[Store] Add support for ClickHouse

[lyrixx]

Q	A
Bug fix?	no
New feature?	yes
Docs?	no
Issues
License	MIT

We already have ClickHouse in our infrastructure, So I'm benchmarkin it.
And, let's share my code 🎉

As you can see, I let some extension points (for the metadata). AFAIK, it's not possible to
store structured / indexed JSON in CH. We have to store it as String instead.
We could instead use better typed structure, but Symfony cannot know the type, obviously.
So with string, if we want to filter the results, we need to use [JSONExtractString](https://clickhouse.com/docs/sql-reference/functions/json-functions#jsonextractstring).
And It's slow!

So with the current extensions points, I could write the following code:

class ClickHouseStore extends SymfonyClickHouseStore
{
    public function initialize(array $options = []): void
    {
        $sql = <<<'SQL'
            CREATE TABLE IF NOT EXISTS {{ table }} (
                id UUID,
                metadata String,
                embedding Array(Float32),
                crawlId UUID,
            ) ENGINE = MergeTree()
            ORDER BY id
        SQL;

        $this->execute('POST', $sql);
    }

    protected function formatVectorDocument(VectorDocument $document): array
    {
        $formatted = parent::formatVectorDocument($document);

        $formatted['crawlId'] = $document->metadata['crawlId'];

        return $formatted;
    }
}

And then:

$clickHouseStore = new ClickHouseStore(
    httpClient: HttpClient::createForBaseUri($_SERVER['CLICKHOUSE_URI']),
    databaseName: 'pocai',
    tableName: $tableName,
);

$documents = $clickHouseStore->query(
    $vector,
    [
        'where' => 'crawlId = {crawlId:String} AND id != {currentId:UUID} AND score < 0.1',
        'params' => [
            'crawlId' => $crawlPoId,
            'currentId' => $row['id'],
        ],
    ],
);

---

If the PR could be accepted, I'll write tests

[chr-hertel]

the need for extension is not really ClickHouse specific, is it?
I think we need an extension point somewhere here anyways. a low level mapping/hydration thingy.
with #18 we have a similar topic i'd say - and with #158 we also have an issue of making too many assumption. totally open for ideas. :)

[lyrixx]

the need for extension is not really ClickHouse specific, is it?

Indeed 👍🏼 The current StoreInterface is a bit limited, but I eventually managed to do what I need to POC something.

I think we need an extension point somewhere here anyways. a low level mapping/hydration thingy.

Yes!

---

Should I work on tests? edit: done ✅

[OskarStark]

Suggested change

	return $this
	->httpClient
	return $this->httpClient

[OskarStark]

Let's use MockHttpClient?