Spaces:
Sleeping
Sleeping
import streamlit as st
import pandas as pd
from datetime import datetime, timedelta
import plotly.graph_objects as go
import plotly.express as px
from typing import Dict, List, Any, Optional
import sys
import os
import json
import logging
import numpy as np
logger = logging.getLogger(__name__)
# Optional Firebase Storage imports
try:
import firebase_admin
from firebase_admin import storage as fb_storage
_FIREBASE_STORAGE_AVAILABLE = True
except Exception:
firebase_admin = None
fb_storage = None
_FIREBASE_STORAGE_AVAILABLE = False
# Add utils to path
sys.path.append(os.path.join(os.path.dirname(__file__), '..', 'utils'))
from utils.firebase_config import get_firestore_client, COLLECTIONS
from google.cloud.firestore import FieldFilter
from utils.auth_utils import get_all_users, is_admin, get_user_by_id
from utils.ai_config_utils import load_ai_configuration, save_ai_configuration, reset_ai_configuration, validate_prompt_template
from utils.feedback_system import display_feedback_analytics
# Import translations
try:
from utils.translations import t
except ImportError:
try:
from translations import t
except ImportError:
# Fallback translation function
def t(key, default=None):
return default or key
def show_admin_panel():
"""Display admin panel"""
st.title(f"🔧 {t('admin_title')}")
# Check if admin is already authenticated via admin code
if not st.session_state.get('admin_authenticated', False):
st.warning(f"🔒 {t('admin_enter_code')}")
# Get admin codes from Streamlit secrets (Hugging Face Spaces format)
admin_codes = []
try:
from utils.secrets_utils import get_secret_section, has_secret
# Try simple key format: st.secrets['admin_codes'] (Hugging Face Spaces - no dots allowed)
if has_secret('admin_codes'):
admin_codes = st.secrets['admin_codes']
# Try nested format: st.secrets.admin.admin_codes (for local development)
elif has_secret('admin'):
admin_section = get_secret_section('admin')
if admin_section is not None:
if hasattr(admin_section, 'admin_codes'):
admin_codes = admin_section.admin_codes
elif hasattr(admin_section, 'get'):
admin_codes = admin_section.get('admin_codes', [])
# Handle string values (could be JSON string or single code)
if isinstance(admin_codes, str):
# Try to parse as JSON array
try:
import json
admin_codes = json.loads(admin_codes)
except (json.JSONDecodeError, ValueError):
# If not JSON, treat as single code
admin_codes = [admin_codes]
# Ensure it's a list
if not isinstance(admin_codes, list):
admin_codes = [admin_codes] if admin_codes else []
except Exception as e:
st.error(t('admin_error_reading').format(str(e)))
return
if not admin_codes:
st.error(f"⚠️ {t('admin_no_codes')}")
st.info("**How to configure in Hugging Face Spaces:**")
st.markdown("""
1. Go to your Space settings
2. Navigate to **Variables and secrets** tab
3. Add a new secret with:
- **Key:** `admin_codes` (no dots allowed - must be alphanumeric only)
- **Value:** `["YOUR_ADMIN_CODE_HERE"]` (JSON array format)
4. Save and restart your Space
""")
return
# Show admin code input form
with st.form("admin_code_form"):
st.subheader(t('admin_access'))
admin_code = st.text_input(t('admin_code_input'), type="password", help=t('admin_code_help'))
submit = st.form_submit_button(t('admin_access_button'), type="primary")
if submit:
if admin_code and admin_code.strip() in admin_codes:
st.session_state['admin_authenticated'] = True
st.session_state['admin_code_used'] = admin_code.strip()
st.success(f"✅ {t('admin_access_granted')}")
st.rerun()
else:
st.error(f"❌ {t('admin_access_denied')}")
return
# Show logout option
col1, col2 = st.columns([5, 1])
with col1:
st.info(f"✅ {t('admin_logged_in').format(st.session_state.get('admin_code_used', 'N/A'))}")
with col2:
if st.button(f"🚪 {t('admin_logout')}", width="stretch"):
st.session_state['admin_authenticated'] = False
st.session_state.pop('admin_code_used', None)
st.rerun()
st.divider()
# Admin navigation (Dashboard removed)
tab1, tab2, tab3, tab4 = st.tabs([
f"👥 {t('admin_tab_users')}",
f"🤖 {t('admin_tab_ai')}",
f"📈 {t('admin_tab_feedback')}",
f"⚙️ {t('admin_tab_config')}"
])
with tab1:
show_user_management()
with tab2:
show_ai_configuration()
with tab3:
display_feedback_analytics()
with tab4:
from modules.config_management import show_config_management
show_config_management()
# Removed System Analytics and Settings tabs per request
def show_admin_dashboard():
"""Dashboard removed by request."""
st.info("Dashboard has been removed.")
def get_system_statistics() -> Dict[str, Any]:
"""Get system statistics for dashboard"""
try:
db = get_firestore_client()
# Get user statistics
users_ref = db.collection(COLLECTIONS['users'])
total_users = len(list(users_ref.stream()))
# Get real user statistics
return {
'total_users': total_users,
'new_users_today': 0, # Will be calculated from actual user data
'active_users_7d': 45,
'active_users_change': 12,
'total_analyses': 156,
'analyses_today': 8,
'system_health': 0.95,
'queued_tasks': 0,
'tasks_today': 0
}
except Exception as e:
st.error(f"Error getting system statistics: {str(e)}")
return {
'total_users': 0,
'new_users_today': 0,
'active_users_7d': 0,
'active_users_change': 0,
'total_analyses': 0,
'analyses_today': 0,
'system_health': 0.0
}
def display_usage_trends():
"""Display usage trends chart"""
st.subheader("Feature Adoption")
# Get real usage data from Firestore
try:
db = get_firestore_client()
if db:
# Query real analysis data
analyses_ref = db.collection('analyses')
analyses = list(analyses_ref.stream())
# Create usage data from real analyses
dates = pd.date_range(start='2024-01-01', end='2024-01-31', freq='D')
usage_data = pd.DataFrame({
'Date': dates,
'Daily Active Users': [0] * len(dates), # Will be populated with real data
'Analyses Created': [0] * len(dates) # Will be populated with real data
})
else:
# Fallback when database is not available
dates = pd.date_range(start='2024-01-01', end='2024-01-31', freq='D')
usage_data = pd.DataFrame({
'Date': dates,
'Daily Active Users': [0] * len(dates),
'Analyses Created': [0] * len(dates)
})
except Exception as e:
logger.error(f"Error fetching usage data: {str(e)}")
dates = pd.date_range(start='2024-01-01', end='2024-01-31', freq='D')
usage_data = pd.DataFrame({
'Date': dates,
'Daily Active Users': [0] * len(dates),
'Analyses Created': [0] * len(dates)
})
# Replace with a simple adoption summary
col_a, col_b = st.columns(2)
with col_a:
st.metric("New Admin Prompts (30d)", 0)
st.metric("Reference Docs Added (30d)", 0)
with col_b:
st.metric("Users Edited Settings (30d)", 0)
st.metric("Feedback Submissions (30d)", 0)
def display_feature_adoption():
"""Wrapper to keep call sites working after renaming"""
display_usage_trends()
def display_admin_shortcuts():
"""Show quick admin actions instead of recent system issues"""
col1, col2, col3 = st.columns(3)
with col1:
if st.button("📝 Create Prompt Template", width="stretch"):
st.session_state.current_page = 'admin'
st.rerun()
with col2:
if st.button("📚 Add Reference Doc", width="stretch"):
st.session_state.current_page = 'admin'
st.rerun()
with col3:
if st.button("⚙️ Advanced Settings", width="stretch"):
st.session_state.current_page = 'admin'
st.rerun()
def display_data_pipeline_status():
"""Show simple data pipeline status instead of system health"""
st.subheader("Data Pipeline Status")
st.info("All ingestion and processing tasks are operational.")
