🦔

50.browser_use試してみる

2025/05/26に公開

仮想環境作成と有効化

python3 -m venv ./venv
source ./venv/bin/activate

インストール

pip install browser-use
playwright install chromium --with-deps --no-shell

`main.py`作成

import asyncio
from dotenv import load_dotenv
load_dotenv()
from browser_use import Agent
from langchain_openai import ChatOpenAI

async def main():
    agent = Agent(
        task="Compare the price of gpt-4o and DeepSeek-V3",
        llm=ChatOpenAI(model="gpt-4o"),
    )
    await agent.run()

asyncio.run(main())

内容は公式そのまま

`.env`作成

OPENAI_API_KEY=
ANTHROPIC_API_KEY=
AZURE_OPENAI_ENDPOINT=
AZURE_OPENAI_KEY=
GOOGLE_API_KEY=
DEEPSEEK_API_KEY=
GROK_API_KEY=
NOVITA_API_KEY=

使用するAPI_KEYを記述

実行

python3 main.py

結果

% python3 main.py

INFO     [telemetry] Anonymized telemetry enabled. See https://docs.browser-use.com/development/telemetry for more information.
INFO     [agent] 🧠 Starting a browser-use agent v0.2.4 with base_model=gpt-4o-mini +tools +vision +memory extraction_model=gpt-4o-mini 
path/venv/lib/python3.13/site-packages/faiss/loader.py:49: DeprecationWarning: numpy.core._multiarray_umath is deprecated and has been renamed to numpy._core._multiarray_umath. The numpy._core namespace contains private NumPy internals and its use is discouraged, as NumPy internals can change without warning in any release. In practice, most real-world usage of numpy.core is to access functionality in the public NumPy API. If that is the case, use the public NumPy API. If not, you are using NumPy internals. If you would still like to access an internal attribute, use numpy._core._multiarray_umath.__cpu_features__.
  from numpy.core._multiarray_umath import __cpu_features__
<frozen importlib._bootstrap>:488: DeprecationWarning: builtin type SwigPyPacked has no __module__ attribute
<frozen importlib._bootstrap>:488: DeprecationWarning: builtin type SwigPyObject has no __module__ attribute
<frozen importlib._bootstrap>:488: DeprecationWarning: builtin type swigvarlink has no __module__ attribute
INFO     [agent] 🚀 Starting task: Compare the price of gpt-4o-mini and DeepSeek-V3
INFO     [browser] 🌎 Launching local browser driver=playwright channel=chromium user_data_dir=~/.config/browseruse/profiles/default
INFO     [agent] 📍 Step 1: Evaluating page with 0 interactive elements on: about:blank
========================================================================================================================================================================================
INFO     [agent] 🧠 LLM call => ChatOpenAI [✉️ 7 msg, ~3153 tk, 30512 char, 📷 img] => JSON out + 🔨 20 tools (function_calling)
INFO     [agent] ❓ Eval: Unknown - The current state does not provide any elements to navigate or interact with, since it's an empty page. I will need to perform a search to gather information about the pricing.
INFO     [agent] 🧠 Memory: Current URL is about:blank with no interactive elements. I need to carry out a search for 'gpt-4o-mini price' and 'DeepSeek-V3 price'. I am at step 1/100 and no actions have been performed.
INFO     [agent] 🎯 Next goal: Perform a Google search for "gpt-4o-mini price" to gather pricing information.
INFO     [controller] 🔍  Searched for "gpt-4o-mini price" in Google
INFO     [agent] ☑️ Executed action 1/1: search_google
INFO     [agent] 📍 Step 2: Ran 1 actions in 9.72s: ✅ 1
INFO     [agent] 📍 Step 2: Evaluating page with 82 interactive elements on: https://www.google.com/search?q=gpt-4o-mini%20pric...
========================================================================================================================================================================================
INFO     [agent] 🧠 LLM call => ChatOpenAI [✉️ 10 msg, ~4484 tk, 615906 char, 📷 img] => JSON out + 🔨 20 tools (function_calling)
INFO     [agent] ❓ Eval: Unknown - The search for 'gpt-4o-mini price' is successful, but I still need to extract the relevant pricing information from the results.
INFO     [agent] 🧠 Memory: I have performed a search for 'gpt-4o-mini price' and located potential sources of information with a summary indicating low pricing of $0.04 and mentioning its comparison with GPT-4. However, I still need to check for the comparison price of DeepSeek-V3. Progress is at step 2/100, with no actions yet performed to extract content or check the price of DeepSeek-V3.
INFO     [agent] 🎯 Next goal: Extract relevant content from the search results for 'gpt-4o-mini price'.
INFO     [controller] 📄  Extracted from page
: json
{
  "extracted_info": {
    "gpt-4o-mini_price": {
      "input_cost_per_million_tokens": 0.15,
      "output_cost_per_million_tokens": 0.60,
      "comparison": "60% cheaper than GPT-3.5 Turbo"
    },
    "additional_sources": [
      {
        "source": "OpenAI API Pricing",
        "link": "https://openai.com/ja-JP/api/pricing"
      },
      {
        "source": "ChatGPT Pricing",
        "link": "https://openai.com/ja-JP/chatgpt/pricing"
      },
      {
        "source": "Chowagiken blog on API pricing",
        "link": "https://www.chowagiken.co.jp/blog/chatgpt_api"
      },
      {
        "source": "AIsmiley news",
        "link": "https://aismiley.co.jp/ニュース"
      }
    ]
  }
}


