diff --git a/etl-pipeline/custom/construct_send_telegram_message.py b/etl-pipeline/custom/construct_send_telegram_message.py
new file mode 100644
index 0000000..90a8b87
--- /dev/null
+++ b/etl-pipeline/custom/construct_send_telegram_message.py
@@ -0,0 +1,84 @@
+import telebot
+from mage_ai.data_preparation.shared.secrets import get_secret_value
+if 'custom' not in globals():
+    from mage_ai.data_preparation.decorators import custom
+if 'test' not in globals():
+    from mage_ai.data_preparation.decorators import test
+
+AQI_LEVELS = [
+    (0, 50, "Bueno 😁"),  # Good
+    (51, 100, "Moderado 🙂"),  # Moderate
+    (101, 150, "Insalubre para grupos sensibles 😷"),  # Unhealthy for sensitive groups
+    (151, 200, "Insalubre 😶‍🌫️"),  # Unhealthy
+    (201, 300, "Muy insalubre 😰"),  # Very Unhealthy
+    (301, 500, "Peligroso 💀")  # Hazardous
+]
+
+AQI_MESSAGES = [
+    (0, 50,"🟢 ¡Hoy es un excelente día para estar al aire libre! 🏃‍♂️🍃"),  # Good
+    (51, 100, "🟡 Hoy es un buen día para estar al aire libre para la mayoría de las personas. 👍🏻"),  # Moderate
+    (101, 150, "🟠 Las personas sensibles deben prestar atención a síntomas 😷"),  # Unhealthy for sensitive groups
+    (151, 200, "🔴 Hoy es un mal día para estar al aire libre.\n Se recomienda el uso de tapabocas para la población general💨😷"),  # Unhealthy
+    (201, 300, "🟣 Hoy es un muy mal día para estar al aire libre. \n Se recomienda el uso de tapabocas para la población general🔥😷"),  # Very Unhealthy
+    (301, 500, "⚫ Evite la exposición al aire libre. 💀😷\nAnte la aparición de síntomas, consulte a su médico.")  # Hazardous
+
+]
+
+def get_aqi_label(aqi_value):
+    """Returns the AQI level label for a given AQI value."""
+    for min_val, max_val, label in AQI_LEVELS:
+        if min_val <= aqi_value <= max_val:
+            return label
+    return "Fuera del AQI 💀🤯"  
+
+def get_aqi_message(aqi_value):
+    for min_val, max_val, message in AQI_MESSAGES:
+        if min_val <= aqi_value <= max_val:
+            return message
+    return " "
+
+def get_latest_aqi_summary(df):
+    avg_value = int(df['forecast_avg'].mean())
+    max_value = int(df['forecast_avg'].max())
+    min_value = int(df['forecast_avg'].min())
+    
+    return avg_value, max_value, min_value
+
+def construct_message(avg_aqi, max_aqi, min_aqi):
+    avg_label = get_aqi_label(avg_aqi)
+    max_label = get_aqi_label(max_aqi)
+    min_label = get_aqi_label(min_aqi)
+    aqi_message = get_aqi_message(avg_aqi)
+
+    message = (
+        f"📊 **Reporte de Calidad del Aire para Gran Asunción - Pronóstico de 12 horas**\n"
+        f"🔹 **AQI Promedio:** {avg_aqi} ({avg_label})\n"
+        f"🔺 **AQI Máximo:** {max_aqi} ({max_label})\n"
+        f"🔻 **AQI Mínimo:** {min_aqi} ({min_label})\n\n"
+        f"{aqi_message}\n\n"
+        "🔗 Podés acceder al pronóstico actualizado en tu zona ingresando a www.proyectorespira.net"
+    )
+    return message
+
+def send_message(token, chat_id, msg=''):
+    bot = telebot.TeleBot(token)
+    return bot.send_message(chat_id=chat_id, text=msg, parse_mode='Markdown')
+
+@custom
+def transform_custom(data, *args, **kwargs):
+    telegram_token = get_secret_value('TELEGRAM_BOT_TOKEN')
+    telegram_chat_id = get_secret_value('TELEGRAM_CHAT_ID')
+    aqi_summary = get_latest_aqi_summary(data)
+    if aqi_summary:
+        avg_aqi, max_aqi, min_aqi = aqi_summary
+    message = construct_message(avg_aqi, max_aqi, min_aqi)
+    send_message(token=telegram_token, chat_id=telegram_chat_id, msg=message)
+    return data
+
+
+@test
+def test_output(output, *args) -> None:
+    """
+    Template code for testing the output of the block.
