Use SynapCores from Python: pip install, SQL and vector search
The synapcores package on PyPI talks to your engine over its REST API. Install it with pip, log in, and run SQL from Python, including vector search with embeddings the database computes for you.
Install free, one command
curl -fsSL https://get.synapcores.com | SYNAPCORES_REF=docs-python shFree Community Edition. Linux (x86_64, aarch64) and macOS on Apple Silicon; Windows via WSL2. Docker and other options
You need a running engine (quickstart or Docker), its admin password, and Python 3.8 or later.
Install the SDK
python3 -m venv venv
source venv/bin/activate
pip install synapcores
python -c "import synapcores; print(synapcores.__version__)"0.5.0Run SQL and a vector search
Save this as quickstart.py. It creates a small table of help articles, embeds them inside the database, and finds the article closest in meaning to a question:
import os
from synapcores import SynapCores
client = SynapCores(host="localhost", port=8080)
client.login(os.environ["SYNAPCORES_USER"], os.environ["SYNAPCORES_PASSWORD"])
# A throwaway database keeps the demo away from your own tables.
client.sql("CREATE DATABASE IF NOT EXISTS py_quickstart")
client.sql("""CREATE TABLE IF NOT EXISTS py_quickstart.articles (
id INTEGER PRIMARY KEY, title TEXT, body TEXT, embedding VECTOR(384))""")
articles = [
[1, "Reset your password", "Click Forgot password at sign-in to get a reset link."],
[2, "Update billing details", "Change the card on file under Settings > Billing."],
[3, "Export your data", "Download your records as a CSV from Settings > Export."],
[4, "Invite teammates", "Add colleagues to your workspace in Settings > Members."],
]
# EMBED() turns each body into a 384-dim vector inside the database.
for row in articles:
client.sql("INSERT INTO py_quickstart.articles VALUES ($1, $2, $3, EMBED($3))", row)
print(client.sql("SELECT id, title FROM py_quickstart.articles ORDER BY id"))
question = "I'm locked out of my account"
hits = client.sql("""SELECT title, COSINE_SIMILARITY(embedding, EMBED($1)) AS score
FROM py_quickstart.articles ORDER BY score DESC LIMIT 3""", [question])
print(f"\nTop matches for {question!r}:\n{hits}")
client.sql("DROP DATABASE py_quickstart")
client.close()export SYNAPCORES_USER=admin
export SYNAPCORES_PASSWORD='your-admin-password'
python quickstart.py id title
0 1 Reset your password
1 2 Update billing details
2 3 Export your data
3 4 Invite teammates
Top matches for "I'm locked out of my account":
title score
0 Reset your password 0.537934
1 Update billing details 0.199136
2 Invite teammates 0.174465The question and the top article share no keywords; the match comes from the embeddings. The script drops its database at the end, so you can run it again.
How it works
client.sql()sends a statement toPOST /v1/query/execute. Use$1,$2, … placeholders and pass the values as a list.- Rows come back as a pandas DataFrame, so you can filter, join or plot them straight away.
EMBED(text)creates a 384-dimension embedding with the local all-MiniLM model inside the engine, andCOSINE_SIMILARITY()ranks rows against the question. Your Python code never handles vectors.- Database-qualified names such as
py_quickstart.articleslet one client work across databases.
Connect with an API key
For services and scripts that run unattended, pass an API key instead of logging in (the MCP guide shows how to create one):
import os
from synapcores import SynapCores
client = SynapCores(host="localhost", port=8080, api_key=os.environ["SYNAPCORES_API_KEY"])
print(client.sql("SELECT 1 + 1 AS two")) two
0 2Go further
client.sql()runs any statement the engine accepts over REST. The SQL reference lists every function, from agent memory to AutoML.- Building a RAG app? The LlamaIndex guide stores chunks in the engine, retrieves them with a vector search like the one above and answers with a chat model the same engine serves through its Ollama-compatible API, so retrieval and generation stay on your machine.
Next steps
Full reference
- SQL reference: Every statement and function: EMBED, VECTOR, COSINE_SIMILARITY, GENERATE, Cypher, AutoML, agent memory.
- API reference: The REST API. Your engine also serves its own OpenAPI schema at /v1/openapi.json.
- System requirements: Supported platforms, memory and disk sizing.