INFO     [agent] ☑️ Executed action 1/1: extract_content
INFO     [agent] 📍 Step 3: Ran 1 actions in 11.83s: ✅ 1
INFO     [agent] 📍 Step 3: Evaluating page with 82 interactive elements on: https://www.google.com/search?q=gpt-4o-mini%20pric...
================================================================
INFO     [agent] 🧠 LLM call => ChatOpenAI [✉️ 13 msg, ~5016 tk, 616655 char, 📷 img] => JSON out + 🔨 20 tools (function_calling)
INFO     [agent] 👍 Eval: Success - Extracted pricing information for gpt-4o-mini successfully.
INFO     [agent] 🧠 Memory: Extracted the following pricing data: 
- gpt-4o-mini:
  - Input cost per million tokens: $0.15
  - Output cost per million tokens: $0.60
  - Comparison: 60% cheaper than GPT-3.5 Turbo.

I need to now search for the pricing information for DeepSeek-V3. Progress: 3/100 actions completed.
INFO     [agent] 🎯 Next goal: Perform a Google search for "DeepSeek-V3 price" to gather pricing information.
INFO     [browser] 👁️ Foregound tab changed by human from [0]google.com/search?q=gp… ➡️ [1]google.com/search?q=De… (agent will stay on [1]google.com/search?q=De…)
INFO     [controller] 🔍  Searched for "DeepSeek-V3 price" in Google
INFO     [agent] ☑️ Executed action 1/1: search_google
INFO     [agent] 📍 Step 4: Ran 1 actions in 7.29s: ✅ 1
INFO     [agent] 📍 Step 4: Evaluating page with 76 interactive elements on: https://www.google.com/search?q=DeepSeek-V3%20pric...
===================================================================================================================
INFO     [agent] 🧠 LLM call => ChatOpenAI [✉️ 16 msg, ~5320 tk, 583363 char, 📷 img] => JSON out + 🔨 20 tools (function_calling)
INFO     [agent] ❓ Eval: Unknown - The search for 'DeepSeek-V3 price' is successful, but I still need to extract the relevant pricing information from the results.
INFO     [agent] 🧠 Memory: Current URL is the Google search results page for 'DeepSeek-V3 price'. I need to extract content to find the pricing information for DeepSeek-V3. So far, I have extracted the pricing for 'gpt-4o-mini'. Currently at step 4/100.
INFO     [agent] 🎯 Next goal: Extract relevant content from the search results for 'DeepSeek-V3 price'.
INFO     [controller] 📄  Extracted from page
: json
{
  "price_info": {
    "input_price_cache_hit": "$0.07 per 1M tokens",
    "input_price_cache_miss": "$0.27 per 1M tokens",
    "output_price": "$1.10 per 1M tokens",
    "additional_fees": "$0.0004 per request"
  }
}


