How does schema work?

[pdavis68]

I've been trying to set the schema and I've tried a number of variations, but I can't seem to hit upon the magic combination to get it to work.

For example, with this code:

from scrapegraphai.graphs import OmniScraperGraph

# Define the configuration for the graph
graph_config = {
    "llm": {
        "api_key": "<key>",
        "model": "gpt-4-turbo",
    },
    "embeddings": {
        "model": "ollama/nomic-embed-text",
        "base_url": "http://localhost:11434/",  # set ollama URL arbitrarily
    },
    "max_results": 5,
    "verbose": True
}

schema= """
    { 
    "top_stories": [
            { 
                "title": "...", 
                "url": "...", 
            }, 
            { 
                "title": "...", 
                "url": "...", 
            } 
        ] 
    } 
"""

omni_scraper_graph = OmniScraperGraph(
   prompt="List the 10 most important things in the news right now. Give the headline and the url.",
   source="http://news.google.com",
   config=graph_config,
   schema=schema
)

result = omni_scraper_graph.run()
print(result)

But what I get is this:

{'news': [{'headline': 'Mexico election 2024: Live updates, results and latest news', 'url': 'http://news.google.com/articles/CBMiN2h0dHBzOi8vYXBuZXdzLmNvbS9saXZlL21leGljby1lbGVjdGlvbi1yZXN1bHRzLXVwZGF0ZXPSAQA?hl=en-US≷=US&ceid;=US%3Aen'}, {'headline': 'Mexico election live updates: Voters set to elect first female president', 'url': 'http://news.google.com/articles/CBMiX2h0dHBzOi8vd3d3LmNubi5jb20vYW1lcmljYXMvbGl2ZS1uZXdzL21leGljby1wcmVzaWRlbnRpYWwtZWxlY3Rpb24tcmVzdWx0cy0wNi0wMi0yNC9pbmRleC5odG1s0gFjaHR0cHM6Ly9hbXAuY25uLmNvbS9jbm4vYW1lcmljYXMvbGl2ZS1uZXdzL21leGljby1wcmVzaWRlbnRpYWwtZWxlY3Rpb24tcmVzdWx0cy0wNi0wMi0yNC9pbmRleC5odG1s?hl=en-US≷=US&ceid;=US%3Aen'}, {'headline': 'Mexico election 2024 results live: Will Sheinbaum win the presidency?', 'url': 'http://news.google.com/articles/CBMidGh0dHBzOi8vd3d3LmFsamF6ZWVyYS5jb20vbmV3cy9saXZlYmxvZy8yMDI0LzYvMy9tZXhpY28tZWxlY3Rpb24tMjAyNC1yZXN1bHRzLWxpdmUtd2lsbC1zaGVpbmJhdW0td2luLXRoZS1wcmVzaWRlbmN50gF4aHR0cHM6Ly93d3cuYWxqYXplZXJhLmNvbS9hbXAvbmV3cy9saXZlYmxvZy8yMDI0LzYvMy9tZXhpY28tZWxlY3Rpb24tMjAyNC1yZXN1bHRzLWxpdmUtd2lsbC1zaGVpbmJhdW0td2luLXRoZS1wcmVzaWRlbmN5?hl=en-US≷=US&ceid;=US%3Aen'}, {'headline': "Claudia Sheinbaum Will Be Mexico's Next President. But Which Version of Her Will Govern?", 'url': 'http://news.google.com/articles/CBMiaWh0dHBzOi8vd3d3LnBvbGl0aWNvLmNvbS9uZXdzL21hZ2F6aW5lLzIwMjQvMDYvMDEvY2xhdWRpYS1zaGVpbmJhdW0tbWV4aWNvLXByZXNpZGVudGlhbC1lbGVjdGlvbi0wMDE2MTA4MdIBAA?hl=en-US≷=US&ceid;=US%3Aen'}, {'headline': 'Majority of independents say Trump received ‘fair trial’: Poll', 'url': 'http://news.google.com/articles/CBMicWh0dHBzOi8vdGhlaGlsbC5jb20vcmVndWxhdGlvbi9jb3VydC1iYXR0bGVzLzQ2OTkxNjgtbWFqb3JpdHktb2YtaW5kZXBlbmRlbnRzLXNheS10cnVtcC1yZWNlaXZlZC1mYWlyLXRyaWFsLXBvbGwv0gF1aHR0cHM6Ly90aGVoaWxsLmNvbS9yZWd1bGF0aW9uL2NvdXJ0LWJhdHRsZXMvNDY5OTE2OC1tYWpvcml0eS1vZi1pbmRlcGVuZGVudHMtc2F5LXRydW1wLXJlY2VpdmVkLWZhaXItdHJpYWwtcG9sbC9hbXAv?hl=en-US≷=US&ceid;=US%3Aen'}, {'headline': "Looming