Rag logs not clearing

2025-03-03 12:54 UTC gpt-4o Open in ChatGPT ↗

i have this code that is responsible for creating a langchain class object that is managed through gradio. I have the rag.logs to store history of rag steps. but for some reason it’s not clearing when i reload the app.

def init(self): # Public variables (Accessible directly) self.parameters = None self.workflow_steps = None self.persona_steps = None self.prompts = None self.rag_queries = None self.llm = None self.rag_retriever = None self.rag_results = {} self.generated_inputs = {}

    # ==== chain execution =====
    self.persona_workflow = None
    self.full_workflow = None
    self.step_index = 0
    self.user_inputs = {}  # Stores user-modified prompts
    #self.execution_history = []  # Stores all results
    self.current_persona = ""  # Stores the latest persona

    # ==== gradio =====
    self.history = []
    self.persona_history = []
    self.rag_logs = []

    # Load configurations
    #self._initialize_config()

def _initialize_config(self):
    """Private method to load configurations."""
    # Load environment variables
    load_dotenv()
    HUGGINGFACE_TOKEN = os.getenv("HUGGINGFACE_TOKEN")

    if not HUGGINGFACE_TOKEN:
        raise ValueError("HUGGINGFACE_TOKEN is not set! Make sure to define it in .env.")

    # Load YAML Config
    with open("config_article_silver_fir_story_1_OBEY.yml", "r", encoding="utf-8") as file:
        config = yaml.safe_load(file)

    self.parameters = config["parameters"]
    self.workflow_steps = config["workflow"]
    self.persona_steps = config["persona_workflow"]
    self.prompts = config["prompts"]
    self.rag_queries = config["rag"]

    # Initialize Hugging Face LLM
    self.llm = HuggingFaceAPI(api_url=config["API"]["model_url"], api_token=HUGGINGFACE_TOKEN)

    # Initialize RAG Retriever
    embedding_model = SentenceTransformer('all-MiniLM-L6-v2')
    self.rag_retriever = RagRetrieveWithMeta(
        embedding_model,
        config['rag']['scientific']['faiss'],
        config['rag']['scientific']['embeddings'],
        config['rag']['diary']['faiss'],
        config['rag']['diary']['embeddings'],
        config['rag']['weather']['faiss'],
        config['rag']['weather']['embeddings'],
        config['rag']['insights']['faiss'],
        config['rag']['insights']['embeddings']
    )

If your rag_logs list is not clearing when you reload the app, there are a few possible reasons:

Potential Issues and Fixes:

Gradio State Persistence
- Gradio retains state across interactions unless explicitly reset. If self.rag_logs is tied to a Gradio component (like gr.Textbox() or gr.State()), make sure you reset it when reloading.
Fix: Explicitly Reset rag_logs When Reloading
```
def reload_app(self):
    """Function to reset all relevant states, including rag_logs"""
    self.history = []
    self.persona_history = []
    self.rag_logs = []  # Ensure it resets
```
- Call reload_app() when you reload your app.
Global or Persistent State
- If rag_logs is being stored in a global variable or a singleton instance of the class, it might not reset properly when the app reloads.
- Ensure a new instance of the class is created when reloading.
Fix:
- When initializing your class in Gradio, ensure you reinstantiate it instead of reusing an old instance.
```
my_app_instance = MyClass()  # Ensure this runs on reload
```
Check If rag_logs is Being Reused
- If rag_logs is referenced in Gradio UI elements with state, it may persist across reloads.
Fix: Manually Clear State in UI Handling
```
def reset_rag_logs(self):
    self.rag_logs.clear()  # Explicitly clear instead of reassigning
```
- Call reset_rag_logs() inside any on_click or on_reload handler.
Ensure the Class Constructor is Called on Reload
- If the class object is not being re-created, previous states will persist.
Check where your class is instantiated and confirm it’s being reloaded correctly.
```
app_instance = MyClass()  # This should be inside the reload flow
```

Debugging Steps:

Print logs to confirm when rag_logs is being reset:

print("Before reset:", self.rag_logs)
self.rag_logs = []
print("After reset:", self.rag_logs)

Check Gradio’s state management if it is persisting the object.
Ensure reload_app() is triggered when expected.

Let me know if you need further debugging! 🚀