INFO     [agent] ☑️ Executed action 1/1: extract_content
INFO     [agent] 📍 Step 5: Ran 1 actions in 6.58s: ✅ 1
INFO     [agent] 📍 Step 5: Evaluating page with 76 interactive elements on: https://www.google.com/search?q=DeepSeek-V3%20pric...
===================================================================================================================
INFO     [agent] 🧠 LLM call => ChatOpenAI [✉️ 19 msg, ~5646 tk, 583634 char, 📷 img] => JSON out + 🔨 20 tools (function_calling)
INFO     [agent] ❓ Eval: Unknown - The URL has elements but the current state does not directly specify the needed pricing information. I need to extract details from the search result about DeepSeek-V3 prices for comparison.
INFO     [agent] 🧠 Memory: Gathered prices for gpt-4o-mini: $0.15 per million tokens (input), $0.60 per million tokens (output). 
Extracted prices for DeepSeek-V3 indicating input prices of $0.07 (cache hit) and $0.27 (cache miss), output price $1.10, and additional fees, which need to be consolidated for comparison. Progress: 6/100 actions completed.
INFO     [agent] 🎯 Next goal: Extract relevant pricing information regarding DeepSeek-V3 from the search results.
INFO     [controller] 📄  Extracted from page
: json
{
  "DeepSeek_V3_Price_Details": {
    "Cache_Hit": {
      "Input_Cost": "$0.07 per 1M tokens",
      "Output_Cost": "$1.10 per 1M tokens"
    },
    "Cache_Miss": {
      "Input_Cost": "$0.27 per 1M tokens",
      "Output_Cost": "$0.00456 per 1K tokens"
    },
    "General_Charges": {
      "Input_Tokens": "$0.000140 per 1,000 tokens",
      "Output_Tokens": "$0.000280 per 1,000 tokens",
      "Additional_Fee": "$0.0004"
    }
  }
}


INFO     [agent] ☑️ Executed action 1/1: extract_content
INFO     [agent] 📍 Step 6: Ran 1 actions in 25.17s: ✅ 1
INFO     [agent] 📍 Step 6: Evaluating page with 76 interactive elements on: https://www.google.com/search?q=DeepSeek-V3%20pric...
===================================================================================================================
INFO     [agent] 🧠 LLM call => ChatOpenAI [✉️ 22 msg, ~6106 tk, 584125 char, 📷 img] => JSON out + 🔨 20 tools (function_calling)
INFO     [agent] ❓ Eval: Unknown - Current state shows interaction elements but I need to extract relevant pricing information from the Google search results about the prices concerning DeepSeek-V3.
INFO     [agent] 🧠 Memory: Extracted the following pricing details for DeepSeek-V3: 
- Input cost (Cache Hit): $0.07 per million tokens 
- Input cost (Cache Miss): $0.27 per million tokens 
- Output cost: $1.10 per million tokens 
- General charges: 
  - Input tokens: $0.000140 per 1,000 tokens 
  - Output tokens: $0.000280 per 1,000 tokens 
  - Additional fee: $0.0004
Progress: 6 out of 100 actions completed so far and I need to confirm completion of the task.
INFO     [agent] 🎯 Next goal: Compile the extracted pricing information for comparison.
INFO     [agent] ☑️ Executed action 1/1: done
INFO     [agent] 📄 Result: Pricing comparison completed: 

**gpt-4o-mini:** 
- Input Cost: $0.15 per million tokens 
- Output Cost: $0.60 per million tokens 
- Comparison: 60% cheaper than GPT-3.5 Turbo

**DeepSeek-V3:** 
- Cache Hit Input Cost: $0.07 per million tokens 
- Cache Miss Input Cost: $0.27 per million tokens 
- Output Cost: $1.10 per million tokens 
- General Charges: 
  - Input Tokens: $0.000140 per 1,000 tokens 
  - Output Tokens: $0.000280 per 1,000 tokens 
  - Additional Fee: $0.0004 

