Skip to content

README.md

Latest commit

 

History

History
134 lines (98 loc) · 4.55 KB

README.md

File metadata and controls

134 lines (98 loc) · 4.55 KB

Synalinks Memory Python SDK

Synalinks Memory is the knowledge and context layer for AI agents. It lets your agents always have the right context at the right time. Unlike retrieval systems that compound LLM errors at every step, Synalinks uses logical rules to derive knowledge from your raw data. Every claim can be traced back to evidence, from raw data to insight, no more lies or hallucinations.

This SDK provides a Python client to interact with the Synalinks Memory API, so your agents can store, query, and reason over their knowledge base programmatically.

Installation

pip install synalinks-memory

Or with uv:

uv add synalinks-memory

Quick Start

A Synalinks API key is required to authenticate with your knowledge base.

When you create a knowledge base on app.synalinks.com, a default API key is generated automatically with read/write access and no predicate restrictions — you can use it right away.

To create a key with granular access, go to Profile icon (in the header) > API Keys > Create API Key.

Set your API key as an environment variable:

export SYNALINKS_API_KEY="synalinks_..."

Then query your data:

from synalinks_memory import SynalinksMemory

with SynalinksMemory() as client:
    # List all available tables, concepts, and rules
    predicates = client.list()
    for table in predicates.tables:
        print(f"{table.name}: {table.description}")

    # Fetch rows from a table
    result = client.execute("Users", limit=10)
    for row in result.rows:
        print(row)

    # Search with keywords (fuzzy matching)
    result = client.search("Users", "alice")
    for row in result.rows:
        print(row)

    # Upload a CSV or Parquet file
    upload = client.upload("data/sales.csv", name="Sales", description="Monthly sales data")
    print(f"Uploaded {upload.predicate} ({upload.row_count} rows)")

    # Insert a row
    client.insert("Users", {"name": "Alice", "email": "alice@example.com"})

    # Update rows matching a filter
    result = client.update("Users", filter={"name": "Alice"}, values={"email": "alice@new.com"})
    print(f"Updated {result.updated_count} row(s)")

    # Export data as a file (CSV, Parquet, or JSON)
    client.execute("Users", format="csv", output="users.csv")
    client.execute("Users", format="parquet", output="users.parquet")

    # Chat with the agent (multi-turn)
    answer = client.chat("What were the top 5 products by revenue last month?")
    print(answer)

    # Follow-up question (uses conversation context automatically)
    answer = client.chat("Show me just the top 3")
    print(answer)

    # Reset conversation history
    client.clear()

You can also pass the key directly:

client = SynalinksMemory(api_key="synalinks_...")

Error Handling

from synalinks_memory import (
    SynalinksMemory,
    AuthenticationError,
    NotFoundError,
    RateLimitError,
)

with SynalinksMemory() as client:
    try:
        result = client.execute("MyTable")
    except AuthenticationError:
        print("Invalid API key")
    except NotFoundError as e:
        print(f"Not found: {e.message}")
    except RateLimitError as e:
        print(f"Rate limited, retry after {e.retry_after}s")

API Reference

SynalinksMemory(api_key=None, base_url=None, timeout=30.0)

Parameter Description
api_key Your API key. If omitted, reads from SYNALINKS_API_KEY env var.
base_url Override the API endpoint (defaults to https://app.synalinks.com/api).
timeout Request timeout in seconds.

Methods

Method Description
list() List all tables, concepts, and rules
execute(predicate, *, limit=100, offset=0, format=None, output=None) Fetch rows (or export as json/csv/parquet file when format is set)
search(predicate, keywords, *, limit=100, offset=0) Search rows by keywords (fuzzy matching)
upload(file_path, *, name=None, description=None, overwrite=False) Upload a CSV or Parquet file as a new table
insert(predicate, row) Insert a single row into a table
update(predicate, filter, values) Update rows matching a filter with new values
chat(question) Chat with the agent (multi-turn), returns the answer string
clear() Reset conversation history for a fresh chat
close() Close the HTTP client (not needed with with statement)

License

Licensed under Apache 2.0. See the LICENSE file for full details.