RAGアプリやLLMエージェントをオーケストレーションできるフレームワーク「Dynamiq」を試す

$ docker run --rm -p 8888:8888 -v .:/home/jovyan/work quay.io/jupyter/minimal-notebook:latest

!pip install dynamiq
!pip freeze | grep -i dynamiq

dynamiq==0.3.0

import getpass
import os

os.environ["OPENAI_API_KEY"] = getpass.getpass('OPENAI_API_KEY')

from dynamiq.nodes.llms.openai import OpenAI
from dynamiq.connections import OpenAI as OpenAIConnection
from dynamiq import Workflow
from dynamiq.prompts import Prompt, Message
import json

# 翻訳用のプロンプトテンプレートを定義
prompt_template = """
Translate the following text into Japanese: {{ text }}
"""

# 定義されたテンプレートでプロンプトオブジェクトを作成
prompt = Prompt(messages=[Message(content=prompt_template, role="user")])

# LLMノードをセットアップ
llm = OpenAI(
    id="openai",                    # ノードの一意な識別子
    connection=OpenAIConnection(),  # APIキーを使って接続
    model="gpt-4o-mini",            # 使用するモデル
    temperature=0.3,                # モデルのtemperature
    max_tokens=1000,                # 最大出力トークン
    prompt=prompt                   # モデルに与えるプロンプト
)

# ワークフローオブジェクト
workflow = Workflow()

# ワークフローにLLMノードを追加
workflow.flow.add_nodes(llm)

# 入力データでワークフローを実行
result = workflow.run(
    input_data={
        "text": "Hola Mundo!"  # 翻訳されるテキスト
    }
)

# 翻訳結果を表示
print(json.dumps(result.output, indent=2, ensure_ascii=False))

{
  "openai": {
    "status": "success",
    "input": {
      "text": "Hola Mundo!"
    },
    "output": {
      "content": "こんにちは、世界！",
      "tool_calls": null
    }
  }
}

os.environ["E2B_API_KEY"] = getpass.getpass('E2B_API_KEY')

from dynamiq.nodes.llms.openai import OpenAI
from dynamiq.connections import OpenAI as OpenAIConnection, E2B as E2BConnection
from dynamiq.nodes.agents.react import ReActAgent
from dynamiq.nodes.tools.e2b_sandbox import E2BInterpreterTool

# E2Bツールを初期化
e2b_tool = E2BInterpreterTool(
    connection=E2BConnection()
)

# LLMをセットアップ
llm = OpenAI(
    id="openai",
    connection=OpenAIConnection(),
    model="gpt-4o-mini",
    temperature=0.3,
    max_tokens=1000,
)

# ReActエージェントを作成
agent = ReActAgent(
    name="react-agent",
    llm=llm,
    tools=[e2b_tool],
    role="シニアデータサイエンティスト",
    max_loops=10,
)

# 入力を元にエージェントを実行
result = agent.run(
    input_data={
        "input": "最初の10個の数字を計算し、その結果が素数であるかどうかを判定して。",
    }
)

print(result.output.get("content"))

The prime status of the first 10 natural numbers is as follows:
- 1: Not prime
- 2: Prime
- 3: Prime
- 4: Not prime
- 5: Prime
- 6: Not prime
- 7: Prime
- 8: Not prime
- 9: Not prime
- 10: Not prime

os.environ["TAVILY_API_KEY"] = getpass.getpass('TAVILY_API_KEY')

from dynamiq.connections import (OpenAI as OpenAIConnection,
                                 Tavily as TavilyConnection,
                                 E2B as E2BConnection)
from dynamiq.nodes.llms import OpenAI
from dynamiq.nodes.agents.orchestrators.adaptive import AdaptiveOrchestrator
from dynamiq.nodes.agents.orchestrators.adaptive_manager import AdaptiveAgentManager
from dynamiq.nodes.agents.react import ReActAgent
from dynamiq.nodes.agents.reflection import ReflectionAgent
from dynamiq.nodes.tools.e2b_sandbox import E2BInterpreterTool
from dynamiq.nodes.tools.tavily import TavilyTool


# ツールの初期化
python_tool = E2BInterpreterTool(
    connection=E2BConnection()
)
search_tool = TavilyTool(
    connection=TavilyConnection()
)