This encapsulates the pricing details for both models, fulfilling the task requirements.
INFO     [agent] 📍 Step 7: Ran 1 actions in 10.25s: ✅ 1
INFO     [agent] ✅ Task completed successfully
INFO     [browser] 🛑 Stopped the chromium browser keep_alive=False user_data_dir=~/.config/browseruse/profiles/default cdp_url=None pid=42566

コスト

$0
model: gpt-4o-mini
- Input: 221.098K
- Output: 1.62K

日本語にしたらどうなるか

- task="Compare the price of gpt-4o and DeepSeek-V3",
+ task="gpt-4o-miniとDeepSeek-V3の価格を比較して",

結果

出力される言語に差分は見受けられなかった

% python3 main.py

INFO     [telemetry] Anonymized telemetry enabled. See https://docs.browser-use.com/development/telemetry for more information.
INFO     [agent] 🧠 Starting a browser-use agent v0.2.4 with base_model=gpt-4o-mini +tools +vision +memory extraction_model=gpt-4o-mini 
path/venv/lib/python3.13/site-packages/faiss/loader.py:49: DeprecationWarning: numpy.core._multiarray_umath is deprecated and has been renamed to numpy._core._multiarray_umath. The numpy._core namespace contains private NumPy internals and its use is discouraged, as NumPy internals can change without warning in any release. In practice, most real-world usage of numpy.core is to access functionality in the public NumPy API. If that is the case, use the public NumPy API. If not, you are using NumPy internals. If you would still like to access an internal attribute, use numpy._core._multiarray_umath.__cpu_features__.
  from numpy.core._multiarray_umath import __cpu_features__
<frozen importlib._bootstrap>:488: DeprecationWarning: builtin type SwigPyPacked has no __module__ attribute
<frozen importlib._bootstrap>:488: DeprecationWarning: builtin type SwigPyObject has no __module__ attribute
<frozen importlib._bootstrap>:488: DeprecationWarning: builtin type swigvarlink has no __module__ attribute
INFO     [agent] 🚀 Starting task: gpt-4o-miniとDeepSeek-V3の価格を比較して
INFO     [browser] 🌎 Launching local browser driver=playwright channel=chromium user_data_dir=~/.config/browseruse/profiles/default
INFO     [agent] 📍 Step 1: Evaluating page with 0 interactive elements on: about:blank
===================================================================================================================================================================================
INFO     [agent] 🧠 LLM call => ChatOpenAI [✉️ 7 msg, ~3147 tk, 64607 char, 📷 img] => JSON out + 🔨 20 tools (function_calling)
INFO     [agent] ❓ Eval: Unknown - No actions have been taken yet since the page is currently blank. 
                                                        The ultimate task has not yet begun, and there are no interactive elements available to proceed.
INFO     [agent] 🧠 Memory: Currently at step 1/100, the agent's browser is located at an empty page with no interactive elements. 
                                                        The task of comparing the prices of 'gpt-4o-mini' and 'DeepSeek-V3' has not yet been started.
INFO     [agent] 🎯 Next goal: The first step will be to perform a Google search for 'gpt-4o-mini price' to begin gathering information.
INFO     [controller] 🔍  Searched for "gpt-4o-mini price" in Google
INFO     [agent] ☑️ Executed action 1/1: search_google
INFO     [agent] 📍 Step 2: Ran 1 actions in 10.36s: ✅ 1
INFO     [agent] 📍 Step 2: Evaluating page with 82 interactive elements on: https://www.google.com/search?q=gpt-4o-mini%20pric...
===================================================================================================================
INFO     [agent] 🧠 LLM call => ChatOpenAI [✉️ 10 msg, ~4471 tk, 576906 char, 📷 img] => JSON out + 🔨 20 tools (function_calling)
INFO     [agent] 👍 Eval: Success - The Google search for 'gpt-4o-mini price' returned relevant results. 
                                                        I now have found potential price data for the gpt-4o-mini model.
INFO     [agent] 🧠 Memory: Currently at step 2/100, I have initiated a Google search for 'gpt-4o-mini price'. 
                                                        Relevant search results are displayed, providing different pricing information for 'gpt-4o-mini' from various sources. 
                                                        Next, I will extract the necessary price data from the top result regarding 'gpt-4o-mini' to include in the comparison.
INFO     [agent] 🎯 Next goal: Extract the price for 'gpt-4o-mini' from the first relevant search result.
INFO     [controller] 📄  Extracted from page
: json
{
  "gpt-4o-mini_price_information": {
    "input_cost_per_million_tokens": "$0.15",
    "output_cost_per_million_tokens": "$0.60",
    "comparison": "60% cheaper than GPT-3.5 Turbo"
  }
}


