updated

MCP_RESOURCES.md

@@ -0,0 +1,130 @@
+# MCP Resources Documentation
+
+## Overview
+This MCP server provides agricultural data analysis resources for weed pressure prediction and crop recommendations. The resources are designed to be consumed by LLMs through the Model Context Protocol.
+
+## Available Resources
+
+### Static Resources
+
+#### `agricultural://plots`
+- **Description**: List of all available agricultural plots
+- **Returns**: String with plot names and count
+- **Example**: "Available plots (106): Charbonnerie Entrée, Grand champ 5 (MH), ..."
+
+#### `agricultural://crops`
+- **Description**: List of all crop types in the dataset
+- **Returns**: String with crop names and count
+- **Example**: "Available crops (42): blé tendre hiver, pois de conserve, ..."
+
+#### `agricultural://years`
+- **Description**: Range of years available in the dataset
+- **Returns**: String with year range and total count
+- **Example**: "Available years: 2014-2025 (12 years total)"
+
+#### `agricultural://dataset-info`
+- **Description**: Comprehensive dataset information
+- **Returns**: String with statistics about the agricultural dataset
+- **Includes**: Record counts, plot/crop/intervention counts, herbicide usage stats
+
+#### `agricultural://herbicide-usage`
+- **Description**: Summary of herbicide usage patterns
+- **Returns**: String with IFT statistics and risk distribution
+- **Includes**: Total applications, average IFT, risk levels, most used herbicides
+
+#### `agricultural://predictions/2025-2027`
+- **Description**: Summary of weed pressure predictions for 2025-2027
+- **Returns**: String with prediction statistics and risk distribution
+- **Includes**: Total predictions, average IFT, risk levels, best plots
+
+#### `agricultural://recommendations/sensitive-crops`
+- **Description**: Summary of plot recommendations for sensitive crops
+- **Returns**: String with recommendation statistics and top plots
+- **Includes**: Suitable plots count, average score, top 5 recommendations
+
+### Parameterized Resources
+
+#### `agricultural://plot/{plot_name}`
+- **Description**: Detailed information about a specific agricultural plot
+- **Parameters**: `plot_name` (string) - Name of the plot
+- **Returns**: String with plot statistics and characteristics
+- **Includes**: Interventions count, years active, herbicide usage, surface, main crops
+
+#### `agricultural://crop/{crop_type}`
+- **Description**: Information about a specific crop type
+- **Parameters**: `crop_type` (string) - Name of the crop
+- **Returns**: String with crop cultivation patterns
+- **Includes**: Total interventions, years cultivated, plots count, herbicide usage
+
+#### `agricultural://year/{year}`
+- **Description**: Summary of agricultural activities for a specific year
+- **Parameters**: `year` (int) - Year to analyze
+- **Returns**: String with year statistics
+- **Includes**: Total interventions, active plots, crop types, most active plot/crop
+
+#### `agricultural://plot/{plot_name}/predictions`
+- **Description**: Weed pressure predictions for a specific plot
+- **Parameters**: `plot_name` (string) - Name of the plot
+- **Returns**: String with predictions for 2025-2027
+- **Includes**: IFT predictions by year, risk levels, historical average, recent crops
+
+## Usage Examples
+
+### For LLMs
+```python
+# Get dataset overview
+dataset_info = get_dataset_info()
+
+# Get specific plot information
+plot_info = get_plot_info("Champ ferme W du sol")
+
+# Get crop cultivation patterns
+crop_info = get_crop_info("blé tendre hiver")
+
+# Get year summary
+year_summary = get_year_summary(2023)
+
+# Get herbicide usage patterns
+herbicide_usage = get_herbicide_usage_summary()
+
+# Get predictions summary
+predictions = get_predictions_summary()
+
+# Get recommendations for sensitive crops
