Scott O.

Hi Paul - This is what we have in documentation: SnapCode is a conversational, developer-oriented interface for building, deploying, and running SnapLogic pipelines using plain English — powered by Claude Code (by Anthropic). It was introduced in the July 2026 release. Think of it as a code-first complement to SnapLogic Designer — instead of dragging Snaps onto a canvas, you describe what you want and SnapCode handles the rest. ⚙️ Prerequisites Before installing SnapCode, you'll need: Requirement Details Docker Desktop Installed and running Git To clone the SnapCode repository GitHub Account With a classic Personal Access Token (PAT) with read:packages scope Claude Code Account An Anthropic subscription (or AWS Bedrock as an alternative) SnapLogic Credentials Platform URL, API username (email), and API password ⚠️ Note: SSO and MFA authentication are not currently supported. 🛠️ Installation SnapCode supports three setup environments: 1. VS Code Dev Container – Best for macOS/Linux users working in VS Code 2. Terminal / Docker – Best for Windows users or anyone preferring a terminal 3. Local no-Docker mode – Via a setup script Basic Steps: 1. Clone the SnapCode repository 2. Create your credentials file by copying .env.example to .env and filling in your SnapLogic (and optionally AWS) credentials 3. Run the installer for your environment: macOS/Linux: ./install.sh docker or ./install.sh vscode Windows: .\install.cmd docker or .\install.cmd vscode 4. Start SnapCode by running claude from the sl-procode directory 5. Type /mcp to confirm the SnapCode MCP servers are connected 💬 How to Use SnapCode Once running, simply describe your pipeline in plain English: Build a pipeline that reads customer orders from Salesforce, filters cancelled ones, and writes the results to Snowflake SnapCode will automatically: 1. 🔍Discover Snaps – Looks up correct Snap names and parameters 2. 📝Generate a .py source file – Written in SLPy (readable Python-like format) 3. 🔄Translate to .slp – Converts to SnapLogic's native JSON format 4. 🧪Run a local test – Validates transformation logic locally 5. 🚀Deploy to your org – Uploads to your SnapLogic environment 6. ✅Validate on the platform – Equivalent to clicking Validate in Designer 📁 File Types Extension Description .py SLPy pipeline source — your editable source of truth .slp SnapLogic's native JSON format (auto-generated) .expr Reusable expression libraries ⚠️ Known Limitations Limitation Details Connection accounts Cannot create/configure accounts (Salesforce, Snowflake, etc.) — must be set up manually in Designer SSO/MFA Not supported for authentication Downloading existing pipelines Cannot pull pipelines already on the platform (export manually from Designer first) Execution history Can only show status of the run it just triggered Session state Environment/project/Snaplex preferences reset each new session Getting Started - docs.snaplogic.com/snapcode/snapcode-about.html

Hi Jean-Francois - Other than Match, I queried docs and SnapGPT and other options are: Option 2: Deduplicate Snap (Machine Learning Snap Pack) The Deduplicate Snap uses the same set of comparators (including Levenshtein) but is intended for deduplication within a single dataset. It can still be leveraged to identify how "different" records are from each other with a configurable threshold, but it's less suited for general-purpose pairwise string comparison. Option 3: Script Snap (Custom Workaround for Raw Distance) If you need the raw Levenshtein integer distance (e.g., "kitten" vs "sitting" = 3), you can implement it using a Script Snap with Python or JavaScript: Example (Python-style pseudologic inside a Script Snap): import Levenshtein # if available, or implement manually distance = Levenshtein.distance(input['string1'], input['string2']) output['levenshtein_distance'] = distance Note: Library availability depends on your Snaplex configuration. A custom implementation of the algorithm can also be coded directly in the script. For most use cases involving quantitative string similarity, the Match Snap from the Machine Learning Snap Pack is the recommended starting point due to its built-in support for Levenshtein and many other algorithms.