INFO     [agent] ☑️ Executed action 1/1: extract_content
INFO     [agent] 📍 Step 3: Ran 1 actions in 12.60s: ✅ 1
INFO     [agent] 📍 Step 3: Evaluating page with 82 interactive elements on: https://www.google.com/search?q=gpt-4o-mini%20pric...
===================================================================================================================
INFO     [agent] 🧠 LLM call => ChatOpenAI [✉️ 13 msg, ~4844 tk, 577147 char, 📷 img] => JSON out + 🔨 20 tools (function_calling)
INFO     [agent] ❓ Eval: Unknown - I have extracted price data for 'gpt-4o-mini', but have not yet gathered information for 'DeepSeek-V3'. 
                                                        The current state indicates that the task has not been fully completed yet.
INFO     [agent] 🧠 Memory: Currently at step 3/100. Extracted pricing information for 'gpt-4o-mini': input cost per million tokens is $0.15, output cost per million tokens is $0.60, and it's stated to be 60% cheaper than GPT-3.5 Turbo. Now, I need to do a Google search for 'DeepSeek-V3 price' to continue my comparison task.
INFO     [agent] 🎯 Next goal: Perform a Google search for 'DeepSeek-V3 price'.
INFO     [controller] 🔍  Searched for "DeepSeek-V3 price" in Google
INFO     [agent] ☑️ Executed action 1/1: search_google
INFO     [agent] 📍 Step 4: Ran 1 actions in 7.10s: ✅ 1
INFO     [agent] 📍 Step 4: Evaluating page with 76 interactive elements on: https://www.google.com/search?q=DeepSeek-V3%20pric...
===================================================================================================================
INFO     [agent] 🧠 LLM call => ChatOpenAI [✉️ 16 msg, ~5201 tk, 582810 char, 📷 img] => JSON out + 🔨 20 tools (function_calling)
INFO     [agent] 👍 Eval: Success - The Google search for 'DeepSeek-V3 price' returned relevant results. 
                                                        I now have found potential price data for the DeepSeek-V3 model.
INFO     [agent] 🧠 Memory: Currently at step 4/100. I have performed a Google search for 'DeepSeek-V3 price'. I found several relevant results indicating various costs associated with DeepSeek-V3. Next, I will extract the necessary price data from the first relevant search result regarding DeepSeek-V3 to include in the comparison.
INFO     [agent] 🎯 Next goal: Extract the price for 'DeepSeek-V3' from the first relevant search result.
INFO     [controller] 📄  Extracted from page
: json
{
  "price_information": {
    "input_tokens_cache_hit": "$0.07 per 1M tokens",
    "input_tokens_cache_miss": "$0.27 per 1M tokens",
    "output_tokens": "$1.10 per 1M tokens"
  },
  "source": "first_search_result"
}