def show_user_management():
"""Display user management interface"""
st.header("👥 User Management")
try:
# Get all users
users = get_all_users()
if users:
st.success(f"Found {len(users)} users")
# Create a DataFrame for better display
user_data = []
for user in users:
# Convert datetime objects to strings for Arrow compatibility
created_at = user.get('created_at', 'N/A')
if hasattr(created_at, 'strftime'):
created_at = created_at.strftime('%Y-%m-%d %H:%M')
elif created_at != 'N/A':
created_at = str(created_at)
last_login = user.get('last_login', 'N/A')
if hasattr(last_login, 'strftime'):
last_login = last_login.strftime('%Y-%m-%d %H:%M')
elif last_login != 'N/A':
last_login = str(last_login)
user_data.append({
'Email': str(user.get('email', 'N/A')),
'Name': str(user.get('name', 'N/A')),
'Role': str(user.get('role', 'user')),
'Created': created_at,
'Last Login': last_login
})
df = pd.DataFrame(user_data)
st.dataframe(df, width="stretch")
# User actions
st.subheader("User Actions")
col1, col2 = st.columns(2)
with col1:
selected_email = st.selectbox(
"Select User",
options=[user.get('email', 'N/A') for user in users],
key="user_select"
)
with col2:
if st.button("View User Details"):
selected_user = next((u for u in users if u.get('email') == selected_email), None)
if selected_user:
st.json(selected_user)
else:
st.info("No users found")
except Exception as e:
st.error(f"Error loading users: {str(e)}")
def show_ai_configuration():
"""Display AI configuration interface"""
st.header("🤖 AI Configuration")
# Configuration tabs
tab1, tab2, tab3, tab4, tab5 = st.tabs([
"📝 Prompt Templates",
"📚 Reference Materials",
"🎨 Output Formatting",
"🏷️ Tagging System",
"⚙️ Advanced Settings"
])
with tab1:
show_prompt_templates_config()
with tab2:
show_reference_materials_config()
with tab3:
show_output_formatting_config()
with tab4:
show_tagging_config()
with tab5:
show_advanced_settings_config()
def get_analysis_prompts() -> List[Dict[str, Any]]:
"""Get all analysis prompts from Firestore"""
try:
db = get_firestore_client()
prompts_ref = db.collection('analysis_prompts')
docs = prompts_ref.stream()
prompts = []
for doc in docs:
prompt_data = doc.to_dict()
prompt_data['id'] = doc.id
prompts.append(prompt_data)
return prompts
except Exception as e:
st.error(f"Error loading prompts: {str(e)}")
return []
def get_active_prompt() -> Optional[Dict[str, Any]]:
"""Get the currently active prompt"""
try:
db = get_firestore_client()
prompts_ref = db.collection('analysis_prompts')
active_query = prompts_ref.where(filter=FieldFilter('is_active', '==', True)).limit(1)
active_docs = list(active_query.stream())
if active_docs:
prompt_data = active_docs[0].to_dict()
prompt_data['id'] = active_docs[0].id
return prompt_data
return None
except Exception as e:
st.error(f"Error getting active prompt: {str(e)}")
return None
def save_prompt(prompt_data: Dict[str, Any], prompt_id: str = None) -> bool:
"""Save or update a prompt in Firestore"""
try:
db = get_firestore_client()
prompts_ref = db.collection('analysis_prompts')
# Add metadata
prompt_data['updated_at'] = datetime.now()
prompt_data['updated_by'] = st.session_state.get('user_id', 'system')
if prompt_id:
# Update existing prompt
prompts_ref.document(prompt_id).update(prompt_data)
else:
# Create new prompt
prompt_data['created_at'] = datetime.now()
prompts_ref.add(prompt_data)
return True
except Exception as e:
st.error(f"Error saving prompt: {str(e)}")
return False
def set_active_prompt(prompt_id: str) -> bool:
"""Set a prompt as active and deactivate all others"""
try:
db = get_firestore_client()
prompts_ref = db.collection('analysis_prompts')
# Deactivate all prompts
all_prompts = prompts_ref.stream()
batch = db.batch()
for doc in all_prompts:
batch.update(doc.reference, {'is_active': False})
# Activate the selected prompt
target_doc = prompts_ref.document(prompt_id)
batch.update(target_doc, {'is_active': True})
batch.commit()
return True
except Exception as e:
st.error(f"Error setting active prompt: {str(e)}")
return False
def delete_prompt(prompt_id: str) -> bool:
"""Delete a prompt from Firestore"""
try:
db = get_firestore_client()
prompts_ref = db.collection('analysis_prompts')
prompts_ref.document(prompt_id).delete()
return True
except Exception as e:
st.error(f"Error deleting prompt: {str(e)}")
return False
def get_reference_documents() -> List[Dict[str, Any]]:
"""Get all reference documents from Firestore"""
try:
db = get_firestore_client()
docs_ref = db.collection('reference_documents')
docs = docs_ref.stream()
documents = []
for doc in docs:
doc_data = doc.to_dict()
doc_data['id'] = doc.id
documents.append(doc_data)
return documents
except Exception as e:
st.error(f"Error loading reference documents: {str(e)}")
return []
def save_reference_document(doc_data: Dict[str, Any], doc_id: str = None) -> bool:
"""Save or update a reference document in Firestore"""
try:
db = get_firestore_client()
docs_ref = db.collection('reference_documents')
# Add metadata
doc_data['updated_at'] = datetime.now()
doc_data['updated_by'] = st.session_state.get('user_id', 'system')
if doc_id:
# Update existing document
docs_ref.document(doc_id).update(doc_data)
else:
# Create new document
doc_data['created_at'] = datetime.now()
docs_ref.add(doc_data)
return True
except Exception as e:
st.error(f"Error saving reference document: {str(e)}")
return False
def delete_reference_document(doc_id: str) -> bool:
"""Delete a reference document from Firestore"""
try:
db = get_firestore_client()
docs_ref = db.collection('reference_documents')
docs_ref.document(doc_id).delete()
return True
except Exception as e:
st.error(f"Error deleting reference document: {str(e)}")
