Feature/langchain agent

README.md

@@ -8,4 +8,7 @@ firebase emulators:start --only functions
 
 ```bash
 firebase deploy --only functions
-```
+```
+
+## Reference
+- https://cloud.google.com/vertex-ai/generative-ai/docs/agent-engine/develop/langchain#vertex-ai-extension

functions/agent.py

@@ -0,0 +1,44 @@
+from vertexai import agent_engines, init
+from langchain_google_vertexai import HarmBlockThreshold, HarmCategory
+
+init(project="ai-agent-hackathon-3-471422", location="us-central1")
+
+safety_settings = {
+    HarmCategory.HARM_CATEGORY_UNSPECIFIED: HarmBlockThreshold.BLOCK_NONE,
+    HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT: HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE,
+    HarmCategory.HARM_CATEGORY_HATE_SPEECH: HarmBlockThreshold.BLOCK_ONLY_HIGH,
+    HarmCategory.HARM_CATEGORY_HARASSMENT: HarmBlockThreshold.BLOCK_LOW_AND_ABOVE,
+    HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT: HarmBlockThreshold.BLOCK_NONE,
+}
+
+model_kwargs = {
+    # temperature (float): The sampling temperature controls the degree of
+    # randomness in token selection.
+    "temperature": 0.28,
+    # max_output_tokens (int): The token limit determines the maximum amount of
+    # text output from one prompt.
+    "max_output_tokens": 1000,
+    # top_p (float): Tokens are selected from most probable to least until
+    # the sum of their probabilities equals the top-p value.
+    "top_p": 0.95,
+    # top_k (int): The next token is selected from among the top-k most
+    # probable tokens. This is not supported by all model versions. See
+    # https://cloud.google.com/vertex-ai/generative-ai/docs/multimodal/image-understanding#valid_parameter_values
+    # for details.
+    "top_k": None,
+    # safety_settings (Dict[HarmCategory, HarmBlockThreshold]): The safety
+    # settings to use for generating content.
+    # (you must create your safety settings using the previous step first).
+    "safety_settings": safety_settings,
+}
+
+model = "gemini-2.0-flash"
+
+agent = agent_engines.LangchainAgent(
+    model=model,                # Required.
+    model_kwargs=model_kwargs,  # Optional.
+)
+
+if __name__ == "__main__":
+    response = agent.query(input="What is the exchange rate from US dollars to SEK today?")
+    print(response)

functions/gee_ndvi.py

@@ -0,0 +1,54 @@
+import ee
+import requests
+
+def initialize_gee():
+    """認証と初期化。ローカル認証が必要な場合はee.Authenticate()を使う。"""
+    try:
+        ee.Initialize()
+    except Exception:
+        ee.Authenticate()
+        ee.Initialize()
+
+def get_ndvi_image(lat, lon, start_date, end_date, out_path):
+    """
+    指定座標・期間でNDVI画像を生成し、ファイル保存する。
+    Args:
+        lat (float): 緯度
+        lon (float): 経度
+        start_date (str): 取得開始日（YYYY-MM-DD）
+        end_date (str): 取得終了日（YYYY-MM-DD）
+        out_path (str): 保存先ファイルパス
+    Returns:
+        out_path (str): 保存した画像ファイルパス
+    """
+    # Sentinel-2画像コレクション取得
+    point = ee.Geometry.Point([lon, lat])
+    collection = ee.ImageCollection('COPERNICUS/S2') \
+        .filterBounds(point) \
+        .filterDate(start_date, end_date) \
+        .sort('CLOUD_COVER')
+    image = collection.first()
+    # NDVI計算
+    ndvi = image.normalizedDifference(['B8', 'B4']).rename('NDVI')
+    # 可視化パラメータ
+    vis_params = {'min': 0, 'max': 1, 'palette': ['blue', 'white', 'green']}
+    # 画像をエクスポート
+    url = ndvi.getThumbURL({'region': point.buffer(10000).bounds().getInfo(), 'dimensions': 512, 'format': 'png', **vis_params})
+    # 画像ダウンロード
+    r = requests.get(url)
+    with open(out_path, 'wb') as f:
+        f.write(r.content)
+    return out_path
+
+if __name__ == "__main__":
+    # サンプル座標（東京駅付近）
+    lat = 35.681236
+    lon = 139.767125
+    start_date = "2023-08-01"
+    end_date = "2023-08-31"
+    out_path = "ndvi_sample.png"
+    print("GEE認証・初期化中...")
+    initialize_gee()
+    print("NDVI画像生成中...")
+    result = get_ndvi_image(lat, lon, start_date, end_date, out_path)
+    print(f"画像保存先: {result}")

functions/requirements.txt

@@ -1 +1,2 @@
-firebase_functions~=0.1.0
+firebase_functions~=0.1.0
+earthengine-api