# LLMを初期化
llm = OpenAI(
    connection=OpenAIConnection(),
    model="gpt-4o-mini",
    temperature=0.1,
)

# エージェントを定義
coding_agent = ReActAgent(
    name="coding-agent",
    llm=llm,
    tools=[python_tool],
    role=(
        "コーディングスキルを持つエキスパート・エージェント。"
        "ゴールは、Pythonのソフトウェアエンジニアリングスキルを使って、"
        "入力タスクに対するソリューションを提供すること。"
    ),
    max_loops=15,
)

planner_agent = ReflectionAgent(
    name="planner-agent",
    llm=llm,
    role=(
        "企画力のあるエキスパートエージェント。"
        "ゴールは、複雑な要求を分析し、詳細な行動計画を提供すること。"
    ),
)

search_agent = ReActAgent(
    name="search-agent",
    llm=llm,
    tools=[search_tool],
    role=(
        "Web検索スキルを持つエキスパート・エージェント。"
        "ゴールは、ウェブ検索と要約のスキルを使って、"
        "入力タスクの解決策を提供すること"
    ),
    max_loops=10,
)

# adaptive agent managerを初期化
agent_manager = AdaptiveAgentManager(llm=llm)

# オーケストレータを作成
orchestrator = AdaptiveOrchestrator(
    name="adaptive-orchestrator",
    agents=[coding_agent, planner_agent, search_agent],
    manager=agent_manager,
)

# 入力タスクの定義
input_task = (
    "コーディングスキルを使って、過去10年間のNvidiaとIntelの株価に関するデータを収集し、"
    "各社の年平均を計算し、表を作成する。そしてレポートを作成し、10年前に100ドルを投資して"
    "いたらどうなっていたか、という結論を加える。"
)

# オーケストレータを実行
result = orchestrator.run(
    input_data={"input": input_task},
)

# 結果を出力
print(result.output.get("content"))

### Final Report on Nvidia and Intel Stock Performance Over the Past Decade

#### Introduction
This report analyzes the stock performance of Nvidia Corporation (NVDA) and Intel Corporation (INTC) over the past ten years. It includes the calculation of annual average stock prices for both companies and evaluates the potential returns on a hypothetical investment of $100 made ten years ago.

#### Annual Average Stock Prices
The following table summarizes the annual average stock prices for Nvidia and Intel from 2014 to 2024:

| Year | Average Price NVDA | Average Price INTC |
|------|---------------------|---------------------|
| 2014 | $0.504400           | $35.669778          |
| 2015 | $0.592956           | $32.179127          |
| 2016 | $1.344030           | $33.329127          |
| 2017 | $3.744593           | $37.794661          |
| 2018 | $5.809604           | $49.054980          |
| 2019 | $4.364692           | $51.375198          |
| 2020 | $9.890709           | $54.892055          |
| 2021 | $19.521993          | $55.851786          |
| 2022 | $18.568956          | $38.594502          |
| 2023 | $36.569260          | $33.826800          |
| 2024 | $100.900029         | $32.910000          |

#### Investment Analysis
To assess the investment potential of both stocks, we consider a hypothetical investment of $100 made ten years ago (in 2014):

- **Nvidia (NVDA)**: 
  - The average price in 2014 was approximately $0.5044. 
  - With an investment of $100, the number of shares purchased would be:
    \[
    \text{Shares} = \frac{100}{0.5044} \approx 198.51 \text{ shares}
    \]
  - As of October 2024, the price of Nvidia stock is approximately $100.90.
  - The current value of the investment would be:
    \[
    \text{Current Value} = 198.51 \times 100.90 \approx 20,073.59
    \]

- **Intel (INTC)**:
  - The average price in 2014 was approximately $35.6698.
  - With an investment of $100, the number of shares purchased would be:
    \[
    \text{Shares} = \frac{100}{35.6698} \approx 2.80 \text{ shares}
    \]
  - As of October 2024, the price of Intel stock is approximately $32.91.
  - The current value of the investment would be:
    \[
    \text{Current Value} = 2.80 \times 32.91 \approx 92.15
    \]

#### Conclusion
The analysis reveals a stark contrast in the investment performance of Nvidia and Intel over the past decade. If $100 had been invested in Nvidia ten years ago, it would be worth approximately **$20,073.59** today, reflecting a remarkable return on investment. In contrast, the same investment in Intel would yield a value of approximately **$92.15**, indicating minimal growth.

This significant disparity highlights Nvidia's exceptional growth trajectory compared to Intel, making it a more lucrative investment option over the past ten years. Investors should consider these trends when evaluating future investment opportunities in the semiconductor industry.
Click to add a cell.