Phidata Refresher

Agents with built-in Memory (Postgres) and Knowledge.

1. Philosophy

Phidata solves the "Amensia" problem. It connects Agents directly to a database (Postgres) to store sessions, and uses the same database for Vector Search (pgvector).

Technique: Persistent Storage

from phi.assistant import Assistant
from phi.storage.agent.postgres import PgAgentStorage

# 1. Define Storage (Postgres)
storage = PgAgentStorage(
    table_name="agent_sessions",
    db_url="postgresql://user:pass@localhost:5432/db"
)

# 2. Agent with Storage
agent = Assistant(
    name="MemoryAgent",
    storage=storage,
    add_history_to_messages=True, # Sends chat history to LLM
)

# 3. Resume Session
# If I run this with session_id="123", it loads previous chats.
agent.print_response("Hello", session_id="123")

Technique: Knowledge Base

from phi.knowledge.pdf import PDFUrlKnowledgeBase
from phi.vectordb.pgvector import PgVector2

kb = PDFUrlKnowledgeBase(
    urls=["https://.../report.pdf"],
    vector_db=PgVector2(collection="reports", db_url="...")
)
kb.load()

agent = Assistant(knowledge_base=kb, search_knowledge=True)