+    """
+    assert output is not None, 'The output is undefined'
diff --git a/etl-pipeline/custom/construct_send_twitter_message.py b/etl-pipeline/custom/construct_send_twitter_message.py
new file mode 100644
index 0000000..7d96092
--- /dev/null
+++ b/etl-pipeline/custom/construct_send_twitter_message.py
@@ -0,0 +1,68 @@
+import tweepy
+from mage_ai.data_preparation.shared.secrets import get_secret_value
+
+if 'custom' not in globals():
+    from mage_ai.data_preparation.decorators import custom
+if 'test' not in globals():
+    from mage_ai.data_preparation.decorators import test
+
+AQI_LEVELS = [
+    (0, 50, "🟢 Bueno 😁"),  # Good
+    (51, 100, "🟡 Moderado 🙂"),  # Moderate
+    (101, 150, "🟠 Insalubre para grupos sensibles 😷"),  # Unhealthy for sensitive groups
+    (151, 200, "🔴 Insalubre 😶‍🌫️"),  # Unhealthy
+    (201, 300, "🟣 Muy insalubre 😰"),  # Very Unhealthy
+    (301, 500, "⚫ Peligroso 💀")  # Hazardous
+]
+
+def get_aqi_label(aqi_value):
+    for min_val, max_val, label in AQI_LEVELS:
+        if min_val <= aqi_value <= max_val:
+            return label
+    return "Fuera del AQI 💀🤯"  
+
+def get_latest_aqi_summary(df):
+    avg_value = int(df['forecast_avg'].mean())
+    max_value = int(df['forecast_avg'].max())
+    min_value = int(df['forecast_avg'].min())
+    
+    return avg_value, max_value, min_value
+
+def construct_message(avg_aqi, max_aqi, min_aqi):
+    avg_label = get_aqi_label(avg_aqi)
+    max_label = get_aqi_label(max_aqi)
+    min_label = get_aqi_label(min_aqi)
+
+    message = (
+        f"📊 Calidad del Aire para Gran Asunción - Próximas 12 hs\n"
+        f"🔹 AQI Promedio: {avg_aqi} ({avg_label})\n"
+        f"🔺 AQI Máximo: {max_aqi} ({max_label})\n"
+        f"🔻 AQI Mínimo: {min_aqi} ({min_label})\n\n"
+        "🔗 ¡Pronto podrás acceder al pronóstico en tu zona! www.proyectorespira.net"
+    )
+    return message
+
+def send_message(msg=''):
+    consumer_key = get_secret_value('TWITTER_CONSUMER_KEY')
+    consumer_secret = get_secret_value('TWITTER_CONSUMER_SECRET')
+    access_token = get_secret_value('TWITTER_ACCESS_TOKEN')
+    access_token_secret = get_secret_value('TWITTER_ACCESS_TOKEN_SECRET')
+    
+    client = tweepy.Client(consumer_key=consumer_key,
+                            consumer_secret=consumer_secret,
+                            access_token=access_token,
+                            access_token_secret=access_token_secret)
+    response = client.create_tweet(text=msg)
+    
+    return 
+
+@custom
+def transform_custom(data, *args, **kwargs):
+
+
+    aqi_summary = get_latest_aqi_summary(data)
+    if aqi_summary:
+        avg_aqi, max_aqi, min_aqi = aqi_summary
+    message = construct_message(avg_aqi, max_aqi, min_aqi)
+    send_message(msg=message)
+    return data
diff --git a/etl-pipeline/data_loaders/get_latest_forecast_12h.sql b/etl-pipeline/data_loaders/get_latest_forecast_12h.sql
new file mode 100644
index 0000000..e2a0f4f
--- /dev/null
+++ b/etl-pipeline/data_loaders/get_latest_forecast_12h.sql
@@ -0,0 +1,12 @@
+-- Docs: https://docs.mage.ai/guides/sql-blocks
+WITH inference_run_id AS (
+    SELECT id