+recommendations = get_recommendations_summary()
+
+# Get specific plot predictions
+plot_predictions = get_plot_predictions("Etang 5")
+```
+
+## Activation Instructions
+
+To activate MCP resources, uncomment the `@gr.mcp.resource` decorators in `mcp_server.py`:
+
+```python
+# Change from:
+# @gr.mcp.resource("agricultural://plots")
+def get_available_plots_resource() -> str:
+
+# To:
+@gr.mcp.resource("agricultural://plots")
+def get_available_plots_resource() -> str:
+```
+
+## Data Sources
+
+- **Station**: Station Expérimentale de Kerguéhennec
+- **Period**: 2014-2025
+- **Records**: 5,756 agricultural interventions
+- **Plots**: 106 agricultural plots
+- **Crops**: 42 different crop types
+- **Herbicide Applications**: 985 applications
+
+## Key Metrics
+
+- **IFT (Indice de Fréquence de Traitement)**: Number of herbicide applications per hectare
+- **Risk Levels**: 
+  - Low (IFT < 1.0): Suitable for sensitive crops
+  - Moderate (1.0 ≤ IFT < 2.0): Requires monitoring
+  - High (IFT ≥ 2.0): Not recommended for sensitive crops
+- **Recommendation Score**: 100 - (predicted_ift × 30)

mcp_server.py

@@ -89,6 +89,274 @@ class WeedPressureAnalyzer:
 # Initialize analyzer
 analyzer = WeedPressureAnalyzer()
 
+# MCP Resources (to be activated when @gr.mcp.resource is available)
+# These functions provide structured data access for LLMs and can be exposed as MCP resources
+# To activate: uncomment the @gr.mcp.resource decorators below
+
+# @gr.mcp.resource("agricultural://plots")
+def get_available_plots_resource() -> str:
+    """Get list of all available agricultural plots from the dataset"""
+    try:
+        plots = get_available_plots()
+        return f"Available plots ({len(plots)-1}): " + ", ".join(plots[1:6]) + f" ... and {len(plots)-6} more"
+    except Exception as e:
+        return f"Error loading plots: {str(e)}"
+
+# @gr.mcp.resource("agricultural://crops")
+def get_available_crops_resource() -> str:
+    """Get list of all crop types in the dataset"""
+    try:
+        crops = get_available_crops()
+        return f"Available crops ({len(crops)-1}): " + ", ".join(crops[1:6]) + f" ... and {len(crops)-6} more"
+    except Exception as e:
+        return f"Error loading crops: {str(e)}"
+
+# @gr.mcp.resource("agricultural://years")
+def get_available_years_resource() -> str:
+    """Get range of years available in the dataset"""
+    try:
+        df = analyzer.load_data()
+        years = sorted(df['year'].dropna().unique())
+        return f"Available years: {min(years)}-{max(years)} ({len(years)} years total)"
+    except Exception as e:
+        return f"Error loading years: {str(e)}"
+
+# @gr.mcp.resource("agricultural://dataset-info")
+def get_dataset_info() -> str:
+    """Get comprehensive information about the agricultural dataset"""
+    try:
+        df = analyzer.load_data()
+        herbicide_data = df[df['is_herbicide'] == True]
+        
+        info = f"""
+Agricultural Dataset Information:
+- Total records: {len(df):,}
+- Years covered: {df['year'].min()}-{df['year'].max()}
+- Number of plots: {df['plot_name'].nunique()}
+- Number of crop types: {df['crop_type'].nunique()}
+- Number of intervention types: {df['intervention_type'].nunique()}
+- Herbicide applications: {len(herbicide_data):,}
+- Average IFT (herbicides): {herbicide_data.groupby(['plot_name', 'year']).size().mean():.2f}
+- Data source: Station Expérimentale de Kerguéhennec
+- Last updated: 2025
+        """
+        return info.strip()
+    except Exception as e:
+        return f"Error loading dataset info: {str(e)}"
+
+# @gr.mcp.resource("agricultural://plot/{plot_name}")
+def get_plot_info(plot_name: str) -> str:
+    """Get detailed information about a specific agricultural plot"""
+    try:
+        df = analyzer.load_data()
+        plot_data = df[df['plot_name'] == plot_name]
