Question: Is StructuredQueryOutputParser compatible with GPT-3.5 for nested filters?

[zyloveslego]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

This issue can be reproduced in the official notebook:
[`self_query_hotel_search.ipynb`](https://github.com/langchain-ai/langchain/blob/master/cookbook/self_query_hotel_search.ipynb), specifically cell [18] to [20].


examples = [
    (
        "I want a hotel in the Balkans with a king sized bed and a hot tub. Budget is $300 a night",
        {
            "query": "king-sized bed, hot tub",
            "filter": 'and(in("country", ["Bulgaria", "Greece", "Croatia", "Serbia"]), lte("onsiterate", 300))',
        },
    ),
    (
        "A room with breakfast included for 3 people, at a Hilton",
        {
            "query": "Hilton",
            "filter": 'and(eq("mealsincluded", true), gte("maxoccupancy", 3))',
        },
    ),
]
prompt = get_query_constructor_prompt(
    doc_contents, filter_attribute_info, examples=examples
)

chain.invoke({
    "query": "Find a 2-person room in Vienna or London, preferably with meals included and AC"
})


This results in a malformed filter:

```json
{
    "query": "2-person room, Vienna, London, meals included, AC",
    "filter": "and(or(eq(\"city\", \"Vienna\"), eq(\"city\", \"London\"), eq(\"mealsincluded\", true))"
}

...which throws a OutputParserException due to the missing closing parenthesis.

Note: Using gpt-4-turbo instead produces the correct output.

### Description

I'm referencing the official notebook `self_query_hotel_search.ipynb` to build a structured query chain that extracts filters from natural language hotel search queries.

I expect the model to return a valid logical DSL like:

```json
{
  "filter": "and(or(eq(\"city\", \"Vienna\"), eq(\"city\", \"London\")), gte(\"maxoccupancy\", 2), eq(\"mealsincluded\", true))"
}

However, when using gpt-3.5-turbo, the consistently often collapses all conditions into the or(...) clause, and fails to close the parentheses properly.

I’ve tried adding more structurally complex few-shot examples, but it still happens.

Switching to gpt-4-turbo significantly reduce the issue.

Notebook reference: self_query_hotel_search.ipynb, cell [20].

System Info

langchain==0.3.24
langchain-chroma==0.2.3
langchain-community==0.3.22
langchain-core==0.3.56
langchain-elasticsearch==0.3.2
langchain-openai==0.3.12
langchain-text-splitters==0.3.8