How to pass secret keys to exported flow that is triggered using Python ?

[gauravdhiman]

I exported my flow as vector_store_rag.json and I am triggering it in my basic python app / script like below.

from langflow.load import run_flow_from_json
TWEAKS = {
  "ChatInput-9FOSJ": {},
  "ParseData-f3zpS": {},
  "Prompt-DJHnM": {},
  "ChatOutput-sHzFM": {},
  "SplitText-Tb2qB": {},
  "File-EXg8c": {},
  "OpenAIModel-N0opB": {},
  "HuggingFaceInferenceAPIEmbeddings-uzQoe": {},
  "OpenRouter-HUmNL": {},
  "Chroma-dFezZ": {}
}

# HuggingFace_API_Key = "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"
result = run_flow_from_json(flow="./flows/vector_store_rag.json",
                            input_value="message",
                            fallback_to_env_vars=True, # False by default
                            tweaks=TWEAKS)

print(result)

When I run this script, I get an error, saying

  File "/Users/gauravdhiman/projects/python/ai/langflow-server/.venv/lib/python3.12/site-packages/langflow/graph/graph/base.py", line 605, in _run
    raise ValueError(f"Error running graph: {exc}") from exc
ValueError: Error running graph: Error building Component HuggingFace Embeddings: 

1 validation error for SecretStrInput
value
  Value error, Invalid value type `<class 'NoneType'>` for input `api_key` [type=value_error, input_value=None, input_type=NoneType]
    For further information visit https://errors.pydantic.dev/2.8/v/value_error

I understand the problem that its not able to see / load the secret strings used in flow. In flow I used Global variable to save and use secret strings (different API keys).

When I use the flow like this in Python, how can I pass the secrets to flow so that all the components in flow that need secrets get it as expected ?

I read here that the global variables used are actually stored in DB used by langflow. There seems to be some way (also explained on the same page) to override the variables and pass them through .env file, but I could not understand it well.

If anyone know how to do it, please share some working examples of passing global variables values through .env or any other way from python program..

[dosubot[bot]]

Hey there, @gauravdhiman! I'm here to help you with any bugs or questions you have while waiting for a human maintainer. Let's tackle this together!

To pass secret keys to an exported flow triggered using Python and ensure that all components in the flow receive the necessary secrets, you can use environment variables. Here are examples of how to pass global variable values through a .env file or directly from a Python program.

Using a .env File

1. Create a .env file with your secret keys:

   HUGGINGFACE_API_KEY=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
   

2. Load the .env file in your Python script using the dotenv package:

   from dotenv import load_dotenv
   import os
   
   # Load environment variables from .env file
   load_dotenv()
   
   # Access the secret key
   huggingface_api_key = os.getenv('HUGGINGFACE_API_KEY')
   
   # Use the secret key in your flow
   print(f"HuggingFace API Key: {huggingface_api_key}")
   
   result = run_flow_from_json(
       flow="./flows/vector_store_rag.json",
       input_value="message",
       fallback_to_env_vars=True,  # Ensure this is set to True
       tweaks=TWEAKS
   )
   
   print(result)

Using Environment Variables Directly in Python

1. Set the environment variable in your operating system or directly in your Python script:

   import os
   
   # Set the environment variable
   os.environ['HUGGINGFACE_API_KEY'] = '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'
   
   # Access the secret key
   huggingface_api_key = os.getenv('HUGGINGFACE_API_KEY')
   
   # Use the secret key in your flow
   print(f"HuggingFace API Key: {huggingface_api_key}")
   
   result = run_flow_from_json(
       flow="./flows/vector_store_rag.json",
       input_value="message",
       fallback_to_env_vars=True,  # Ensure this is set to True
       tweaks=TWEAKS
   )
   
   print(result)

Using Langflow's Global Variables

If you are using Langflow, you can create and manage global variables that are accessible from any Text field in your project. Here’s how to create a global variable for your secret keys:

1. Create and Add a Global Variable:

   • In the HuggingFace API Key text field, click the 🌐 button, then Add New Variable.
   • Enter huggingface_api_key in the Variable Name field.
   • Paste your HuggingFace API Key (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) in the Value field.
   • Select Credential for the Type.
   • Choose HuggingFace API Key in the Apply to Fields field to apply this variable to all fields named HuggingFace API Key.
   • Click Save Variable.

2. Configure Environment Variables in your .env file:

   • Set LANGFLOW_STORE_ENVIRONMENT_VARIABLES to true in your .env file to add all variables in LANGFLOW_VARIABLES_TO_GET_FROM_ENVIRONMENT to your user's Global Variables.
   • List the variables to get from the environment in LANGFLOW_VARIABLES_TO_GET_FROM_ENVIRONMENT, as a comma-separated list (e.g., HUGGINGFACE_API_KEY).

   LANGFLOW_STORE_ENVIRONMENT_VARIABLES=true
   LANGFLOW_VARIABLES_TO_GET_FROM_ENVIRONMENT=HUGGINGFACE_API_KEY
   

By following these methods, you can ensure that all components in the flow receive the necessary secrets securely ^[1][2].

To continue talking to Dosu, mention @dosu.

[gauravdhiman]

Setting an env variable in .env file did not help.

[gauravdhiman]

On digging in code found that we can pass the API keys through Tweak object that we pass --OR-- explicitly set fallback_to_env_vars parameter to True while calling run_flow_from_json() and pass the path to .env file in env parameter.

Either of these approaches should work. I tried the first one (passing it using Tweak and it worked. Here is the tweak object that I passed and worked for me. Imp think to note here is also that we need to set load_from_db key as False, else langflow will look into DB not in tweak.

TWEAKS = {
  "ChatInput-9FOSJ": {},
  "ParseData-f3zpS": {},
  "Prompt-DJHnM": {},
  "ChatOutput-sHzFM": {},
  "SplitText-Tb2qB": {},
  "File-EXg8c": {},
  "OpenAIModel-N0opB": {},
  "HuggingFaceInferenceAPIEmbeddings-uzQoe": {
    "api_key": {
        "load_from_db": False,
        "value": os.getenv("HUGGINGFACE_API_KEY")
    }
  },
  "OpenRouter-HUmNL": {
    "api_key": {
        "load_from_db": False,
        "value": os.getenv("OPENROUTER_API_KEY")
    }
  },
  "Chroma-dFezZ": {}
}