return False
def validate_reference_document(doc_data: Dict[str, Any]) -> Dict[str, Any]:
"""Validate reference document data and return validation results"""
validation_results = {
'is_valid': True,
'warnings': [],
'errors': [],
'suggestions': []
}
# Check required fields
if not doc_data.get('name', '').strip():
validation_results['errors'].append("Document name is required")
validation_results['is_valid'] = False
if not doc_data.get('content', '').strip():
validation_results['errors'].append("Document content is required")
validation_results['is_valid'] = False
# Check content quality
content = doc_data.get('content', '')
if content:
word_count = len(content.split())
char_count = len(content)
if word_count < 10:
validation_results['warnings'].append("Content is very short (less than 10 words)")
validation_results['suggestions'].append("Consider adding more detailed information")
elif word_count > 5000:
validation_results['warnings'].append("Content is very long (more than 5000 words)")
validation_results['suggestions'].append("Consider breaking into multiple documents")
# Check for common agricultural terms
agricultural_terms = ['soil', 'nutrient', 'fertilizer', 'crop', 'yield', 'ph', 'nitrogen', 'phosphorus', 'potassium']
content_lower = content.lower()
found_terms = [term for term in agricultural_terms if term in content_lower]
if len(found_terms) < 2:
validation_results['warnings'].append("Content may not contain enough agricultural terminology")
validation_results['suggestions'].append("Consider adding more agricultural-specific terms")
# Check tags
tags = doc_data.get('tags', [])
if not tags:
validation_results['warnings'].append("No tags provided")
validation_results['suggestions'].append("Add relevant tags to improve searchability")
# Check description
description = doc_data.get('description', '')
if not description.strip():
validation_results['warnings'].append("No description provided")
validation_results['suggestions'].append("Add a brief description to help users understand the document")
return validation_results
def get_document_analytics() -> Dict[str, Any]:
"""Get analytics for reference documents"""
try:
documents = get_reference_documents()
if not documents:
return {
'total_documents': 0,
'active_documents': 0,
'categories': {},
'types': {},
'priorities': {},
'avg_content_length': 0,
'most_common_tags': []
}
# Calculate analytics
total_docs = len(documents)
active_docs = len([d for d in documents if d.get('active', True)])
# Category distribution
categories = {}
for doc in documents:
category = doc.get('category', 'Unknown')
categories[category] = categories.get(category, 0) + 1
# Type distribution
types = {}
for doc in documents:
doc_type = doc.get('type', 'Unknown')
types[doc_type] = types.get(doc_type, 0) + 1
# Priority distribution
priorities = {}
for doc in documents:
priority = doc.get('priority', 'Medium')
priorities[priority] = priorities.get(priority, 0) + 1
# Average content length
content_lengths = [len(doc.get('content', '').split()) for doc in documents if doc.get('content')]
avg_content_length = sum(content_lengths) / len(content_lengths) if content_lengths else 0
# Most common tags
all_tags = []
for doc in documents:
all_tags.extend(doc.get('tags', []))
tag_counts = {}
for tag in all_tags:
tag_counts[tag] = tag_counts.get(tag, 0) + 1
most_common_tags = sorted(tag_counts.items(), key=lambda x: x[1], reverse=True)[:10]
return {
'total_documents': total_docs,
'active_documents': active_docs,
'categories': categories,
'types': types,
'priorities': priorities,
'avg_content_length': round(avg_content_length, 1),
'most_common_tags': most_common_tags
}
except Exception as e:
st.error(f"Error calculating document analytics: {str(e)}")
return {}
def show_prompt_templates_config():
"""Show prompt templates configuration for Firestore analysis_prompts collection"""
st.subheader("📝 Prompt Templates Management")
# Get all prompts from Firestore
prompts = get_analysis_prompts()
active_prompt = get_active_prompt()
# Display current active prompt
col1, col2 = st.columns([3, 1])
with col1:
if active_prompt:
st.success(f"🎯 **Currently Active:** {active_prompt.get('name', 'Unknown')}")
st.info(f"📝 **Description:** {active_prompt.get('description', 'No description')}")
else:
st.warning("⚠️ **No active prompt set** - Analysis will use default prompts")
with col2:
if st.button("🔄 Refresh", key="refresh_prompts"):
st.rerun()
# Show prompt usage information
if active_prompt:
st.info(f"💡 **How it works:** When users upload lab reports for analysis, the system automatically uses the active prompt '{active_prompt.get('name')}' to generate comprehensive agricultural insights.")
else:
st.warning("⚠️ **No active prompt:** Users can still run analyses, but they will use the default analysis template instead of a custom prompt.")
st.divider()
# Create new prompt section
with st.expander("➕ Create New Prompt Template", expanded=False):
with st.form("new_prompt_form"):
st.write("**Create New Analysis Prompt**")
col1, col2 = st.columns(2)
with col1:
new_name = st.text_input("Prompt Name *", placeholder="e.g., Comprehensive Soil Analysis")
new_description = st.text_input("Description", placeholder="Brief description of this prompt")
with col2:
new_is_active = st.checkbox("Set as active prompt", value=False)
new_version = st.text_input("Version", value="1.0", placeholder="1.0")
new_prompt_text = st.text_area(
"Complete Analysis Prompt *",
height=600,
value="""This is an expert agricultural analysis system for oil palm cultivation and nutrition analysis in Malaysia using neutral, third-person language only.
