unable to run python ode after importing baml functions.

[resonancejb]

---

title: "Unable to run Python file using Baml code"
labels: bug, question

Issue Description

I'm new to the generative AI field and recently started using Baml for structured outputs via LLMs. I noticed that using Baml increased the accuracy of my outputs significantly, but I'm encountering issues executing my Python script that integrates with my Baml code. I've been stuck for a couple of days and would greatly appreciate any assistance in resolving this.

Baml Code

// Defining a data model.
client MyClient {
  provider "openai-generic"
  options {
    base_url "my_api_url"
    api_key env.Runpod_API_KEY
    model "Qwen/Qwen2.5-VL-7B-Instruct"
    temperature 0
    top_p 0.25
    presence_penalty 0
    frequency_penalty 0
  }
}

class AccountDetails {
  account_number string @description("Account number as shown on statement")
  account_holder string @description("Name of account holder")
  bank_name string @description("Name of the bank")
  branch string @description("Branch name/code")
  sheet_number string @description("Statement sheet/page number")
  year string @description("Year of the statement")
}

class Transaction {
  date string @description("ISO8601 formatted date")
  details string @description("Transaction description")
  transaction_type "credit" | "debit" | null @description("Type of transaction")
  amount float? @description("Transaction amount")
  balance float @description("Running balance after transaction don't calculate anything just copy the given running balance")
}

class BankStatement {
  account_details AccountDetails
  transactions Transaction[]
}

function ExtractBankStatement(statement_image: image) -> BankStatement {
  client MyClient
  prompt #"
    {{ _.role("user") }}
    Extract info from this bank statement:
    {{ statement_image }}
    {# special macro to print the output schema instructions. #}
    {{ ctx.output_format }}
  "#
}

test ExtractFromSampleStatement {
  functions [ExtractBankStatement]
  args {
    statement_image {
      file "statement_ocr_2.jpeg"
    }
  }
}

##Python Code: Extracting Bank Statement with Baml
Python

import base64
from baml_py import Image
from baml_client.sync_client import b
from baml_client.types import BankStatement

def extract_bank_statement(image_path: str) -> BankStatement:
    # Open the image file and read its bytes.
    try:
        with open(image_path, "rb") as f:
            image_data = f.read()
    except Exception as e:
        raise ValueError("Error reading image file: " + str(e))

    # Encode the image bytes to a base64 string.
    image_b64 = base64.b64encode(image_data).decode("utf-8")

    # Create the Baml image object using the helper.
    # The first parameter is the media type (MIME type), e.g. "image/jpeg".
    img_obj = Image.from_base64("image/jpeg", image_b64)

    # Call the Baml function with the image object.
    response = b.ExtractBankStatement(statement_image=img_obj)
    return response

def main():
    image_path = "statement_ocr_2.jpeg"  # Ensure this file exists in your working directory.
    bank_statement = extract_bank_statement(image_path)
    print("Extracted Bank Statement:")
    print(bank_statement)

if __name__ == "__main__":
    main()

Steps Taken
I have followed all the steps from baml_cli init to generate the necessary files and configurations.

[aaronvg]

do you have the error?

[resonancejb]

Client: MyClient (<unknown>) - 319ms
    ---PROMPT---
    user: Extract info from this bank statement:<image_placeholder base64>Answer in JSON using this schema:
    {
      account_details: {
        // Account number as shown on statement
        account_number: string,
        // Name of account holder
        account_holder: string,
        // Name of the bank
        bank_name: string,
        // Branch name/code
        branch: string,
        // Statement sheet/page number
        sheet_number: string,
        // Year of the statement
        year: string,
      },
      transactions: [
        {
          // ISO8601 formatted date
          date: string,
          // Transaction description
          details: string,
          // Type of transaction
          transaction_type: "credit" or "debit" or null,
          // Transaction amount
          amount: float or null,
          // Running balance after transaction don't calculate anything just copy the given running balance
          balance: float,
        }
      ],
    }
    
    ---REQUEST OPTIONS---
    model: "Qwen/Qwen2.5-VL-7B-Instruct"
    temperature: 0
    top_p: 0.25
    presence_penalty: 0
    frequency_penalty: 0
    ---ERROR (InvalidAuthentication (401))---
    LLM call failed: LLMErrorResponse { client: "MyClient", model: None, prompt: Chat([RenderedChatMessage { role: "user", allow_duplicate_role: false, parts: [Text("Extract info from this bank statement:"), Media(BamlMedia { media_type: Image, mime_type: Some("image/jpeg"), content: Base64(MediaBase64 { base64: "/9j/4AAQSkZJRgA...KKACiiigD//2Q==" }) }), Text("Answer in JSON using this schema:\n{\n  account_details: {\n    // Account number as shown on statement\n    account_number: string,\n    // Name of account holder\n    account_holder: string,\n    // Name of the bank\n    bank_name: string,\n    // Branch name/code\n    branch: string,\n    // Statement sheet/page number\n    sheet_number: string,\n    // Year of the statement\n    year: string,\n  },\n  transactions: [\n    {\n      // ISO8601 formatted date\n      date: string,\n      // Transaction description\n      details: string,\n      // Type of transaction\n      transaction_type: \"credit\" or \"debit\" or null,\n      // Transaction amount\n      amount: float or null,\n      // Running balance after transaction don't calculate anything just copy the given running balance\n      balance: float,\n    }\n  ],\n}")] }]), request_options: {"model": String("Qwen/Qwen2.5-VL-7B-Instruct"), "temperature": Number(0), "top_p": Number(0.25), "presence_penalty": Number(0), "frequency_penalty": Number(0)}, start_time: SystemTime { intervals: 133869244781653924 }, latency: 319.6103ms, message: "Failed to parse JSON: EOF while parsing a value at line 1 column 0", code: InvalidAuthentication }

[resonancejb]

i am getting this error
i am passing image to the function after converting into base64

[resonancejb]

Can you please help with it.

[aaronvg]

here's a good guide as well on processing PDFs into images: https://blog.kuzudb.com/post/unstructured-data-to-graph-baml-kuzu/

[resonancejb]

thanks i will look into this

[aaronvg]

The problem is that it says "invalid authentication", which means baml isn't picking up your environment API key. Make sure you are on version 0.80.1 of baml-py and the vscode extension. And try importing the baml_client into your python file after you load_dotenv() or load your credentials into python