over Trump's conviction: reversal by the '13th juror'", 'url': 'http://news.google.com/articles/CBMiT2h0dHBzOi8vd3d3LnBvbGl0aWNvLmNvbS9uZXdzLzIwMjQvMDYvMDIvdHJ1bXAtY29udmljdGlvbi1hcHBlYWwtanVyb3ItMDAxNjExMTDSAQA?hl=en-US≷=US&ceid;=US%3Aen'}, {'headline': 'Half say Trump verdict correct, should end campaign: Poll', 'url': 'http://news.google.com/articles/CBMiZmh0dHBzOi8vdGhlaGlsbC5jb20vaG9tZW5ld3MvY2FtcGFpZ24vNDY5OTMyOS1oYWxmLXNheS10cnVtcC12ZXJkaWN0LWNvcnJlY3Qtc2hvdWxkLWVuZC1jYW1wYWlnbi1wb2xsL9IBamh0dHBzOi8vdGhlaGlsbC5jb20vaG9tZW5ld3MvY2FtcGFpZ24vNDY5OTMyOS1oYWxmLXNheS10cnVtcC12ZXJkaWN0LWNvcnJlY3Qtc2hvdWxkLWVuZC1jYW1wYWlnbi1wb2xsL2FtcC8?hl=en-US≷=US&ceid;=US%3Aen'}, {'headline': 'Opinion | Holy Cow, 34 for 45!', 'url': 'http://news.google.com/articles/CBMiQmh0dHBzOi8vd3d3Lm55dGltZXMuY29tLzIwMjQvMDYvMDEvb3Bpbmlvbi9kb25hbGQtdHJ1bXAtdHJpYWwuaHRtbNIBAA?hl=en-US≷=US&ceid;=US%3Aen'}, {'headline': "US has 'every expectation' Israel will expect ceasefire proposal", 'url': 'http://news.google.com/articles/CBMiLmh0dHBzOi8vd3d3LmJiYy5jb20vbmV3cy9hcnRpY2xlcy9jZDExbDU5MHFxd2_SATJodHRwczovL3d3dy5iYmMuY29tL25ld3MvYXJ0aWNsZXMvY2QxMWw1OTBxcXdvLmFtcA?hl=en-US≷=US&ceid;=US%3Aen'}, {'headline': "As Hunter Biden goes to trial on gun charges, here's how his attorneys plan to defend a 'simple case'", 'url': 'http://news.google.com/articles/CBMiYWh0dHBzOi8vYWJjbmV3cy5nby5jb20vVVMvaHVudGVyLWJpZGVuLXRyaWFsLWd1bi1jaGFyZ2VzLWF0dG9ybmV5cy1wbGFuLWRlZmVuZC9zdG9yeT9pZD0xMTA3MjM0NjbSAQA?hl=en-US≷=US&ceid;=US%3Aen'}]}

Very cool tool, BTW. Really enjoying exploring it.

[PeriniM]

Thank you for the feedback! Yes, right now it is more like a "soft" schema constraint. We are about to add Pydantic schema validation, I will open an Issue and start working on it, fell free to contribute!

[pdavis68]

Sadly, python is a weak spot for me. I'm primarily a C# developer. I'll wrap ScrapeGraphAi in a Flask app and call it from C# :-)

Trying to decide whether or not to invest the time in Python. I'm only a few years from retirement, but programming was a hobby for before it was a job and I imagine I'll still be programming in retirement. I think there's part of me that's hoping at some point people are going to start writing AI tools in other languages. I mean, there's no reason not to.

Appreciate the response, though. Makes sense. I don't really know how Pydantic works, but I've heard of it. I'll check it out.

[archidealhome]

Ah, the schema seems to be defining the expected output format. Try removing the schema definition and see if the scraper can produce results in its natural format. That might be easier to work with for now. Contact Xylostar for more guidance!

[VinciGit00]

you should have used pydantic, look at the exampels