, I am facing this error after I update llama-index pac...

oopenmind

, I am facing this error after I update llama-index package.

Error code: 400 - {'error': {'message': "'name' is a required property - 'functions.0'", 'type': 'invalid_request_error', 'param': None, 'code': None}}

37 comments

kkapa.ai

Please stand by @openmind as I search the knowledge base. I normally take about 20 seconds or so. :wave:

kkapa.ai

----------------------
The error message you're seeing suggests that the 'name' property is missing in the first function of the 'functions' array you're passing to the LlamaIndex API.

In the LlamaIndex API, when you're making a function call, you need to define a class with the required properties and then convert it to an OpenAI function using the to_openai_function method. Here's an example from the Llama API documentation ():

Plain Text

from pydantic import BaseModel
from llama_index.llms.openai_utils import to_openai_function

class Song(BaseModel):
    """A song with name and artist"""
    name: str
    artist: str

song_fn = to_openai_function(Song)

In this example, 'Song' is the name of the function. When you call the complete method, you need to pass this function in the 'functions' array:

Plain Text

llm = LlamaAPI(api_key=api_key)
response = llm.complete("Generate a song", functions=[song_fn])

Please ensure that you're correctly defining and passing the function in your code. If the problem persists, it would be helpful to review the exact code that's causing the error.
----------------------
Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:
----------------------

oopenmind

@Logan M , Function calling is not working

LLogan M

How did you setup your function/tool?

oopenmind

Like Above

oopenmind

Expecting value: line 1 column 1 (char 0)
Gettign this error

oopenmind

why do we need to use api key from llama-index?
it was working with openai api key before
@Logan M

LLogan M

Don't listen to kapa there, it's using some other LLM, not openai

LLogan M

In your code, how did you setup the function/tool? Exactly like above?

oopenmind

llm = LlamaAPI(api_key="key")
func_response = llm.complete(func_prompt, functions=[compilance_fn])
function_call = func_response.additional_kwargs["function_call"]

This is what kapa said.

oopenmind

func_response = OpenAI(model="gpt-4").complete(func_prompt, functions=[compilance_fn])

Using this code, it worked well before using low version, but in latest version, it doesn't work.

LLogan M

Kapa is crazy, ignore it.

In any case, newer versions of llama-index use a newer openai client.

It looks like this now

Plain Text

from pydantic import BaseModel
from llama_index.llms import OpenAI
from llama_index.llms.openai_utils import to_openai_tool

class Song(BaseModel):
    """A song with name and artist"""
    name: str
    artist: str

song_fn = to_openai_tool(Song)

llm = OpenAI(model="gpt-4")
response = llm.complete(func_prompt, tools=[song_fn])

oopenmind

just replace functions to tools?

oopenmind

func_response = OpenAI(model="gpt-4").complete(func_prompt, tools=[compilance_fn])func_response = OpenAI(model="gpt-4").complete(func_prompt, tools=[compilance_fn])

LLogan M

And different function here

LLogan M

song_fn = to_openai_tool(Song)

oopenmind

let me check

oopenmind

function_call = func_response.additional_kwargs["function_call"]

oopenmind

this is necessary?

oopenmind

@Logan M

LLogan M

not really, that's only used to tell which function was used.

If you need it, it might be under a different key now, not function_call

oopenmind

class Compilance(BaseModel):
"""Compilance score between requirement and answer, Score should be number from 1 to 10."""
score: str

compilance_fn = to_openai_tool(Compilance)
func_response = OpenAI(model="gpt-4").complete(func_prompt, tools=[compilance_fn])

Fullcode, and not working, @Logan M

LLogan M

Plain Text

>>> response = llm.complete("Tell me a song", tools=[song_fn])
>>> song = Song.parse_raw(response.additional_kwargs['tool_calls'][0].function.arguments)
>>> song
Song(name='Bohemian Rhapsody', artist='Queen')
>>>

LLogan M

works fine

LLogan M

for me

oopenmind

what version are you using?

LLogan M

llama-index==0.9.11.post1, openai==1.3.5

oopenmind

llama-index==0.9.11

I am using this version

oopenmind

let me try to update llama-index version again

oopenmind

still same

LLogan M

Full code

Plain Text

from pydantic import BaseModel
from llama_index.llms.openai_utils import to_openai_tool
 
class Song(BaseModel):
    """A song with name and artist"""
    name: str
    artist: str
 
song_fn = to_openai_tool(Song)

llm = OpenAI(model="gpt-4")
response = llm.complete("Tell me a song", tools=[song_fn])
song = Song.parse_raw(response.additional_kwargs['tool_calls'][0].function.arguments)
print(song)

LLogan M

not sure what to tell you otherwise

oopenmind

Song.parse_raw(response.additional_kwargs['tool_calls'][0].function.arguments)

Here , how to access to name?

LLogan M

response.additional_kwargs['tool_calls'][0].function.name

oopenmind

score = func_response.additional_kwargs['tool_calls'][0].function.score

not working. but

oopenmind

@Logan M , ah, finally solved

oopenmind

thank you!

Add a reply

Find answers from the community

, I am facing this error after I update llama-index package. Error code: 400 - {'error':