Skip to content

Changelog

All notable changes to Ellie's Postgres Cookbook will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[Unreleased]

Added

  • Four SQL recipes for the Snowflake 2.6.0 extension: Generating Snowflake IDs, Decoding a Snowflake ID, Converting a Sequence to Snowflake, and Checking the Node Number.
  • Six SQL recipes for pgEdge Vectorizer 1.1: Vectorizing a Table, Searching Vectorized Text, Combining Keyword and Vector Search, Splitting Text into Chunks, Monitoring the Embedding Queue, and Keeping Embeddings Up to Date.
  • Five SQL recipes for pg_semantic_cache 0.1.0-beta4: Caching a Query Result, Finding an Answer to a Reworded Question, Expiring Cached Results, Invalidating Cached Results by Tag, and Measuring the Cache Hit Rate.
  • Five recipes for the pgEdge Postgres MCP Server 1.1.0: Connecting to the MCP Server, Exploring a Schema, Running Read-Only Queries, Searching by Meaning, and Reading Resources and Prompts.
  • Five recipes for the pgEdge Control Plane v0.11.0 API: Checking the Cluster, Creating a Database, Connecting to a Database, Following a Task, and Changing a Database's Settings.
  • A test server image built from pgEdge Postgres with the extensions installed, scripts that start the Postgres MCP Server and the Control Plane in containers, and continuous integration jobs for the extension, Postgres MCP Server, and Control Plane recipes.
  • Nine recipes for the pgEdge RAG Server 2.0.0 API: Preparing Data for the RAG Server, Asking a Question, Streaming an Answer, Showing the Sources of an Answer, Filtering Retrieved Documents, Holding a Conversation, Checking Health and Listing Pipelines, Handling RAG Server Errors, and Limiting Answers per User.
  • Seven recipes for the pgEdge AI DBA Workbench 1.0.0 API: Authenticating, Registering a Server, Listing and Acknowledging Alerts, Querying Metrics, Scheduling a Maintenance Blackout, Calling MCP Tools, and Chatting with the LLM.
  • Seven recipes for the pgEdge Starfleet API: Getting an Access Token, Listing Regions, Sizes, and Versions, Creating a Database and Waiting for It, Connecting to a Starfleet Database, Creating a Branch for a Test Run, Taking a Backup and Tracking Its Task, and Rotating a Role Password.
  • A language set for HTTP API recipes (Python, TypeScript, Java, C#, Go, Rust, and Ruby, which is every language except C) with a curl tab, --dir and --exclude options for the test runner, scripts that start the RAG Server and AI DBA Workbench in containers, and continuous integration jobs for each product, with the Starfleet job run weekly or by hand when its credentials are configured.
  • Eleven PostgreSQL recipes: Passing a List of Values, Using Transactions, Inserting Rows and Reading Generated Keys, Bulk Loading with COPY, Reading Large Results in Batches, Using a Connection Pool (which has no C program), Storing and Searching Vectors with pgvector, Listening for Notifications, Setting a Statement Timeout, Retrying Serialisation Failures, and Connecting Securely with TLS.
  • Eight Go-only recipes for the pgEdge Go LLM Library: Sending a Chat Request, Streaming a Response, Handling Provider Errors and Retries, Storing Embeddings in PostgreSQL, Answering Questions from Your Own Data, Reranking Search Results, Letting a Model Query the Database, and Extracting Structured Data into a Table.
  • Support for recipes whose output varies (check_output), a custom unreachable-server environment (failure_env.txt), and a TLS-enabled test server script.
  • The Connecting to PostgreSQL and Running a Query recipe, in Python, TypeScript, Java, C#, Go, Rust, Ruby, and C.
  • Continuous integration that tests every recipe against PostgreSQL.
  • A publish workflow that renders the documentation, with the recipe code inlined, to the published branch for the primary documentation site.