rail-engine 0.1.1

Python SDK for Railtown AI Rail Engine - Retrieval

Rail Engine Retrieval SDK

Python SDK for retrieving and searching data from Railtown AI Rail Engine - handles embeddings, storage documents, and indexed content.

Overview

The rail-engine package provides a Pythonic interface for retrieving and searching data from Rail Engine. It supports async/await patterns, client-side filtering, automatic pagination, and Pydantic model deserialization.

Installation

pip install rail-engine

Or using uv:

uv pip install rail-engine

Quick Start

import asyncio
from railtown.engine import RailEngine

async def main():
    # Initialize client (reads from ENGINE_PAT and ENGINE_ID env vars)
    async with RailEngine() as client:
        # Search vector store
        results = client.search_vector_store(
            engine_id=client.engine_id,
            vector_store="VectorStore1",
            query="apple"
        )
        async for item in results:
            print(item)

asyncio.run(main())

Configuration

Environment Variables

• ENGINE_PAT (required) - Personal Access Token
• ENGINE_ID (required) - Engine ID (can also be passed to constructor)
• RAILTOWN_API_URL (optional) - Base API URL (defaults to https://cndr.railtown.ai/api)

Constructor Parameters

• pat (optional) - PAT token (if not provided, reads from ENGINE_PAT env)
• engine_id (optional) - Engine ID (if not provided, reads from ENGINE_ID env, required if not in env)
• api_url (optional) - Base API URL (if not provided, reads from RAILTOWN_API_URL env or defaults to production)
• model (optional) - Pydantic model type for deserializing retrieved data

Features

• Multiple retrieval methods:
  • search_vector_store() - Semantic search in vector stores
  • get_storage_document_by_id() - Get document by ID
  • get_storage_document_by_customer_key() - Get documents by customer key
  • query_storage_by_jsonpath() - Query using JSONPath
  • list_storage_documents() - List all documents with pagination
  • search_index() - Full-text search using Azure Search
• Client-side filtering - Filter results using filter_fn parameter
• Automatic pagination - Handles pagination automatically
• Model deserialization - Optional Pydantic model support with per-call override
• Graceful error handling - Returns None or empty iterables on errors
• Async/await support - Built for modern async Python applications

API Reference

RailEngine Client

Methods

• search_vector_store(engine_id, vector_store, query, filter_fn=None, model=None) - Search vector store
• get_storage_document_by_id(engine_id, engine_document_id, filter_fn=None, model=None) - Get document by ID
• get_storage_document_by_customer_key(engine_id, customer_key, page_number=1, page_size=25, filter_fn=None, model=None) - Get documents by customer key
• query_storage_by_jsonpath(engine_id, json_path_query, filter_fn=None, model=None) - Query by JSONPath
• list_storage_documents(engine_id, customer_key=None, page_number=1, page_size=100, filter_fn=None, model=None) - List documents
• search_index(project_id, engine_id, query, filter_fn=None, model=None) - Search index

Properties

• pat - PAT token
• engine_id - Engine ID
• api_url - Base API URL
• model - Default model type

Examples

See the samples/ directory for comprehensive examples:

• samples/retrieval_example.py - Retrieval examples
• samples/README.md - Samples documentation

Error Handling

The SDK provides custom exception classes:

• RailtownError - Base exception
• RailtownBadRequestError - 400 Bad Request
• RailtownUnauthorizedError - 401 Unauthorized
• RailtownNotFoundError - 404 Not Found
• RailtownConflictError - 409 Conflict
• RailtownServerError - 5xx Server errors

Returns None or empty iterables on errors (graceful degradation).

Requirements

• Python 3.10+
• httpx >= 0.24.0
• pydantic >= 1.10.0

Related Package

For ingesting data into Rail Engine, see rail-engine-ingest.

License

MIT

Support

For issues, questions, or contributions, please visit the GitHub repository.