pgurazada1 commited on
Commit
b1e21fb
·
verified ·
1 Parent(s): 7562209

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +8 -8
app.py CHANGED
@@ -257,19 +257,19 @@ def claim_generator(input_bill_path):
257
 
258
  return output['final_output']
259
 
260
- agentic_workflow_representation = 'The workflow used to generate an expense claim document is represented below: \n <img src="https://cdn-uploads.huggingface.co/production/uploads/64118e60756b9e455c7eddd6/iqdGjUG7POKJXLItzWb-K.png">'
261
  agentic_workflow_description = """
262
  This demo represents a multi-agent collaborative workflow that generates an expense claim document based on a submitted bill.
263
- The key steps in this workflow once the user uploads a bill are:
264
- 1. Extracting text from an image of the bill.
265
- 2. Categorizing items in the bill to: alcoholic drinks, non-alcoholic drinks and food.
266
- 3. Retrieving relevant sections from the company reimbursement policy and removing items that are non reimbursable.
267
- 4. Computing the total amount that can be reimbursed (along with taxes).
268
- 5. Preparing a claim document in the company format.
269
 
270
  Each step in this workflow are executed by function-calling agents.
271
  For example, the text extractor is an API-calling agent that calls uses the Open AI APIs to extract text from the bill.
272
- Similarly, the verifier is a search agent that extracts relevant portions of the company policy based on the categories identified.
273
  """
274
 
275
  demo = gr.Interface(
 
257
 
258
  return output['final_output']
259
 
260
+ agentic_workflow_representation = 'The agentic workflow used to generate an expense claim document is represented below: \n <img src="https://cdn-uploads.huggingface.co/production/uploads/64118e60756b9e455c7eddd6/iqdGjUG7POKJXLItzWb-K.png">'
261
  agentic_workflow_description = """
262
  This demo represents a multi-agent collaborative workflow that generates an expense claim document based on a submitted bill.
263
+ Once a user uploads a bill to the interface, the following steps are executed:
264
+ 1. Extract text from an image of the bill.
265
+ 2. Categorize items in the bill to: alcoholic drinks, non-alcoholic drinks and food.
266
+ 3. Based on the categories, retrieve relevant sections from the company reimbursement policy; remove items that are non reimbursable.
267
+ 4. Compute the total amount that can be reimbursed (along with taxes).
268
+ 5. Prepare a claim document in the company format (as a markdown document).
269
 
270
  Each step in this workflow are executed by function-calling agents.
271
  For example, the text extractor is an API-calling agent that calls uses the Open AI APIs to extract text from the bill.
272
+ Similarly, the verifier is a search agent that extracts relevant portions of the company policy based on the nature of the bill.
273
  """
274
 
275
  demo = gr.Interface(