "locate fault and solution",          # 0.42
+              "locate fault",                       # 0.40
+              "定位故障",                            # 0.64
+              "fault solution",                     # 0.47
+              "locate",                             # 0.95
+              "5+7"                                 # 1.12
+              ]
+
+    for query in querys:
+        logging.info("*"*50)
+        logging.info(f"query: '{query}'")
+        custom_prompt_rag.search_prompt(query)

{xgae-0.2.1 → xgae-0.2.3}/src/examples/agent/langgraph/react/react_agent.py

@@ -7,7 +7,6 @@ from uuid import uuid4
 from langfuse.callback import CallbackHandler
 from langfuse import Langfuse
 
-
 from langgraph.graph import END, START, StateGraph
 from langgraph.types import interrupt, Command
 from langgraph.checkpoint.memory import MemorySaver
@@ -24,6 +23,7 @@ from examples.agent.langgraph.react.agent_base import AgentContext, TaskState, E
 from examples.agent.langgraph.react.result_eval_agent import TaskResultEvalAgent
 from examples.agent.langgraph.react.custom_prompt_rag import CustomPromptRag
 
+
 class XGAReactAgent:
     MAX_TASK_RETRY = 2
     QUALIFIED_RESULT_SCORE = 0.7
@@ -89,18 +89,24 @@ class XGAReactAgent:
 
 
     def _search_system_prompt(self, user_input: str) -> str:
+        system_prompt = None
         if hasattr(self, 'custom_prompt_rag'):
-            system_prompt = self.custom_prompt_rag.search_prompt(user_input)
+            system_prompt_path = self.custom_prompt_rag.search_prompt(user_input)
+            if system_prompt_path:
+                system_prompt = read_file(system_prompt_path)
         else:
-            system_prompt = None if "fault" not in user_input else read_file("templates/example/fault_user_prompt.txt")
+            if "fault" in user_input:  # only for example
+                system_prompt = read_file("templates/example/fault_user_prompt.txt")
         return system_prompt
 
 
     async def _supervisor_node(self, state: TaskState) -> Dict[str, Any]:
         user_input = state['user_inputs'][0]
         eval_result = state.get('eval_result', None)
+        system_prompt = state.get('system_prompt', None)
 
-        system_prompt = self._search_system_prompt(user_input)
+        if system_prompt is  None and eval_result is None:
+            system_prompt = self._search_system_prompt(user_input)
         is_system_prompt = True if system_prompt is not None else False
 
         general_tools = [] if system_prompt else ["*"]

{xgae-0.2.1 → xgae-0.2.3}/src/examples/agent/langgraph/react/run_react_agent.py

@@ -17,7 +17,7 @@ async def main():
     ]
 
     for user_input in user_inputs:
-        agent = XGAReactAgent(use_prompt_rag=False)
+        agent = XGAReactAgent(use_prompt_rag=True)
         task_no += 1
         context: AgentContext = {
             'task_id': f"agent_task_{uuid4()}",   # can be set with request_id, must be unique

xgae-0.2.3/test/test_chroma.py

@@ -0,0 +1,31 @@
+import chromadb
+
+chroma_client = chromadb.EphemeralClient()
+
+# default use 'ONNXMiniLM_L6_V2' embedding function
+collection = chroma_client.get_or_create_collection(name="fault_collection")
+
+collection.upsert(
+    documents=[
+        "Fault location and analysis of fault causes",
+        "This is a analyse about alarm"
+    ],
+    ids=["id1", "id2"],
+    metadatas=[{"type": 1}, {"type": 2}]
+)
+
+querys = ["locate fault and solution",
+          "alarm search"
+          ]
+
+for query in querys:
+    results = collection.query(
+        query_texts=query,
+        n_results=2,
+        #where={"type": 1},
+        where={"type": {"$gt": 0}},
+        where_document = {"$contains": "fault"}
+    )
+    print("*"*50)
+    print(f"query='{query}', id={results['ids'][0]}, score={results['distances'][0]}")
+