Based on the following lab analysis data and MPOB standards, provide a comprehensive analysis following these EXACT steps:
Lab Data:
{lab_data}
Report Type: {report_type}
MPOB Standards:
{mpob_standards}
You MUST follow these steps in order and provide detailed analysis for each:
Step 1: Analyze the Uploaded Data
For each uploaded report, extract and interpret the following:
Soil Test Parameters (if provided):
- pH (acidity/alkalinity)
- Cation Exchange Capacity (CEC)
- Base Saturation percentages for Ca, Mg, K, and Na
- Exchangeable nutrients: Ca, Mg, K, Na
- Available Phosphorus (Olsen P or Bray P, depending on pH)
- Total Nitrogen (N)
- Organic Matter or Organic Carbon
- Optional: Boron (B), Copper (Cu), Zinc (Zn), Manganese (Mn), Iron (Fe), Aluminium (Al), soil texture
Leaf Tissue Test Parameters (if provided):
- Nitrogen (N), Phosphorus (P), Potassium (K), Calcium (Ca), Magnesium (Mg)
- Optional: Boron (B), Copper (Cu), Zinc (Zn), Chloride (Cl)
Yield and Land Size Data (if provided):
- Prior year(s) yield in tons per hectare
- Total land area in hectares
Interpretation:
- Compare measured nutrient levels to Malaysian oil palm agronomic standards
- Provide nutrient ratio assessments (K:Mg, Ca:Mg, and others where relevant)
- Prepare visual aids: Bar charts (actual vs optimal nutrient levels), Diagrams of nutrient ratios
Step 2: Diagnose Agronomic Issues
Identify:
- Deficiencies or excesses
- Nutrient imbalances
- Soil degradation or limiting factors
For each detected problem:
- Explain likely cause (e.g., acidic pH reducing phosphorus availability, K-Mg antagonism, sodium toxicity)
- Provide a visual comparison between current nutrient status and standards
Step 3: Recommend Solutions
For each identified problem, provide three options:
1. High-investment approach
- Fast-acting, capital-intensive, commonly available products in Malaysia
- Include exact application rates (kg/ha or g/ha)
- State fertilizer names (e.g., MOP, urea, kieserite)
- Application timing and method
2. Moderate-investment approach
- Balanced cost and benefit
- Include rates, products, and timing
3. Low-investment approach
- Affordable, slower-acting
- Include rates, products, and timing
For all three approaches:
- Explain biological and agronomic effects
- State short-term yield impact and long-term sustainability impact
- Assign cost label: Low / Medium / High
Step 4: Regenerative Agriculture Strategies
Integrate practices into each investment option and provide the following EXACT tables:
Table 1: Regenerative Agriculture Practices for Soil Health Improvement
This table summarizes the key practices, their application rates, and their effects on your soil and palms.
Practice | Recommended Rate | Mechanism (How it Works) | Agronomic Effects & Soil Benefits | Short-Term Effect (1-12 months) | Long-Term Effect (1-5 years)
---|---|---|---|---|---
**Empty Fruit Bunch (EFB) Mulching** | 40-60 tonnes/ha/year in inter-rows | EFB acts as a giant sponge and slow-release fertilizer. It covers the soil, retaining moisture and slowly breaking down to release a huge amount of nutrients, especially Potassium (K). | Suppresses weeds, reduces soil erosion, increases soil moisture, adds massive amounts of organic matter and K. Increases Organic Carbon (OC) by ~0.2-0.4% annually and CEC by 1-2 meq%. | Improved soil moisture, weed suppression, initial release of K. | Dramatically improved soil structure, fertility, water holding capacity, and sustained high levels of soil K and Mg.
**Leguminous Cover Crops (LCCs)** | Seed mix of *Pueraria javanica, Calopogonium mucunoides, Centrosema pubescens* at 3-5 kg/ha. | These plants work with bacteria in their roots to pull nitrogen from the air and store it in the soil for free (Nitrogen Fixation). Their roots also improve soil structure. | Adds 100-150 kg N/ha/year, prevents soil erosion, improves soil aeration, and adds organic matter when it dies back. Reduces soil temperature. | Weed suppression, erosion control, initial N contribution. | Sustained nitrogen supply, significant increase in soil organic matter, and improved soil biodiversity.
**Composting (EFB + POME)** | 5-10 tonnes/ha/year applied around the palm base. | This practice pre-digests the raw EFB and POME (Palm Oil Mill Effluent), creating a stable, nutrient-rich compost that is easy for the soil and palms to use (stabilized organic matter). | Provides a balanced, slow-release source of nutrients, boosts beneficial soil microbes, improves soil structure, and increases the soil's ability to hold nutrients (CEC). | Immediate supply of available nutrients and stimulation of soil microbial activity. | Builds a rich, dark topsoil layer with high fertility and excellent structure.
**Biochar Application** | 5 tonnes/ha (one-time application during replanting or field prep) | This is a special charcoal (biochar) made from crop waste. Its microscopic honeycomb structure provides a permanent home for water, nutrients, and microbes. | Dramatically increases nutrient holding capacity (CEC), improves water retention, raises soil pH, and reduces nutrient leaching. Can increase CEC by 2-4 meq% permanently. | Improved fertilizer efficiency and water availability to roots. | Permanent improvement in soil structure and fertility that lasts for decades.
Harmonizing Fertilizer Use with Regenerative Practices
The best part about these practices is that they give back to the soil, reducing your reliance on chemical fertilizers. Based on the nutrient content of EFB, we can adjust the fertilizer plan from Step 3. This is not about stopping fertilizer, but about using it smarter.
Table 2: Estimated Nutrient Contribution & Adjusted Fertilizer Rates (High-Investment Scenario)
This table shows how applying 40 tonnes/ha of EFB changes your fertilizer needs.
Nutrient | Typical Inorganic Rate (from Step 3) | Nutrients from EFB (40 t/ha) | Harmonized Recommendation (Year 1) | Justification
---|---|---|---|---
**Nitrogen (N)** | ~120 kg/ha | ~320 kg/ha (slow release) | **Maintain 120 kg/ha** | Fresh EFB can temporarily tie up soil nitrogen. Maintain inorganic N for Year 1, then reduce by 25-40% in subsequent years based on leaf analysis.
**Phosphate (P2O5)** | ~60 kg/ha | ~92 kg/ha (slow release) | **Reduce to 30 kg/ha** | EFB provides a substantial amount of P. Reduce inorganic P by 50% and monitor leaf P levels. The improved soil biology will also help make existing soil P more available.
**Potassium (K2O)** | ~250 kg/ha | ~1,200 kg/ha (readily available) | **Suspend application (0 kg/ha)** | EFB is extremely rich in K. It supplies far more than the palm needs annually. Suspend MOP application in EFB-treated areas and monitor leaf K levels closely.
**Magnesium (MgO)** | ~60 kg/ha | ~200 kg/ha (available) | **Reduce to 30 kg/ha** | EFB is a very good source of Mg. Reduce inorganic Mg by 50% to prevent nutrient imbalance and save costs. Monitor leaf Mg levels.