+    FROM inference_runs
+    WHERE run_date BETWEEN ('{{ execution_date }}'::timestamp - INTERVAL '2 hours') 
+                      AND ('{{ execution_date }}'::timestamp + INTERVAL '1 hours')
+    ORDER BY run_date DESC
+    LIMIT 1
+)
+SELECT inference_run, station, forecasts_12h
+FROM inference_results
+WHERE inference_run = (SELECT id FROM inference_run_id);
\ No newline at end of file
diff --git a/etl-pipeline/pipelines/telegram_bot/metadata.yaml b/etl-pipeline/pipelines/telegram_bot/metadata.yaml
index 5493ee1..70a923c 100755
--- a/etl-pipeline/pipelines/telegram_bot/metadata.yaml
+++ b/etl-pipeline/pipelines/telegram_bot/metadata.yaml
@@ -1,4 +1,66 @@
-blocks: []
+blocks:
+- all_upstream_blocks_executed: true
+  color: null
+  configuration:
+    data_provider: postgres
+    data_provider_profile: default
+    dbt: {}
+    disable_query_preprocessing: false
+    export_write_policy: append
+    file_path: data_loaders/get_latest_forecast_12h.sql
+    file_source:
+      path: data_loaders/get_latest_forecast_12h.sql
+    limit: 1000
+    use_raw_sql: true
+  downstream_blocks:
+  - calculate_regional_forecast
+  executor_config: null
+  executor_type: local_python
+  has_callback: false
+  language: sql
+  name: get_latest_forecast_12h
+  retry_config: null
+  status: executed
+  timeout: null
+  type: data_loader
+  upstream_blocks: []
+  uuid: get_latest_forecast_12h
+- all_upstream_blocks_executed: true
+  color: null
+  configuration:
+    data_provider: postgres
+    data_provider_profile: default
+    export_write_policy: append
+  downstream_blocks:
+  - construct_send_telegram_message
+  executor_config: null
+  executor_type: local_python
+  has_callback: false
+  language: python
+  name: calculate_regional_forecast
+  retry_config: null
+  status: updated
+  timeout: null
+  type: transformer
+  upstream_blocks:
+  - get_latest_forecast_12h
+  uuid: calculate_regional_forecast
+- all_upstream_blocks_executed: false
+  color: null
+  configuration: {}
+  downstream_blocks: []
+  executor_config: null
+  executor_type: local_python
+  has_callback: false
+  language: python
+  name: construct_send_telegram_message
+  retry_config: null
+  status: executed
+  timeout: null
+  type: custom
+  upstream_blocks:
+  - calculate_regional_forecast
+  uuid: construct_send_telegram_message
 cache_block_output_in_memory: false
 callbacks: []
 concurrency_config: {}
@@ -22,5 +84,5 @@ tags:
 - bots
 type: python
 uuid: telegram_bot
-variables_dir: /app/.mage_data/
+variables_dir: /app/.mage_data
 widgets: []
diff --git a/etl-pipeline/pipelines/twitter_bot/metadata.yaml b/etl-pipeline/pipelines/twitter_bot/metadata.yaml
index 96c4134..5b5b945 100755
--- a/etl-pipeline/pipelines/twitter_bot/metadata.yaml
+++ b/etl-pipeline/pipelines/twitter_bot/metadata.yaml
@@ -1,4 +1,66 @@
-blocks: []
+blocks:
+- all_upstream_blocks_executed: true
+  color: null
+  configuration:
+    data_provider: postgres
+    data_provider_profile: default
+    dbt: {}
+    disable_query_preprocessing: false
+    export_write_policy: append
+    file_path: data_loaders/get_latest_forecast_12h.sql