INFO     [agent] ☑️ Executed action 1/1: extract_content
INFO     [agent] 📍 Step 5: Ran 1 actions in 6.61s: ✅ 1
INFO     [agent] 📍 Step 5: Evaluating page with 76 interactive elements on: https://www.google.com/search?q=DeepSeek-V3%20pric...
===================================================================================================================
INFO     [agent] 🧠 LLM call => ChatOpenAI [✉️ 19 msg, ~5568 tk, 583080 char, 📷 img] => JSON out + 🔨 20 tools (function_calling)
INFO     [agent] 👍 Eval: Success - I have extracted the pricing information for 'DeepSeek-V3'. 
                                                        Now I have both price details needed for comparison with 'gpt-4o-mini'.
INFO     [agent] 🧠 Memory: Currently at step 5/100. Extracted pricing information for 'DeepSeek-V3': input tokens cache hit is $0.07 per million tokens, input tokens cache miss is $0.27 per million tokens, and output tokens are $1.10 per million tokens. Now, I have all the necessary price data for completing my original task of comparing the prices between 'gpt-4o-mini' and 'DeepSeek-V3'.
INFO     [agent] 🎯 Next goal: Prepare the final comparison of prices between 'gpt-4o-mini' and 'DeepSeek-V3'.
INFO     [agent] ☑️ Executed action 1/1: done
INFO     [agent] 📄 Result: Comparison of prices between 'gpt-4o-mini' and 'DeepSeek-V3' is completed:

**gpt-4o-mini Pricing:**
- Input Cost per Million Tokens: $0.15
- Output Cost per Million Tokens: $0.60
- Comparison: 60% cheaper than GPT-3.5 Turbo

**DeepSeek-V3 Pricing:**
- Input Tokens Cache Hit: $0.07 per Million Tokens
- Input Tokens Cache Miss: $0.27 per Million Tokens
- Output Tokens: $1.10 per Million Tokens

In conclusion, DeepSeek-V3 offers a lower price for input tokens cache hits compared to gpt-4o-mini, while the output token cost for gpt-4o-mini is lower than that of DeepSeek-V3. Overall, DeepSeek-V3 is generally cheaper for input tokens.
INFO     [agent] 📍 Step 6: Ran 1 actions in 8.64s: ✅ 1
INFO     [agent] ✅ Task completed successfully
INFO     [browser] 🛑 Stopped the chromium browser keep_alive=False user_data_dir=~/.config/browseruse/profiles/default cdp_url=None pid=53879