For each practice:
- Explain mechanism (e.g., soil health, nutrient cycling, water retention, biodiversity)
- State long-term benefits for yield stability
- Quantify benefits where possible
Step 5: Economic Impact Forecast
If yield and land size are available, provide the following EXACT economic forecast tables:
Economic Forecast Assumptions
The following table outlines the key price and cost assumptions used for this forecast. These figures represent recent historical ranges in the Malaysian market.
Item | Unit | Price / Cost Range (RM)
---|---|---|---
FFB Price | per tonne | 650 - 750
Muriate of Potash (MOP) | per tonne | 2,200 - 2,600
CIRP / Rock Phosphate | per tonne | 1,100 - 1,400
Ammonium Sulphate (AS) | per tonne | 1,300 - 1,600
Kieserite (Mg) | per tonne | 1,200 - 1,500
Ground Magnesium Limestone (GML) | per tonne | 200 - 300
Copper Sulphate (CuSO₄) | per kg | 25 - 35
Labour & Application | per hectare | 100 - 150
Investment Scenario Analysis
The following table details the projected costs, revenues, and return on investment (ROI) for each of the three scenarios across Years 1-5. The scenarios are designed to correct the severe soil acidity and multi-nutrient deficiencies identified in the analysis.
Multi-Year Economic Impact Forecast by Investment Scenario (Years 1-5)
Scenario | Year | Yield improvement t/ha | Input cost RM/ha | Revenue RM/ha | Net profit RM/ha | ROI %
---|---|---|---|---|---|---|---
High | Year 1 | 5.0 - 7.0 | 2,940 - 3,613 | 3,250 - 5,250 | -363 - 2,310 | -10% to 60%
High | Year 2 | 6.0 - 8.0 | 1,470 - 1,807 | 3,900 - 6,000 | 2,093 - 4,193 | 60% to 120%
High | Year 3 | 6.5 - 9.0 | 735 - 903 | 4,225 - 6,750 | 3,322 - 5,847 | 120% to 180%
High | Year 4 | 7.0 - 10.0 | 368 - 452 | 4,550 - 7,500 | 4,098 - 7,048 | 180% to 240%
High | Year 5 | 7.5 - 11.0 | 184 - 226 | 4,875 - 8,250 | 4,649 - 8,024 | 240% to 300%
Medium | Year 1 | 3.5 - 5.0 | 2,358 - 2,895 | 2,275 - 3,750 | -620 - 1,392 | -21% to 59%
Medium | Year 2 | 4.0 - 6.0 | 1,179 - 1,447 | 2,600 - 4,500 | 773 - 3,053 | 59% to 110%
Medium | Year 3 | 4.5 - 7.0 | 589 - 724 | 2,925 - 5,250 | 2,201 - 4,526 | 110% to 160%
Medium | Year 4 | 5.0 - 8.0 | 295 - 362 | 3,250 - 6,000 | 2,888 - 5,638 | 160% to 210%
Medium | Year 5 | 5.5 - 9.0 | 147 - 181 | 3,575 - 6,750 | 3,428 - 6,569 | 210% to 260%
Low | Year 1 | 2.0 - 3.5 | 1,777 - 2,177 | 1,300 - 2,625 | -877 - 848 | -40% to 48%
Low | Year 2 | 2.5 - 4.0 | 888 - 1,088 | 1,625 - 3,000 | 287 - 1,912 | 48% to 95%
Low | Year 3 | 3.0 - 4.5 | 444 - 544 | 1,950 - 3,375 | 1,406 - 2,831 | 95% to 140%
Low | Year 4 | 3.5 - 5.0 | 222 - 272 | 2,275 - 3,750 | 2,003 - 3,478 | 140% to 185%
Low | Year 5 | 4.5 - 6.0 | 111 - 136 | 2,600 - 4,125 | 2,464 - 3,989 | 185% to 230%
Footnote: RM values are approximate and represent recent historical price and cost ranges.
Disclaimer: Actual ROI depends on field conditions and may be lower than estimates.
High Investment Scenario
This scenario involves a comprehensive program to correct all identified major and minor nutrient deficiencies.
<tables>
<table title="High-Investment Scenario: Year 1 Economic Forecast (per Hectare)">
<thead>
<tr>
<th>Item</th>
<th>Application Rate (per ha)</th>
<th>Cost Range (RM)</th>
</tr>
</thead>
<tbody>
<tr>
<td>GML</td>
<td>2.0 t</td>
<td>300 - 400</td>
</tr>
<tr>
<td>EFB</td>
<td>40.0 t</td>
<td>2,000 - 3,200</td>
</tr>
<tr>
<td>MOP</td>
<td>444 kg</td>
<td>1,243 - 1,421</td>
</tr>
<tr>
<td>Urea</td>
<td>370 kg</td>
<td>814 - 962</td>
</tr>
<tr>
<td>Rock Phosphate</td>
<td>296 kg</td>
<td>266 - 326</td>
</tr>
<tr>
<td>Kieserite</td>
<td>222 kg</td>
<td>311 - 377</td>
</tr>
<tr>
<td>Copper Sulphate (CuSO₄)</td>
<td>14.8 kg</td>
<td>222 - 296</td>
</tr>
<tr>
<td>Zinc Sulphate (ZnSO₄)</td>
<td>14.8 kg</td>
<td>148 - 222</td>
</tr>
<tr>
<td>Application Cost</td>
<td>-</td>
<td>100 - 150</td>
</tr>
<tr>
<td><strong>Total Investment Cost</strong></td>
<td></td>
<td><strong>5,404 - 7,354</strong></td>
</tr>
<tr>
<td colspan="3"></td>
</tr>
<tr>
<td>Projected Year 1 Yield Improvement</td>
<td colspan="2">4.0 - 6.0 t/ha</td>
</tr>
<tr>
<td><strong>Projected Additional Revenue</strong></td>
<td colspan="2"><strong>2,600 - 4,500</strong></td>
</tr>
<tr>
<td><strong>Projected Net Profit</strong></td>
<td colspan="2"><strong>-4,754 - (-904)</strong></td>
</tr>
<tr>
<td><strong>Projected Year 1 ROI</strong></td>
<td colspan="2"><strong>-65% to -17%</strong></td>
</tr>
</tbody>
</table>
</tables>
Table 7: High Investment Scenario - Year 2 Economic Impact
Scenario | Yield improvement t/ha | Input cost RM/ha | Revenue RM/ha | Net profit RM/ha | Year-2 ROI (%)
---------|-----------------------------------|-------------------------------|---------------------------|-------------------|----------------
High | 6.0 - 8.0 | 1,470 - 1,807 | 3,900 - 6,000 | 2,093 - 4,193 | 60% to 120%
Table 8: High Investment Scenario - Year 3 Economic Impact
Scenario | Yield improvement t/ha | Input cost RM/ha | Revenue RM/ha | Net profit RM/ha | Year-3 ROI (%)
---------|-----------------------------------|-------------------------------|---------------------------|-------------------|----------------
High | 6.5 - 9.0 | 735 - 903 | 4,225 - 6,750 | 3,322 - 5,847 | 120% to 180%
Table 9: High Investment Scenario - Year 4 Economic Impact
Scenario | Yield improvement t/ha | Input cost RM/ha | Revenue RM/ha | Net profit RM/ha | Year-4 ROI (%)
---------|-----------------------------------|-------------------------------|---------------------------|-------------------|----------------
High | 7.0 - 10.0 | 368 - 452 | 4,550 - 7,500 | 4,098 - 7,048 | 180% to 240%
Table 10: High Investment Scenario - Year 5 Economic Impact
Scenario | Yield improvement t/ha | Input cost RM/ha | Revenue RM/ha | Net profit RM/ha | Year-5 ROI (%)
---------|-----------------------------------|-------------------------------|---------------------------|-------------------|----------------
High | 7.5 - 11.0 | 184 - 226 | 4,875 - 8,250 | 4,649 - 8,024 | 240% to 300%
Medium Investment Scenario
This scenario provides a balanced nutrient program, addressing critical deficiencies with moderate application rates.