+    file_source:
+      path: data_loaders/get_latest_forecast_12h.sql
+    limit: 1000
+    use_raw_sql: true
+  downstream_blocks:
+  - calculate_regional_forecast
+  executor_config: null
+  executor_type: local_python
+  has_callback: false
+  language: sql
+  name: get_latest_forecast_12h
+  retry_config: null
+  status: executed
+  timeout: null
+  type: data_loader
+  upstream_blocks: []
+  uuid: get_latest_forecast_12h
+- all_upstream_blocks_executed: true
+  color: null
+  configuration:
+    file_path: transformers/calculate_regional_forecast.py
+    file_source:
+      path: transformers/calculate_regional_forecast.py
+  downstream_blocks:
+  - construct_send_twitter_message
+  executor_config: null
+  executor_type: local_python
+  has_callback: false
+  language: python
+  name: calculate_regional_forecast
+  retry_config: null
+  status: executed
+  timeout: null
+  type: transformer
+  upstream_blocks:
+  - get_latest_forecast_12h
+  uuid: calculate_regional_forecast
+- all_upstream_blocks_executed: true
+  color: null
+  configuration: {}
+  downstream_blocks: []
+  executor_config: null
+  executor_type: local_python
+  has_callback: false
+  language: python
+  name: construct_send_twitter_message
+  retry_config: null
+  status: updated
+  timeout: null
+  type: custom
+  upstream_blocks:
+  - calculate_regional_forecast
+  uuid: construct_send_twitter_message
 cache_block_output_in_memory: false
 callbacks: []
 concurrency_config: {}
@@ -22,5 +84,5 @@ tags:
 - bots
 type: python
 uuid: twitter_bot
-variables_dir: /app/.mage_data/
+variables_dir: /app/.mage_data
 widgets: []
diff --git a/etl-pipeline/transformers/calculate_regional_forecast.py b/etl-pipeline/transformers/calculate_regional_forecast.py
new file mode 100644
index 0000000..4770222
--- /dev/null
+++ b/etl-pipeline/transformers/calculate_regional_forecast.py
@@ -0,0 +1,34 @@
+import pandas as pd
+if 'transformer' not in globals():
+    from mage_ai.data_preparation.decorators import transformer
+if 'test' not in globals():
+    from mage_ai.data_preparation.decorators import test
+
+
+@transformer
+def transform(data, *args, **kwargs):
+    klogger = kwargs.get('logger')
+    try:
+        if data.empty:
+            klogger.exception('No available forecasts')
+        df_exploded = data.explode('forecasts_12h').reset_index(drop=True)
+        df_exploded[['timestamp', 'value']] = pd.json_normalize(df_exploded['forecasts_12h'])
+        df_transformed = df_exploded.drop(columns=['forecasts_12h'])
+        df_transformed['timestamp'] = pd.to_datetime(df_transformed['timestamp'], utc=True)
+
+        df_grouped = (
+            df_transformed.groupby(['timestamp'], as_index=False)['value']
+            .mean().round()
+            .rename(columns={'value': 'forecast_avg'})
+        )
+        return df_grouped
+    except Exception as e:
+        klogger.exception(f'An error occurred: {e}')
+
+
+@test
+def test_output(output, *args) -> None:
+    """
+    Template code for testing the output of the block.
+    """
+    assert output is not None, 'The output is undefined'
\ No newline at end of file