Skip to main content
Turn a user’s preferences into personalised, up-to-date suggestions with a single Agent that searches the web. 
from praisonaiagents import Agent
from praisonaiagents.tools import duckduckgo

agent = Agent(
    name="Recommender",
    instructions="Provide personalized suggestions based on preferences.",
    tools=[duckduckgo],
)

agent.start("Recommend 5 sci-fi movies from 2024")
Recommendation agent with web search for personalized suggestions.

​
Quick Start

1

Simple Usage

Attach a search tool and describe what you like.
from praisonaiagents import Agent
from praisonaiagents.tools import duckduckgo

agent = Agent(
    name="Recommender",
    instructions="Provide personalized suggestions based on preferences.",
    tools=[duckduckgo],
)

agent.start("Recommend 5 sci-fi movies from 2024")
2

With Configuration

Enable memory so recommendations improve over time.
from praisonaiagents import Agent
from praisonaiagents.tools import duckduckgo

agent = Agent(
    name="Recommender",
    instructions="Learn user taste and refine recommendations each turn.",
    tools=[duckduckgo],
    memory=True,
)

agent.start("Recommend sci-fi movies, then remember I dislike horror.")

​
How It Works

​
Simple

Agents: 1 — Single agent analyzes preferences and generates recommendations.

​
Workflow

  1. Receive user preferences
  2. Search for current options
  3. Generate personalized recommendations

​
Setup

pip install praisonaiagents praisonai duckduckgo-search
export OPENAI_API_KEY="your-key"

​
Run — Python

from praisonaiagents import Agent
from praisonaiagents.tools import duckduckgo

agent = Agent(
    name="Recommender",
    instructions="Provide personalized suggestions based on preferences.",
    tools=[duckduckgo]
)

result = agent.start("Recommend 5 sci-fi movies from 2024")
print(result)

​
Run — CLI

praisonai "Recommend good books about AI" --web-search

​
Run — agents.yaml

framework: praisonai
topic: Recommendations
roles:
  recommender:
    role: Recommendation Specialist
    goal: Generate personalized recommendations
    backstory: You are an expert at finding great content
    tools:
      - duckduckgo
    tasks:
      recommend:
        description: Recommend 5 sci-fi movies from 2024
        expected_output: A list of recommendations

praisonai agents.yaml

​
Serve API

from praisonaiagents import Agent
from praisonaiagents.tools import duckduckgo

agent = Agent(
    name="Recommender",
    instructions="You are a recommendation agent.",
    tools=[duckduckgo]
)

agent.launch(port=8080)

curl -X POST http://localhost:8080/chat \
  -H "Content-Type: application/json" \
  -d '{"message": "Recommend podcasts about technology"}'

​
Advanced Workflow (All Features)

Agents: 1 — Single agent with memory, persistence, structured output, and session resumability.

​
Workflow

  1. Initialize session for preference tracking
  2. Configure SQLite persistence for recommendation history
  3. Search and recommend with structured output
  4. Store preferences in memory for personalization
  5. Resume session for refined recommendations

​
Setup

pip install praisonaiagents praisonai duckduckgo-search pydantic
export OPENAI_API_KEY="your-key"

​
Run — Python

from praisonaiagents import Agent, Task, AgentTeam, Session
from praisonaiagents.tools import duckduckgo
from pydantic import BaseModel

class Recommendation(BaseModel):
    category: str
    items: list[str]
    descriptions: list[str]
    ratings: list[str]

session = Session(session_id="rec-001", user_id="user-1")

agent = Agent(
    name="Recommender",
    instructions="Generate structured recommendations.",
    tools=[duckduckgo],
    memory=True
)

task = Task(
    description="Recommend 5 sci-fi movies from 2024 with ratings",
    expected_output="Structured recommendations",
    agent=agent,
    output_pydantic=Recommendation
)

agents = AgentTeam(
    agents=[agent],
    tasks=[task],
    memory=True
)

result = agents.start()
print(result)

​
Run — CLI

praisonai "Recommend sci-fi movies" --web-search --memory --verbose

​
Run — agents.yaml

framework: praisonai
topic: Recommendations
memory: true
memory_config:
  provider: sqlite
  db_path: recommendations.db
roles:
  recommender:
    role: Recommendation Specialist
    goal: Generate structured recommendations
    backstory: You are an expert at finding great content
    tools:
      - duckduckgo
    memory: true
    tasks:
      recommend:
        description: Recommend 5 sci-fi movies from 2024
        expected_output: Structured recommendations
        output_json:
          category: string
          items: array
          descriptions: array
          ratings: array

praisonai agents.yaml --verbose

​
Serve API

from praisonaiagents import Agent
from praisonaiagents.tools import duckduckgo

agent = Agent(
    name="Recommender",
    instructions="Generate structured recommendations.",
    tools=[duckduckgo],
    memory=True
)

agent.launch(port=8080)

curl -X POST http://localhost:8080/chat \
  -H "Content-Type: application/json" \
  -d '{"message": "Recommend books", "session_id": "rec-001"}'

​
Monitor / Verify

praisonai "test recommendations" --web-search --verbose

​
Cleanup

rm -f recommendations.db

​
Features Demonstrated

FeatureImplementation
WorkflowPersonalized recommendation generation
DB PersistenceSQLite via memory_config
Observability--verbose flag
ToolsDuckDuckGo search
ResumabilitySession with session_id
Structured OutputPydantic Recommendation model

​
Best Practices

Spell out likes, dislikes, and constraints in the prompt. Vague requests yield generic lists; concrete taste signals produce recommendations users act on.
Set memory=True so the agent remembers taste across sessions and stops re-recommending items the user already rejected.
Use output_pydantic with items, descriptions, and ratings fields so a UI can render cards or a table directly.
When a recommendation turns into a buy, switch to the Shopping Agent to compare prices across stores.

Compare prices once a user picks an item.

Research options in depth before recommending.