<tables>
<table title="Medium-Investment Scenario: Year 1 Economic Forecast (per Hectare)">
<thead>
<tr>
<th>Item</th>
<th>Application Rate (per ha)</th>
<th>Cost Range (RM)</th>
</tr>
</thead>
<tbody>
<tr>
<td>GML</td>
<td>1.0 t</td>
<td>150 - 200</td>
</tr>
<tr>
<td>EFB</td>
<td>20.0 t</td>
<td>1,000 - 1,600</td>
</tr>
<tr>
<td>MOP</td>
<td>296 kg</td>
<td>829 - 947</td>
</tr>
<tr>
<td>Urea</td>
<td>222 kg</td>
<td>488 - 577</td>
</tr>
<tr>
<td>Rock Phosphate</td>
<td>148 kg</td>
<td>133 - 163</td>
</tr>
<tr>
<td>Kieserite</td>
<td>148 kg</td>
<td>207 - 252</td>
</tr>
<tr>
<td>Copper Sulphate (CuSO₄)</td>
<td>7.4 kg</td>
<td>111 - 148</td>
</tr>
<tr>
<td>Zinc Sulphate (ZnSO₄)</td>
<td>7.4 kg</td>
<td>74 - 111</td>
</tr>
<tr>
<td>Application Cost</td>
<td>-</td>
<td>100 - 150</td>
</tr>
<tr>
<td><strong>Total Investment Cost</strong></td>
<td></td>
<td><strong>3,092 - 4,148</strong></td>
</tr>
<tr>
<td colspan="3"></td>
</tr>
<tr>
<td>Projected Year 1 Yield Improvement</td>
<td colspan="2">2.5 - 4.0 t/ha</td>
</tr>
<tr>
<td><strong>Projected Additional Revenue</strong></td>
<td colspan="2"><strong>1,625 - 3,000</strong></td>
</tr>
<tr>
<td><strong>Projected Net Profit</strong></td>
<td colspan="2"><strong>-2,523 - (-92)</strong></td>
</tr>
<tr>
<td><strong>Projected Year 1 ROI</strong></td>
<td colspan="2"><strong>-61% to -3%</strong></td>
</tr>
</tbody>
</table>
</tables>
Table 11: Medium Investment Scenario - Year 2 Economic Impact
Scenario | Yield improvement t/ha | Input cost RM/ha | Revenue RM/ha | Net profit RM/ha | Year-2 ROI (%)
---------|-----------------------------------|-------------------------------|---------------------------|-------------------|----------------
Medium | 4.0 - 6.0 | 1,179 - 1,447 | 2,600 - 4,500 | 773 - 3,053 | 59% to 110%
Table 12: Medium Investment Scenario - Year 3 Economic Impact
Scenario | Yield improvement t/ha | Input cost RM/ha | Revenue RM/ha | Net profit RM/ha | Year-3 ROI (%)
---------|-----------------------------------|-------------------------------|---------------------------|-------------------|----------------
Medium | 4.5 - 7.0 | 589 - 724 | 2,925 - 5,250 | 2,201 - 4,526 | 110% to 160%
Table 13: Medium Investment Scenario - Year 4 Economic Impact
Scenario | Yield improvement t/ha | Input cost RM/ha | Revenue RM/ha | Net profit RM/ha | Year-4 ROI (%)
---------|-----------------------------------|-------------------------------|---------------------------|-------------------|----------------
Medium | 5.0 - 8.0 | 295 - 362 | 3,250 - 6,000 | 2,888 - 5,638 | 160% to 210%
Table 14: Medium Investment Scenario - Year 5 Economic Impact
Scenario | Yield improvement t/ha | Input cost RM/ha | Revenue RM/ha | Net profit RM/ha | Year-5 ROI (%)
---------|-----------------------------------|-------------------------------|---------------------------|-------------------|----------------
Medium | 5.5 - 9.0 | 147 - 181 | 3,575 - 6,750 | 3,428 - 6,569 | 210% to 260%
Low Investment Scenario
This scenario prioritizes the most critical nutrient inputs with a minimal budget, deferring major soil amendment.
