Skip to content

lilyjazz/agent-working-memory

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

14 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

🧠 Agent Working Memory

The Serverless Working Memory for AI Agents. Powered by TiDB Cloud Zero.

Just like humans use "Working Memory" to process tasks, solve problems, and hold temporary context, AI Agents need a place to:

  1. Verify Code: Run SQL in a sandbox before outputting it.
  2. Process Data: Clean and analyze CSVs using SQL power.
  3. Hold Context: Store logs, RAG vectors, and state across sessions.

This repo provides standardized "Skills" that give your Agent this capability instantly.

📦 What's Inside?

A collection of self-contained skills. Each skill provisions a serverless TiDB database on-the-fly.

[Phase 1: The Code Verifier]

  • verify-code: The Agent runs generated SQL in a disposable DB to fix syntax errors before they reach the user. (Status: Ready)

[Phase 2: The Data Analyst] (Coming Soon)

  • data-refinery: Load messy CSVs -> SQL Cleaning -> Clean Export.
  • data-diff: Compare two datasets using SQL set operations.

[Phase 3: The Second Brain] (Coming Soon)

  • knowledge-vault: Instant Vector Store for document analysis.
  • agent-teleport: Serialize agent state to cloud for cross-device migration.

🚀 Quick Start

Each skill is standalone. To give your agent the power to verify SQL:

# Install dependencies
pip install pymysql

# Run the skill (Agent simulates this)
python skills/verify-code/run.py --sql "CREATE TABLE test (id INT); INSERT INTO test VALUES (1);"

Output:

{
  "verified": true,
  "infrastructure": "Powered by TiDB Cloud Zero"
}

🏗 Architecture

  • Zero Config: No Docker, no cloud accounts required for the Agent.
  • Ephemeral: Databases auto-recycle after 30 days.
  • Standardized: All skills output strict JSON for easy parsing.

📄 Documentation

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages