📝 Description

This PR enhances the memory persistence mechanism in the graph builder by:

✅ Default Behavior

• Replaces the default in-memory MemorySaver with an SQLite-based checkpointer using langgraph-checkpoint-sqlite.
• SQLite runs in local :memory: mode for dev/testing, ensuring persistence without external services.

🛠️ Multi-Database Support

• Introduces configurable support for several database backends for checkpointing:
  • PostgreSQL (langgraph-checkpoint-postgres)
  • MongoDB (langgraph-checkpoint-mongodb)
  • Redis (langgraph-checkpoint-redis)
• Backend selection and connection string setup are now driven by conf.yaml:

  CHECKPOINTER_MEMORY:
    db_type: "postgres"        # or "sqlite", "mongo", "redis"
    db_uri: "postgresql://user:pass@localhost:5432/dbname"

🧱 Code Structure

• Refactored builder.py to dynamically instantiate checkpointer based on config.
• Reorganized app.py to use the unified graph builder pattern with the new checkpointer.

📘 Docs & Examples

• Extended docs/configuration_guide.md with setup instructions for each supported DB.
• Updated pyproject.toml and dependencies accordingly.

✅ Benefits

• Persistent agent memory across sessions.
• Flexible deployment in cloud or local environments.
• Easy extension for future database types.