累計コスト

$0.01
model: gpt-4o-mini
- Input: 386.406K
- Output: 2.833K

実行の解説

ログの各セクションの解説

初期設定とエージェントの起動
- INFO [telemetry] Anonymized telemetry enabled.
- INFO [agent] 🧠 Starting a browser-use agent v0.2.4 with base_model=gpt-4o-mini +tools +vision +memory extraction_model=gpt-4o-mini
  - 匿名化されたテレメトリーが有効になり、browser-use エージェントのバージョン v0.2.4 が起動したことを示しています。
  - 使用されている基盤モデルは gpt-4o-mini であり、ツール、ビジョン、メモリ抽出機能が有効になっています。
- DeprecationWarning メッセージがいくつか表示されていますが、これは主にPythonライブラリ（numpy, faiss, importlib）の内部的な非推奨機能に関するもので、エージェントの実行には直接的な影響はありません。
タスク開始とブラウザの起動
- INFO [agent] 🚀 Starting task: Compare the price of gpt-4o-mini and DeepSeek-V3
  - エージェントが「gpt-4o-mini と DeepSeek-V3 の価格を比較する」というタスクを開始したことを示しています。
- INFO [browser] 🌎 Launching local browser driver=playwright channel=chromium user_data_dir=~/.config/browseruse/profiles/default
  - Playwright を使用してローカルの Chromium ブラウザを起動したことを示しています。
ステップ1: 初期状態の評価と検索の実行
- INFO [agent] 📍 Step 1: Evaluating page with 0 interactive elements on: about:blank
  - ブラウザが初期の空のページ（about:blank）にあることを評価しています。インタラクティブな要素はゼロです。
- INFO [agent] ❓ Eval: Unknown - The current state does not provide any elements to navigate or interact with, since it's an empty page. I will need to perform a search to gather information about the pricing.
  - 現在のページには情報がないため、価格情報を収集するために検索を行う必要があると判断しています。
- INFO [agent] 🎯 Next goal: Perform a Google search for "gpt-4o-mini price" to gather pricing information.
  - 次の目標として、「gpt-4o-mini price」でGoogle検索を実行することを設定しています。
- INFO [controller] 🔍 Searched for "gpt-4o-mini price" in Google
- INFO [agent] ☑️ Executed action 1/1: search_google
  - Googleで「gpt-4o-mini price」を検索しました。
ステップ2: gpt-4o-miniの価格情報抽出
- INFO [agent] 📍 Step 2: Evaluating page with 82 interactive elements on: https://www.google.com/search?q=gpt-4o-mini%20pric...
  - 検索結果ページを評価しています。82個のインタラクティブな要素があります。
- INFO [agent] ❓ Eval: Unknown - The search for 'gpt-4o-mini price' is successful, but I still need to extract the relevant pricing information from the results.
  - 検索は成功しましたが、関連する価格情報を抽出する必要があると判断しています。
- INFO [agent] 🎯 Next goal: Extract relevant content from the search results for 'gpt-4o-mini price'.
  - 次の目標として、検索結果から関連コンテンツを抽出することを設定しています。
- INFO [controller] 📄 Extracted from page: json {...}
  - ページから以下の情報を抽出しました:
    - gpt-4o-mini:
      - 入力コスト: $0.15 / 100万トークン
      - 出力コスト: $0.60 / 100万トークン
      - 比較: GPT-3.5 Turboより60%安い
ステップ3: DeepSeek-V3の価格情報検索の準備
- INFO [agent] 👍 Eval: Success - Extracted pricing information for gpt-4o-mini successfully.
  - gpt-4o-miniの価格情報の抽出に成功したと判断しています。
- INFO [agent] 🧠 Memory: Extracted the following pricing data: ... I need to now search for the pricing information for DeepSeek-V3.
  - gpt-4o-miniの価格情報を記憶し、次にDeepSeek-V3の価格情報を検索する必要があることを認識しています。
- INFO [agent] 🎯 Next goal: Perform a Google search for "DeepSeek-V3 price" to gather pricing information.
  - 次の目標として、「DeepSeek-V3 price」でGoogle検索を実行することを設定しています。
- INFO [controller] 🔍 Searched for "DeepSeek-V3 price" in Google
  - Googleで「DeepSeek-V3 price」を検索しました。
ステップ4〜6: DeepSeek-V3の価格情報抽出
- エージェントは同様のプロセスを繰り返し、DeepSeek-V3の価格情報を検索結果から抽出します。
- 複数回抽出アクションを実行しているのは、初回の抽出で完全な情報が得られなかったか、より詳細な情報を得るために追加の抽出を試みたためと考えられます。
- INFO [controller] 📄 Extracted from page: json {...}
  - DeepSeek-V3の価格情報が抽出されました:
    - DeepSeek-V3:
      - キャッシュヒット入力コスト: $0.07 / 100万トークン
      - キャッシュミス入力コスト: $0.27 / 100万トークン
      - 出力コスト: $1.10 / 100万トークン
      - 一般的な料金（General Charges）:
        
        入力トークン: $0.000140 / 1,000トークン ($0.14 / 100万トークン)
        
        出力トークン: $0.000280 / 1,000トークン ($0.28 / 100万トークン)
        
        追加料金: $0.0004 / リクエスト
ステップ7: 最終結果のまとめとタスク完了
- INFO [agent] 🎯 Next goal: Compile the extracted pricing information for comparison.
  - 抽出された価格情報を比較のためにまとめることを目標としています。
- INFO [agent] ☑️ Executed action 1/1: done
  - タスクが完了したと判断し、完了アクションを実行しました。
- INFO [agent] 📄 Result: Pricing comparison completed: ...
  - 最終的な比較結果が出力されました。これは、上記で抽出された両モデルの価格情報がまとめられたものです。
- INFO [agent] ✅ Task completed successfully
  - タスクが正常に完了したことを示しています。
- INFO [browser] 🛑 Stopped the chromium browser keep_alive=False ...
  - Chromiumブラウザを停止しました。

感想

もちろんこれだけだと意味がわからないので実際は実アプリケーションの中に組み込んでゴニョゴニョする感じだと思う

仮想環境作成と有効化

インストール

main.py作成

.env作成

実行

結果

コスト

日本語にしたらどうなるか

結果

累計コスト

実行の解説

ログの各セクションの解説

感想

Discussion

`main.py`作成

`.env`作成