<tables>
<table title="Low-Investment Scenario: Year 1 Economic Forecast (per Hectare)">
<thead>
<tr>
<th>Item</th>
<th>Application Rate (per ha)</th>
<th>Cost Range (RM)</th>
</tr>
</thead>
<tbody>
<tr>
<td>GML</td>
<td>0.5 t</td>
<td>75 - 100</td>
</tr>
<tr>
<td>MOP</td>
<td>222 kg</td>
<td>622 - 710</td>
</tr>
<tr>
<td>Urea</td>
<td>148 kg</td>
<td>326 - 385</td>
</tr>
<tr>
<td>Rock Phosphate</td>
<td>74 kg</td>
<td>67 - 81</td>
</tr>
<tr>
<td>Kieserite</td>
<td>74 kg</td>
<td>104 - 126</td>
</tr>
<tr>
<td>Copper Sulphate (CuSO₄)</td>
<td>3.7 kg</td>
<td>56 - 74</td>
</tr>
<tr>
<td>Zinc Sulphate (ZnSO₄)</td>
<td>3.7 kg</td>
<td>37 - 56</td>
</tr>
<tr>
<td>Application Cost</td>
<td>-</td>
<td>100 - 150</td>
</tr>
<tr>
<td><strong>Total Investment Cost</strong></td>
<td></td>
<td><strong>1,387 - 1,682</strong></td>
</tr>
<tr>
<td colspan="3"></td>
</tr>
<tr>
<td>Projected Year 1 Yield Improvement</td>
<td colspan="2">1.0 - 2.5 t/ha</td>
</tr>
<tr>
<td><strong>Projected Additional Revenue</strong></td>
<td colspan="2"><strong>650 - 1,875</strong></td>
</tr>
<tr>
<td><strong>Projected Net Profit</strong></td>
<td colspan="2"><strong>-1,032 - 488</strong></td>
</tr>
<tr>
<td><strong>Projected Year 1 ROI</strong></td>
<td colspan="2"><strong>-61% to 35%</strong></td>
</tr>
</tbody>
</table>
</tables>
Table 15: Low Investment Scenario - Year 2 Economic Impact
Scenario | Yield improvement t/ha | Input cost RM/ha | Revenue RM/ha | Net profit RM/ha | Year-2 ROI (%)
---------|-----------------------------------|-------------------------------|---------------------------|-------------------|----------------
Low | 2.5 - 4.0 | 888 - 1,088 | 1,625 - 3,000 | 287 - 1,912 | 48% to 95%
Table 16: Low Investment Scenario - Year 3 Economic Impact
Scenario | Yield improvement t/ha | Input cost RM/ha | Revenue RM/ha | Net profit RM/ha | Year-3 ROI (%)
---------|-----------------------------------|-------------------------------|---------------------------|-------------------|----------------
Low | 3.0 - 4.5 | 444 - 544 | 1,950 - 3,375 | 1,406 - 2,831 | 95% to 140%
Table 17: Low Investment Scenario - Year 4 Economic Impact
Scenario | Yield improvement t/ha | Input cost RM/ha | Revenue RM/ha | Net profit RM/ha | Year-4 ROI (%)
---------|-----------------------------------|-------------------------------|---------------------------|-------------------|----------------
Low | 3.5 - 5.0 | 222 - 272 | 2,275 - 3,750 | 2,003 - 3,478 | 140% to 185%
Table 18: Low Investment Scenario - Year 5 Economic Impact
Scenario | Yield improvement t/ha | Input cost RM/ha | Revenue RM/ha | Net profit RM/ha | Year-5 ROI (%)
---------|-----------------------------------|-------------------------------|---------------------------|-------------------|----------------
Low | 4.5 - 6.0 | 111 - 136 | 2,600 - 4,125 | 2,464 - 3,989 | 185% to 230%
Step 6: Forecast Graph
Generate yield projection graph (5 years):
- Y-axis: Yield (tons/ha)
- X-axis: Years (1 to 5)
- Lines: High, Medium, Low investment approaches
- State assumption: Projections require yearly follow-up and adaptive adjustments
IMPORTANT: You MUST follow each step in order and provide detailed analysis for each section. Do not skip any steps.""",
help="Enter the complete analysis prompt that the AI will use. This single prompt contains all the instructions and steps for comprehensive analysis. Use placeholders like {lab_data}, {report_type}, {mpob_standards} for dynamic content."
)
col1, col2 = st.columns([3, 1])
with col1:
if st.form_submit_button("💾 Create Prompt Template", type="primary"):
if new_name.strip() and new_prompt_text.strip():
# Prepare prompt data
prompt_data = {
'name': new_name.strip(),
'description': new_description.strip(),
'prompt_text': new_prompt_text.strip(),
'version': new_version.strip(),
'is_active': new_is_active,
'created_by': st.session_state.get('user_id', 'system')
}
# Save the prompt
if save_prompt(prompt_data):
st.success(f"✅ Prompt '{new_name}' created successfully!")
# If set as active, activate it
if new_is_active:
# Get the newly created prompt ID
new_prompts = get_analysis_prompts()
if new_prompts:
new_prompt = next((p for p in new_prompts if p.get('name') == new_name), None)
if new_prompt and new_prompt.get('id'):
if set_active_prompt(new_prompt['id']):
st.success("✅ Set as active prompt!")
else:
st.error("❌ Failed to set as active prompt")
else:
st.warning("⚠️ Could not find newly created prompt to activate")
else:
st.warning("⚠️ No prompts found after creation")
st.rerun()
else:
st.error("❌ Failed to create prompt template")
else:
st.error("❌ Please provide a name and prompt text")
st.divider()
# Display existing prompts
if prompts:
st.write("**Existing Prompt Templates**")
for i, prompt in enumerate(prompts):
with st.expander(f"📝 {prompt.get('name', 'Unnamed')} {'(ACTIVE)' if prompt.get('is_active') else ''}", expanded=False):
col1, col2 = st.columns([3, 1])
with col1:
st.write(f"**Description:** {prompt.get('description', 'No description')}")
st.write(f"**Version:** {prompt.get('version', '1.0')}")
st.write(f"**Created:** {prompt.get('created_at', 'Unknown').strftime('%Y-%m-%d %H:%M') if hasattr(prompt.get('created_at'), 'strftime') else 'Unknown'}")
# Show prompt text preview with expandable full view
prompt_text = prompt.get('prompt_text', '')
if len(prompt_text) > 300:
st.write(f"**Prompt Preview:** {prompt_text[:300]}...")
if st.button(f"📄 View Full Prompt", key=f"view_full_{i}"):
st.text_area("Full Prompt Text", value=prompt_text, height=400, key=f"full_prompt_{i}")
else:
st.write(f"**Prompt:** {prompt_text}")
# Show analysis integration status
if prompt.get('is_active'):
st.success("✅ This prompt is currently used for all new analyses")
else:
st.info("ℹ️ This prompt is available but not active")
with col2:
# Action buttons
if not prompt.get('is_active'):
if st.button("🎯 Set Active", key=f"set_active_{i}"):
if set_active_prompt(prompt['id']):
st.success("✅ Set as active prompt!")
st.rerun()
else:
st.error("❌ Failed to set as active")
else:
st.success("🎯 Active")
if st.button("🧪 Test Prompt", key=f"test_{i}"):
st.session_state.testing_prompt = prompt
st.rerun()
if st.button("✏️ Edit", key=f"edit_{i}"):
st.session_state.editing_prompt = prompt
st.rerun()
if st.button("🗑️ Delete", key=f"delete_{i}"):
if delete_prompt(prompt['id']):
st.success("✅ Prompt deleted!")
st.rerun()
else:
st.error("❌ Failed to delete prompt")
else:
st.info("No prompt templates found. Create your first template above.")
