Add endpoint with similar markets

[kongzii]

We have implementation for fiding similar markets (based on question) in one of the agents here: https://github.com/gnosis/prediction-market-agent/blob/main/prediction_market_agent/db/pinecone_handler.py#L150.

Complication is that this is in the PMA repository, which isn't installable through pypi, but maybe you can install it through git directly into labs-api, not a big trouble for me, we are already doing that with predicitonprophet for example.

So, the task here is to use that in a new endpoint.

[gabrielfior]

The cleanest solution would be to move PineconeHandler to PMAT, so that PMA can use that. That's my preferred option.

Another solution would be to add logic similar to what PMA does (https://github.com/gnosis/prediction-market-agent/blob/31311b72822f4312a6f34f4a7ad3ccd33c97e271/prediction_market_agent/db/pinecone_handler.py#L160-L168), which could be done in just a few lines:

p = PineconeHandler()
r = p.vectorstore.similarity_search_with_relevance_scores(query="Who will win the US election in 2028?", k=10, score_threshold=0.25)
r[0]
# (Document(metadata={'close_time_timestamp': 1734073140.0, 'market_address': '0x9de21ac77fff7e4ca9602525e966b38a431f56a5', 'question_title': 'Will a Democrat win the 2028 Presidential Election?'}, page_content='Will a Democrat win the 2028 Presidential Election?'),
 0.911052972)

[kongzii]

Moving to PMAT sounds good to me as well.

I wouldn't duplicate the logic to keep it together (ie we won't break API if we change some logic in agents)