Skip to content

Database Tools - SQL and a VectorStore Together ​

examples/patterns/database-tools.halo

How to chain a SQL lookup into a vector search Fetch a user's row from Postgres, look up the policies that apply to their type, and have an analyst summarize both.

Open in the playground and play with it →

halo
// ============================================================
// Database Tools - SQL and a VectorStore Together
// ============================================================
// LEARN: How to chain a SQL lookup into a vector search
// WHY:   Fetch a user's row from Postgres, look up the policies that
//        apply to their type, and have an analyst summarize both.
// ============================================================

tool UsersDB : PostgreSQL("postgres://user:pass@localhost:5432/mydb")
tool KnowledgeBase : VectorStore("index-123")

agent Analyst [openai/gpt-4]
  equips UsersDB, KnowledgeBase

pipeline AnalyzeUser(userId: String) -> Result:
  userId
    |> UsersDB.execute(query: "SELECT * FROM users WHERE id = '{{context.input.userId}}'")
    |> KnowledgeBase.execute(method: "search", query: "Policies for user type: {{context.current.type}}")
    |> Analyst.prompt("Summarize the user profile and relevant policies. User: {{context.data.UsersDB.execute}}, Policies: {{context.current}}")

curl -fsSL https://halolang.com/install.sh | sh