json format looks right to me, what is it complaining about?

[iplayfast]

So I'm working with the structured output example, modified for use with ollama. It is calling ollama and getting results, and the results look good to me, (json returned). But the pedantic is still throwing an error.

AttributeError: type object 'CitiesData' has no attribute 'model_validate'

Thanks for any insight you can give me.

Code is here.

from pydantic import BaseModel
import json
import re
import pydantic
from pydantic import ValidationError
from typing import Optional, List
from colorama import Fore
from haystack import component
from haystack.components.builders import PromptBuilder
from haystack_integrations.components.generators.ollama import OllamaGenerator
from haystack import Pipeline

class City(BaseModel):
    name: str
    country: str
    population: int
    
class CitiesData(BaseModel):
    cities: List[City]
    
#json_schema = CitiesData.model_json_schema(indent = 2)# schema_json(indent=2)
# Create an instance of CitiesData to access the schema
cities_data_instance = CitiesData(cities=[])

# Generate the JSON schema for CitiesData
json_schema = cities_data_instance.schema(by_alias=True)

@component
class OutputValidator:
    def __init__(self, pydantic_model: pydantic.BaseModel):
        self.pydantic_model = pydantic_model
        self.iteration_counter = 0

    @component.output_types(valid_replies=List[str], invalid_replies=Optional[List[str]],error_message=Optional[str])
    def run(self, replies: List[str]):
        self.iteration_counter += 1
        ## Try to parse the LLM's reply ##
        # if the LLm's reply is a valid object, return "valid_replies"
        try:            
            output_dict = json.loads(replies[0])         
            self.pydantic_model.model_validate(output_dict)#  parse_obj(output_dict)
            print(
                Fore.GREEN
                + f"OutputValidator at Iteration {self.iteration_counter}: Valid JSON from LLM - No need for looping:{ replies[0]} \n"
            )
            return {"valid_replies": replies}
        except(ValueError, ValidationError) as e:
            print(
                Fore.RED
                + f"OutputValidator at Iteration {self.iteration_counter}: Invalid JSON from LLM - Let's try again\n"
                f"Output from LLM:\n {replies[0]} \n"
                f"Error from OutputValidator: {e}"
            )
            return {"invalid_replies": replies, "error_message": str(e)}
        
output_validator = OutputValidator(pydantic_model=CitiesData)
        
prompt_template = """
    Creat a JSON object from the information present in this passage: {{passage}}.
    Only use information that is present in the passage. Follow this JSON schema, but only return the actual instances without any additional schema defintion:\n",
    "{{schema}}\n",
    "Make sure your response is a dict and not a list. \n",
    "{% if invalid_replies and error_message %}\n",
    "  You already created the following output in a previous attempt: {{invalid_replies}}\n",
    "  However, this doesn't comply with the format requirements from above and triggered this Python exception: {{error_message}}\n",
    "  Correct the output and try again. Just return the corrected output without any extra explanations.",
    "{% endif %}\n",
    """        
    
total_prompt_will_be = """Creat a JSON object from the information present in this passage: "Berlin is the capital of Germany. It has a population of 3,850,809. Paris, "
            "France's capital, has 2.161 million residents. Lisbon is the capital and the "
            "largest city of Portugal with the population of 504,718.\n".
    Only use information that is present in the passage. Follow this JSON schema, but only return the actual instances without any additional schema defintion:\n",
    "'{\n  "title": "CitiesData",\n  "type": "object",\n  "properties": {\n    "cities": {\n      "title": "Cities",\n      "type": "array",\n      "items": {\n        "$ref": "#/definitions/City"\n      }\n    }\n  },\n  "required": [\n    "cities"\n  ],\n  "definitions": {\n    "City": {\n      "title": "City",\n      "type": "object",\n      "properties": {\n        "name": {\n          "title": "Name",\n          "type": "string"\n        },\n        "country": {\n          "title": "Country",\n          "type": "string"\n        },\n        "population": {\n          "title": "Population",\n          "type": "integer"\n        }\n      },\n      "required": [\n        "name",\n        "country",\n        "population"\n      ]\n    }\n  }\n}'\n",
    "Make sure your response is a dict and not a list. \n",
    """
prompt_builder = PromptBuilder(template=prompt_template)
generator = OllamaGenerator(model="mistral",#model="notus:latest", #model="openhermes2.5-mistral:latest",
                            url="http://localhost:11434/api/generate")
                                #url="localhost:11434/api/generate")
pipeline = Pipeline(max_loops_allowed = 3)
#now add componets to your pipeline
pipeline.add_component(instance=prompt_builder,name="prompt_builder")
pipeline.add_component(instance=generator,name="llm") 
pipeline.add_component(instance=output_validator,name="output_validator")
# now connect the components to each other
pipeline.connect( "prompt_builder", "llm")
#pipeline.connect(: "prompt_builder",connect_to: "llm")
pipeline.connect( "llm", "output_validator")
# if a component has more than one output or input, explicitly specify the connections:
pipeline.connect("output_validator.invalid_replies", "prompt_builder.invalid_replies")
pipeline.connect("output_validator.error_message", "prompt_builder.error_message")

pipeline.draw("auto-correct-pipeline.png")


passage = ("Berlin is the capital of Germany. It has a population of 3,850,809. Paris, "
            "France's capital, has 2.161 million residents. Lisbon is the capital and the "
            "largest city of Portugal with the population of 504,718.\n")
result = pipeline.run({"prompt_builder": {"passage": passage, "schema": json_schema}})

valid_reply = result["output_validator"]["valid_replies"][0]
valid_json = json.loads(valid_reply)
print("\nValid JSON:")
print(valid_json)

[anakin87]

Hello, @iplayfast...

My impression is that the error is caused by introducing self.pydantic_model.model_validate(output_dict)
instead of self.pydantic_model.parse_obj(output_dict) (see the tutorial).

Can you try and let me know?

[iplayfast]

A different error, goes past that problem but doesn't return any results. Still stuck.

python tutorial.py 
{}
Traceback (most recent call last):
  File "/home/chris/ai/haystack/tutorial.py", line 102, in <module>
    valid_reply = result["output_validator"]["valid_replies"][0]
KeyError: 'output_validator'

I tried printing result and it was {}

[anakin87]

We just discovered this and we are working on the solution...

[iplayfast]

Cool!

[anakin87]

Fixed in https://github.com/deepset-ai/haystack/releases/tag/v2.0.0-beta.7

you can update the version using
pip install -U haystack-ai

[iplayfast]

OutputValidator at Iteration 1: Invalid JSON from LLM - Let's try again
Output from LLM:
"{
'title': 'CitiesData',
'type': 'object',
'properties': {
'cities': {
'title': 'Cities',
'type': 'array',
'items': [
{
'name': 'Berlin',
'country': 'Germany',
'population': 3850809
},
{
'name': 'Paris',
'country': 'France',
'population': 2161000
},
{
'name': 'Lisbon',
'country': 'Portugal',
'population': 504718
}
]
}
},
'required': ['cities']
}"
Error from OutputValidator: Invalid control character at: line 1 column 4 (char 3)

still getting an error. still looks right to me.