idchlife/rag-solution
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

evaluation for rag systems update, script for evaluating against questions

Browse Source
This commit is contained in:
idchlife
2026-03-13 08:20:18 +03:00
parent 6c953a327f
commit 236f44b2c3
7 changed files with 562 additions and 28947 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
services/rag/langchain/agent.py
View File

@@ -20,6 +20,9 @@ from vector_storage import initialize_vector_store
# Load environment variables
load_dotenv()
CHAT_REQUEST_TIMEOUT_SECONDS = float(os.getenv("CHAT_REQUEST_TIMEOUT_SECONDS", "45"))
CHAT_MAX_RETRIES = int(os.getenv("CHAT_MAX_RETRIES", "0"))
def get_llm_model_info(
llm_model: Optional[str] = None,
@@ -149,10 +152,12 @@ def create_chat_agent(
openai_api_base=base_url_or_api_base,
openai_api_key=api_key,
temperature=0.1,
request_timeout=CHAT_REQUEST_TIMEOUT_SECONDS,
max_retries=CHAT_MAX_RETRIES,
)
logger.info(
f"Using OpenAI-compatible model: {model_name} via {base_url_or_api_base}"
f"Using OpenAI-compatible model: {model_name} via {base_url_or_api_base}, timeout={CHAT_REQUEST_TIMEOUT_SECONDS}s, retries={CHAT_MAX_RETRIES}"
)
else: # Default to ollama
# Initialize the Ollama chat model
@@ -160,9 +165,13 @@ def create_chat_agent(
model=model_name,
base_url=base_url_or_api_base, # Default Ollama URL
temperature=0.1,
sync_client_kwargs={"timeout": CHAT_REQUEST_TIMEOUT_SECONDS},
async_client_kwargs={"timeout": CHAT_REQUEST_TIMEOUT_SECONDS},
)
logger.info(f"Using Ollama model: {model_name}")
logger.info(
f"Using Ollama model: {model_name}, timeout={CHAT_REQUEST_TIMEOUT_SECONDS}s"
)
# Create the document retrieval tool
retrieval_tool = DocumentRetrievalTool()

8
services/rag/langchain/vector_storage.py
View File

@@ -25,6 +25,10 @@ OLLAMA_EMBEDDING_MODEL = os.getenv("OLLAMA_EMBEDDING_MODEL", "nomic-embed-text")
OPENAI_EMBEDDING_MODEL = os.getenv("OPENAI_EMBEDDING_MODEL", "text-embedding-ada-002")
OPENAI_EMBEDDING_BASE_URL = os.getenv("OPENAI_EMBEDDING_BASE_URL")
OPENAI_EMBEDDING_API_KEY = os.getenv("OPENAI_EMBEDDING_API_KEY")
EMBEDDING_REQUEST_TIMEOUT_SECONDS = float(
os.getenv("EMBEDDING_REQUEST_TIMEOUT_SECONDS", "30")
)
EMBEDDING_MAX_RETRIES = int(os.getenv("EMBEDDING_MAX_RETRIES", "0"))
def initialize_vector_store(
@@ -53,6 +57,8 @@ def initialize_vector_store(
model=OPENAI_EMBEDDING_MODEL,
openai_api_base=OPENAI_EMBEDDING_BASE_URL,
openai_api_key=OPENAI_EMBEDDING_API_KEY,
request_timeout=EMBEDDING_REQUEST_TIMEOUT_SECONDS,
max_retries=EMBEDDING_MAX_RETRIES,
)
elif EMBEDDING_STRATEGY == "none":
embeddings = None
@@ -63,6 +69,8 @@ def initialize_vector_store(
embeddings = OllamaEmbeddings(
model=OLLAMA_EMBEDDING_MODEL,
base_url="http://localhost:11434", # Default Ollama URL
sync_client_kwargs={"timeout": EMBEDDING_REQUEST_TIMEOUT_SECONDS},
async_client_kwargs={"timeout": EMBEDDING_REQUEST_TIMEOUT_SECONDS},
)
# Check if collection exists and create if needed