+        
+        if len(plot_data) == 0:
+            return f"Plot '{plot_name}' not found in dataset"
+        
+        herbicide_data = plot_data[plot_data['is_herbicide'] == True]
+        years = sorted(plot_data['year'].unique())
+        
+        info = f"""
+Plot Information: {plot_name}
+- Total interventions: {len(plot_data):,}
+- Years active: {min(years)}-{max(years)} ({len(years)} years)
+- Herbicide applications: {len(herbicide_data):,}
+- Average IFT: {herbicide_data.groupby('year').size().mean():.2f}
+- Surface: {plot_data['plot_surface'].iloc[0]:.2f} hectares
+- Main crops: {', '.join(plot_data['crop_type'].value_counts().head(3).index.tolist())}
+- Main interventions: {', '.join(plot_data['intervention_type'].value_counts().head(3).index.tolist())}
+        """
+        return info.strip()
+    except Exception as e:
+        return f"Error loading plot info: {str(e)}"
+
+# @gr.mcp.resource("agricultural://crop/{crop_type}")
+def get_crop_info(crop_type: str) -> str:
+    """Get information about a specific crop type and its cultivation patterns"""
+    try:
+        df = analyzer.load_data()
+        crop_data = df[df['crop_type'] == crop_type]
+        
+        if len(crop_data) == 0:
+            return f"Crop type '{crop_type}' not found in dataset"
+        
+        herbicide_data = crop_data[crop_data['is_herbicide'] == True]
+        years = sorted(crop_data['year'].unique())
+        plots = crop_data['plot_name'].nunique()
+        
+        info = f"""
+Crop Information: {crop_type}
+- Total interventions: {len(crop_data):,}
+- Years cultivated: {min(years)}-{max(years)} ({len(years)} years)
+- Number of plots: {plots}
+- Herbicide applications: {len(herbicide_data):,}
+- Average IFT: {herbicide_data.groupby(['plot_name', 'year']).size().mean():.2f}
+- Main plots: {', '.join(crop_data['plot_name'].value_counts().head(3).index.tolist())}
+- Main interventions: {', '.join(crop_data['intervention_type'].value_counts().head(3).index.tolist())}
+        """
+        return info.strip()
+    except Exception as e:
+        return f"Error loading crop info: {str(e)}"
+
+# @gr.mcp.resource("agricultural://year/{year}")
+def get_year_summary(year: int) -> str:
+    """Get summary of agricultural activities for a specific year"""
+    try:
+        df = analyzer.load_data()
+        year_data = df[df['year'] == year]
+        
+        if len(year_data) == 0:
+            return f"No data available for year {year}"
+        
+        herbicide_data = year_data[year_data['is_herbicide'] == True]
+        plots = year_data['plot_name'].nunique()
+        crops = year_data['crop_type'].nunique()
+        
+        info = f"""
+Year Summary: {year}
+- Total interventions: {len(year_data):,}
+- Active plots: {plots}
+- Crop types: {crops}
+- Herbicide applications: {len(herbicide_data):,}
+- Average IFT: {herbicide_data.groupby('plot_name').size().mean():.2f}
+- Most active plot: {year_data['plot_name'].value_counts().index[0]} ({year_data['plot_name'].value_counts().iloc[0]} interventions)
+- Most common crop: {year_data['crop_type'].value_counts().index[0]} ({year_data['crop_type'].value_counts().iloc[0]} interventions)
+- Most common intervention: {year_data['intervention_type'].value_counts().index[0]} ({year_data['intervention_type'].value_counts().iloc[0]} interventions)
+        """
+        return info.strip()
+    except Exception as e:
+        return f"Error loading year summary: {str(e)}"
+
+# @gr.mcp.resource("agricultural://herbicide-usage")
+def get_herbicide_usage_summary() -> str:
+    """Get comprehensive summary of herbicide usage patterns"""
+    try:
+        df = analyzer.load_data()
+        herbicide_data = df[df['is_herbicide'] == True]
+        
+        if len(herbicide_data) == 0:
+            return "No herbicide data available"
+        
+        # Calculate IFT by plot and year
+        ift_data = herbicide_data.groupby(['plot_name', 'year']).size().reset_index(name='applications')
+        ift_data['ift'] = ift_data['applications'] / herbicide_data.groupby(['plot_name', 'year'])['plot_surface'].first().values
+        
+        avg_ift = ift_data['ift'].mean()
+        max_ift = ift_data['ift'].max()
+        min_ift = ift_data['ift'].min()
+        
+        # Risk distribution
+        low_risk = len(ift_data[ift_data['ift'] < 1.0])
+        moderate_risk = len(ift_data[(ift_data['ift'] >= 1.0) & (ift_data['ift'] < 2.0)])
+        high_risk = len(ift_data[ift_data['ift'] >= 2.0])
+        
+        info = f"""
+Herbicide Usage Summary:
+- Total applications: {len(herbicide_data):,}
+- Plots with herbicides: {herbicide_data['plot_name'].nunique()}
+- Years with data: {herbicide_data['year'].nunique()}
+- Average IFT: {avg_ift:.2f}
+- IFT range: {min_ift:.2f} - {max_ift:.2f}
+- Risk distribution:
+  * Low risk (IFT < 1.0): {low_risk} plot-years ({low_risk/len(ift_data)*100:.1f}%)
+  * Moderate risk (1.0 ≤ IFT < 2.0): {moderate_risk} plot-years ({moderate_risk/len(ift_data)*100:.1f}%)
+  * High risk (IFT ≥ 2.0): {high_risk} plot-years ({high_risk/len(ift_data)*100:.1f}%)
+- Most used herbicides: {', '.join(herbicide_data['produit'].value_counts().head(3).index.tolist())}
+        """
+        return info.strip()
+    except Exception as e:
+        return f"Error loading herbicide usage summary: {str(e)}"
+
+# @gr.mcp.resource("agricultural://predictions/2025-2027")
+def get_predictions_summary() -> str:
+    """Get summary of weed pressure predictions for 2025-2027"""
+    try:
+        predictions = analyzer.predict_weed_pressure()
+        
+        if len(predictions) == 0:
+            return "No predictions available - insufficient historical data"
+        
+        low_risk = len(predictions[predictions['risk_level'] == 'Faible'])
+        moderate_risk = len(predictions[predictions['risk_level'] == 'Modéré'])
+        high_risk = len(predictions[predictions['risk_level'] == 'Élevé'])
+        
+        avg_ift = predictions['predicted_ift'].mean()
+        
+        info = f"""
+Weed Pressure Predictions 2025-2027:
+- Total predictions: {len(predictions)}
+- Average predicted IFT: {avg_ift:.2f}
+- Risk distribution:
+  * Low risk (IFT < 1.0): {low_risk} predictions ({low_risk/len(predictions)*100:.1f}%)
+  * Moderate risk (1.0 ≤ IFT < 2.0): {moderate_risk} predictions ({moderate_risk/len(predictions)*100:.1f}%)
+  * High risk (IFT ≥ 2.0): {high_risk} predictions ({high_risk/len(predictions)*100:.1f}%)
+- Best plots (lowest IFT): {', '.join(predictions.nsmallest(3, 'predicted_ift')['plot_name'].tolist())}
+- Method: Linear regression on historical IFT data
+        """
+        return info.strip()
+    except Exception as e:
+        return f"Error loading predictions summary: {str(e)}"
+
+# @gr.mcp.resource("agricultural://recommendations/sensitive-crops")
+def get_recommendations_summary() -> str:
+    """Get summary of plot recommendations for sensitive crops (pois, haricot)"""
+    try:
+        predictions = analyzer.predict_weed_pressure()
+        suitable_plots = predictions[predictions['risk_level'] == "Faible"].copy()
+        
+        if len(suitable_plots) == 0:
+            return "No plots recommended for sensitive crops - all plots have high predicted weed pressure"
+        
+        suitable_plots['recommendation_score'] = 100 - (suitable_plots['predicted_ift'] * 30)
+        suitable_plots = suitable_plots.sort_values('recommendation_score', ascending=False)
+        
+        top_plots = suitable_plots.head(5)
+        avg_score = suitable_plots['recommendation_score'].mean()
+        
+        info = f"""
