"Invalid LLM response": using LLM call/ generation for assigning variable value

[tjf-fxp]

What I have done so far:

• I have read all the NeMo-Guardrails docs
• I have updated the package to the latest version before submitting this issue
• I have searched the existing issues of NeMo-Guardrails

What i'm trying to do is make an LLM to assign a variable value. The LLM correctly responds but gets flagged with "Invalid LLM response".

Below is the code snippet:

flow bot extract job position
    global $job_position
    # Extract job position and location from user input 
    $job_position = ..."Extract job position from $last_user_transcript. This can be a specific title like 'CEO', 'CTO', 'CFO', or a more general term like 'Manager'."

flow bot extract location
    global $location    
    # Extract job position and location from user input 
    $location = ..."If mentioned, extract the place name from $last_user_transcript. This can be a city, state, or country. If no location is mentioned, use 'Any location.'."

This is the response:

System
This is the current conversation between the user and the bot:
User
user action: user said "This is a CFO in New York"
Bot
user intent: user said job position
System
Follow these instruction `Extract job position from var_last_user_transcript. This can be a specific title like 'CEO', 'CTO',
'CFO', or a more general term like 'Manager'.` to generate a value that is assigned to:
$job_position =
16:07:04.659 | StartTimerBotAction | {"action_uid": "61f6...", "timer_name": "llm_response_polling", "duration": 1.0}        
16:07:05.587 | Output Stats None
16:07:05.588 | LLM call took 0.93 seconds
2025-05-28 16:07:05,589 WARNING  Error while execution 'GenerateValueAction' with parameters '{'var_name': 'job_position', 'instructions': "Extract job position from var_last_user_transcript. This can be a specific title like 'CEO', 'CTO', 'CFO', or a more general term like 'Manager'."}': Invalid LLM response: `CFO`
2025-05-28 16:07:05,589 ERROR    Invalid LLM response: `CFO`

User
user action: user said "This is a CFO in New York"
Bot
user intent: user said job position
bot intent: bot extract job position
System
Follow these instruction `If mentioned, extract the place name from var_last_user_transcript. This can be a city, state, or  
country. If no location is mentioned, use 'Any location.'.` to generate a value that is assigned to:
$location =
16:07:06.518 | Output Stats None
16:07:06.519 | LLM call took 0.89 seconds
2025-05-28 16:07:06,520 WARNING  Error while execution 'GenerateValueAction' with parameters '{'var_name': 'location', 'instructions': "If mentioned, extract the place name from var_last_user_transcript. This can be a city, state, or country. If no location is mentioned, use 'Any location.'."}': Invalid LLM response: `New York`
2025-05-28 16:07:06,521 ERROR    Invalid LLM response: `New York`

Here the user said: This is a CFO in New York. The correct value to assign to job_position is CFO and location is New York, which is the LLM has correctly responded with. But what's causing it to be flagged as: Invalid LLM response for both CFO and New York ?

This is my config.yml:

colang_version: "2.x"

models:
  - type: main
    engine: openai
    model: gpt-4
    parameters: 
      api_key: sk-proj-...

streaming: True 

I am not sure what's going on. There's a similar issue with assigning value to a variable through an LLM call I posted here. I appreciate any help and insights into this issue.

Thanks in advance!