Store and query AI agent memories in PostgreSQL with full-text search.
npx skills add https://github.com/pascalandy/dotfiles --skill pg-memory
Please help me install this Agent Skill. Skill: pg-memory Source: https://github.com/pascalandy/dotfiles/tree/main/dot_config/opencode/skill/util-pg-memory Command: npx skills add https://github.com/pascalandy/dotfiles --skill pg-memory
The pg-memory skill provides a robust, PostgreSQL-backed store for AI agent memories, enabling durable capture of decisions, observations, and learnings to improve consistency and auditing across sessions.
Install PostgreSQL 18+, create a database, apply the pg-memory schema, and verify with the included verification script. Then insert a memory and run a search to confirm the system is working.
High-intent search queries and answers about installing and using this skill.
To persist AI memories in PostgreSQL, you store decisions, observations, and learnings using a dedicated schema with JSONB metadata and tsvector indexing. This enables durable cross-session context retrieval and full-text search for auditability.
Querying past AI decisions and observations is done using full-text search and JSONB metadata filtering within PostgreSQL. This allows you to retrieve specific historical context and rationale to inform current reasoning workflows.
PostgreSQL 18+ is required to use this memory storage approach, as the schema relies on advanced indexing features like tsvector for full-text search and JSONB for fast metadata lookups to capture and query AI memories.
Setting up PostgreSQL for AI memory retrieval involves installing a PostgreSQL 18+ instance, creating a database, and applying a specific schema with proper indexing. You then verify the setup with a script before inserting searchable memories.
Yes, token and cost tracking are supported through optional integration points when storing AI memories in PostgreSQL. This allows you to accumulate a searchable history of prompts and results while monitoring usage metrics.
Using PostgreSQL for AI memory management requires maintaining a PostgreSQL 18+ instance and managing schema indexing like tsvector. It is not a lightweight in-memory store, meaning it requires database administration overhead for memory retrieval.
Mentor-WizardDesign PostgreSQL database schemas with data types, indexing, and constraints.
softsparkSearch past coding session memories using SQLite FTS5 full-text queries.
MicrockTrack LLM token usage and calculate costs for billing reports.
Datus-aiInitialize data engineering projects with lightweight context stores and AGENTS.md skeletons.
