Xata•9mo ago

Storing objects of a specific class

Hello. Could you please tell me how to store an object of type "Document" (from langchain.docstore.document import Document) in a table?

11 Replies

kostas•9mo ago

Hello! Here's a snippet showing what a document is and how to write it to Xata. Note that you need a column named "content" of type “Text” and any fields present in "metadata" (like "source") should correspond to column names in your Xata schema as well.

import os
from langchain_core.documents import Document
from langchain_community.vectorstores.xata import XataVectorStore
from langchain_openai import OpenAIEmbeddings

api_key = os.environ["XATA_API_KEY"]
db_url = "https://workspace-123456.us-east-1.xata.sh/db/langchaindb"

embeddings = OpenAIEmbeddings()

docs = [
            Document(page_content="a user's message", metadata={'source': 'chatroom1'})
        ]

embeddings = OpenAIEmbeddings()

vector_store = XataVectorStore.from_documents(
    docs, embeddings, api_key=api_key, db_url=db_url, table_name="vectors"
)

import os
from langchain_core.documents import Document
from langchain_community.vectorstores.xata import XataVectorStore
from langchain_openai import OpenAIEmbeddings

api_key = os.environ["XATA_API_KEY"]
db_url = "https://workspace-123456.us-east-1.xata.sh/db/langchaindb"

embeddings = OpenAIEmbeddings()

docs = [
            Document(page_content="a user's message", metadata={'source': 'chatroom1'})
        ]

embeddings = OpenAIEmbeddings()

vector_store = XataVectorStore.from_documents(
    docs, embeddings, api_key=api_key, db_url=db_url, table_name="vectors"
)

(OPENAI_API_KEY env var must be set) Ref: https://python.langchain.com/docs/integrations/vectorstores/xata/#create-the-xata-vector-store

kostas•9mo ago

АндрейOP•9mo ago

Thank you, for your reply. My goal is not to create vector storage, but to store selected chunks (Documents) on query to vector database for analysis.

kostas•9mo ago

You can load the Document into a Dictionary and pickup the columns you want to index into Xata from there

docs = [
            Document(page_content="a user's message", metadata={'source': 'chatroom1'})
        ]

myDocument=dict(docs[0])

payload={"content":myDocument["page_content"],"source":myDocument["metadata"]["source"]}

xata = XataClient(db_url=db_url)
record = xata.records().insert("mytable",payload)

docs = [
            Document(page_content="a user's message", metadata={'source': 'chatroom1'})
        ]

myDocument=dict(docs[0])

payload={"content":myDocument["page_content"],"source":myDocument["metadata"]["source"]}

xata = XataClient(db_url=db_url)
record = xata.records().insert("mytable",payload)

АндрейOP•9mo ago

Thank you, just what I needed! Learning ))

kostas•9mo ago

Happy to help!

АндрейOP•9mo ago

Help me, please. I can't understand the syntax of the query. I need to select all sub-queries related to questions related to a certain dialogue. data = xata.data().query("subquery", { "columns": ["number", "subquery", "subanswer", "query.*" ], "filter": { "query.dialog_id": dialog_id }, })

kostas•9mo ago

It queries a table named subquery (xata.data().query("subquery",...), retrieves a list of columns including all the columns under a linked table in a column named query ("query.*") and applies a filter Documentation references that will help you understand the query structure (switch the code tab to Python): https://xata.io/docs/sdk/get#columns-selection, https://xata.io/docs/sdk/get#selecting-columns-from-the-linked-tables, https://xata.io/docs/sdk/filtering#exact-matching Perhaps what's confusing you is that subquery refers to actual table names and query refers to a link column name in the table subquery. Do you really have tables and columns named like that?

АндрейOP•9mo ago

I didn't name the column in the subquery table correctly - "query_id", it should have been "query".

kostas•9mo ago

Great sounds like you made progress

АндрейOP•9mo ago

Only with your help!

Gaming

Programming

Storing objects of a specific class