+Sensitive Crop Recommendations (Pois, Haricot):
+- Suitable plots: {len(suitable_plots)}
+- Average recommendation score: {avg_score:.1f}/100
+- Top 5 recommended plots:
+"""
+        for i, (_, plot) in enumerate(top_plots.iterrows(), 1):
+            info += f"  {i}. {plot['plot_name']} - Score: {plot['recommendation_score']:.1f}, IFT: {plot['predicted_ift']:.2f}\n"
+        
+        info += f"- Criteria: IFT < 1.0 (low weed pressure)\n"
+        info += f"- Score formula: 100 - (predicted_ift × 30)\n"
+        info += f"- Method: Based on 2025-2027 predictions"
+        
+        return info.strip()
+    except Exception as e:
+        return f"Error loading recommendations summary: {str(e)}"
+
+# @gr.mcp.resource("agricultural://plot/{plot_name}/predictions")
+def get_plot_predictions(plot_name: str) -> str:
+    """Get weed pressure predictions for a specific plot"""
+    try:
+        predictions = analyzer.predict_weed_pressure()
+        plot_predictions = predictions[predictions['plot_name'] == plot_name]
+        
+        if len(plot_predictions) == 0:
+            return f"No predictions available for plot '{plot_name}' - insufficient historical data"
+        
+        plot_predictions = plot_predictions.sort_values('year')
+        
+        info = f"""
+Predictions for {plot_name}:
+"""
+        for _, pred in plot_predictions.iterrows():
+            info += f"- {pred['year']}: IFT {pred['predicted_ift']:.2f} ({pred['risk_level']} risk)\n"
+        
+        info += f"- Historical average IFT: {plot_predictions['historical_avg_ift'].iloc[0]:.2f}\n"
+        info += f"- Recent crops: {plot_predictions['recent_crops'].iloc[0]}\n"
+        info += f"- Recommendation: {'Suitable for sensitive crops' if plot_predictions['risk_level'].iloc[0] == 'Faible' else 'Not recommended for sensitive crops'}"
+        
+        return info.strip()
+    except Exception as e:
+        return f"Error loading plot predictions: {str(e)}"
+
 def analyze_herbicide_trends(year_start, year_end, plot_filter):
     """
     Analyze herbicide usage trends over time by calculating IFT (Treatment Frequency Index).

test_mcp_resources.py

@@ -0,0 +1,47 @@
+"""
+Test des resources MCP
+"""
+
+from mcp_server import (
+    get_available_plots_resource, 
+    get_available_crops_resource, 
+    get_available_years_resource,
+    get_dataset_info,
+    get_plot_info,
+    get_crop_info,
+    get_year_summary,
+    get_herbicide_usage_summary,
+    get_predictions_summary,
+    get_recommendations_summary,
+    get_plot_predictions
+)
+
+def test_mcp_resources():
+    """Test des resources MCP"""
+    print("🧪 Test des resources MCP...")
+    
+    # Test resources statiques
+    print("\n📋 Test resources statiques:")
+    print("Plots:", get_available_plots_resource())
+    print("Crops:", get_available_crops_resource())
+    print("Years:", get_available_years_resource())
+    print("Dataset info:", get_dataset_info()[:200] + "...")
+    
+    # Test resources avec paramètres
+    print("\n🏞️ Test resources avec paramètres:")
+    print("Plot info:", get_plot_info("Champ ferme W du sol")[:200] + "...")
+    print("Crop info:", get_crop_info("blé tendre hiver")[:200] + "...")
+    print("Year summary:", get_year_summary(2023)[:200] + "...")
+    
+    # Test resources d'analyse
+    print("\n📊 Test resources d'analyse:")
+    print("Herbicide usage:", get_herbicide_usage_summary()[:200] + "...")
+    print("Predictions:", get_predictions_summary()[:200] + "...")
+    print("Recommendations:", get_recommendations_summary()[:200] + "...")
+    
+    # Test resource prédictions par parcelle
+    print("\n🔮 Test prédictions par parcelle:")
+    print("Plot predictions:", get_plot_predictions("Champ ferme W du sol")[:200] + "...")
+
+if __name__ == "__main__":
+    test_mcp_resources()