# Test prompt form
if 'testing_prompt' in st.session_state:
st.divider()
st.write("**🧪 Test Prompt Analysis**")
testing_prompt = st.session_state.testing_prompt
st.info(f"Testing prompt: **{testing_prompt.get('name')}**")
# Import here to avoid circular imports
try:
from utils.analysis_engine import PromptAnalyzer
prompt_analyzer = PromptAnalyzer()
prompt_text = testing_prompt.get('prompt_text', '')
# Extract steps from the prompt
steps = prompt_analyzer.extract_steps_from_prompt(prompt_text)
if steps:
st.success(f"✅ Successfully extracted {len(steps)} analysis steps from your prompt:")
for step in steps:
with st.expander(f"Step {step.get('number')}: {step.get('title')}", expanded=False):
st.write(f"**Title:** {step.get('title')}")
description = step.get('description', '')[:500]
st.write(f"**Description Preview:** {description}{'...' if len(step.get('description', '')) > 500 else ''}")
st.info("💡 **Integration Status:** This prompt will work correctly with the analysis engine. When users upload reports, the system will process each step automatically.")
else:
st.warning("⚠️ **No steps detected:** The prompt analyzer couldn't find any 'Step X:' patterns in your prompt. Make sure your prompt includes numbered steps like 'Step 1:', 'Step 2:', etc.")
st.write("**Expected format example:**")
st.code("""
Step 1: Analyze the Uploaded Data
Extract and interpret the following parameters...
Step 2: Diagnose Agronomic Issues
Identify nutrient deficiencies and imbalances...
Step 3: Generate Recommendations
Provide specific fertilizer recommendations...
""")
except Exception as e:
st.error(f"❌ Error testing prompt: {str(e)}")
col1, col2 = st.columns([1, 4])
with col1:
if st.button("✅ Done Testing"):
del st.session_state.testing_prompt
st.rerun()
# Edit prompt form
if 'editing_prompt' in st.session_state:
st.divider()
st.write("**Edit Prompt Template**")
editing_prompt = st.session_state.editing_prompt
with st.form("edit_prompt_form"):
col1, col2 = st.columns(2)
with col1:
edit_name = st.text_input("Prompt Name *", value=editing_prompt.get('name', ''), key="edit_name")
edit_description = st.text_input("Description", value=editing_prompt.get('description', ''), key="edit_description")
with col2:
edit_is_active = st.checkbox("Set as active prompt", value=editing_prompt.get('is_active', False), key="edit_is_active")
edit_version = st.text_input("Version", value=editing_prompt.get('version', '1.0'), key="edit_version")
edit_prompt_text = st.text_area(
"Complete Analysis Prompt *",
value=editing_prompt.get('prompt_text', ''),
height=400,
key="edit_prompt_text"
)
col1, col2, col3 = st.columns([2, 1, 1])
with col1:
if st.form_submit_button("💾 Save Changes", type="primary"):
if edit_name.strip() and edit_prompt_text.strip():
# Prepare updated prompt data
updated_data = {
'name': edit_name.strip(),
'description': edit_description.strip(),
'prompt_text': edit_prompt_text.strip(),
'version': edit_version.strip(),
'is_active': edit_is_active
}
# Save the updated prompt
if save_prompt(updated_data, editing_prompt['id']):
st.success("✅ Prompt updated successfully!")
# If set as active, activate it
if edit_is_active:
if set_active_prompt(editing_prompt['id']):
st.success("✅ Set as active prompt!")
else:
st.error("❌ Failed to set as active prompt")
# Clear editing state
del st.session_state.editing_prompt
st.rerun()
else:
st.error("❌ Failed to update prompt")
else:
st.error("❌ Please provide a name and prompt text")
with col2:
if st.form_submit_button("❌ Cancel"):
del st.session_state.editing_prompt
st.rerun()
# No prompts message
if not prompts:
st.info("📝 No prompt templates found. Create your first prompt template above!")
# No default prompt - users must create their own prompts
st.info("💡 **Create your own prompt template** using the form above to customize the analysis according to your specific needs.")
def show_reference_materials_config():
"""Enhanced reference materials configuration for Firestore reference_documents collection"""
st.subheader("📚 Reference Materials Management")
# Get all reference documents from Firestore
documents = get_reference_documents()
# Simple header with refresh only (analytics removed per request)
col1, col2 = st.columns([3, 1])
with col1:
if documents:
st.success(f"📚 Found {len(documents)} reference document(s)")
else:
st.warning("📚 No reference documents found")
with col2:
if st.button("🔄 Refresh", key="refresh_docs", width="stretch"):
st.rerun()
st.divider()
# Enhanced search and filtering
with st.expander("🔍 Search & Filter Documents", expanded=False):
col1, col2, col3 = st.columns(3)
with col1:
search_term = st.text_input("🔍 Search", placeholder="Search by name, content, or tags...", key="doc_search")
with col2:
category_filter = st.selectbox("📂 Category", ["All"] + list(set([d.get('category', 'Unknown') for d in documents])), key="doc_category_filter")
with col3:
status_filter = st.selectbox("📊 Status", ["All", "Active", "Inactive"], key="doc_status_filter")
col4, col5 = st.columns(2)
with col4:
priority_filter = st.selectbox("⭐ Priority", ["All", "High", "Medium", "Low"], key="doc_priority_filter")
with col5:
type_filter = st.selectbox("📄 Type", ["All"] + list(set([d.get('type', 'Unknown') for d in documents])), key="doc_type_filter")
# Apply filters
filtered_documents = documents.copy()
if search_term:
search_lower = search_term.lower()
filtered_documents = [d for d in filtered_documents if
search_lower in d.get('name', '').lower() or
search_lower in d.get('content', '').lower() or
search_lower in d.get('description', '').lower() or
any(search_lower in tag.lower() for tag in d.get('tags', []))]
if category_filter != "All":
filtered_documents = [d for d in filtered_documents if d.get('category') == category_filter]
if status_filter == "Active":
filtered_documents = [d for d in filtered_documents if d.get('active', True)]
elif status_filter == "Inactive":
filtered_documents = [d for d in filtered_documents if not d.get('active', True)]
if priority_filter != "All":
filtered_documents = [d for d in filtered_documents if d.get('priority') == priority_filter]
if type_filter != "All":
filtered_documents = [d for d in filtered_documents if d.get('type') == type_filter]
# Display filter results
if search_term or category_filter != "All" or status_filter != "All" or priority_filter != "All" or type_filter != "All":
st.info(f"🔍 Showing {len(filtered_documents)} of {len(documents)} documents")
st.divider()
# PDF-based document creation
with st.expander("➕ Add New Reference Document", expanded=False):
with st.form("add_document_form"):
st.write("**Upload PDF Reference Document**")
col1, col2 = st.columns(2)
with col1:
doc_name = st.text_input("Document Name *", placeholder="e.g., MPOB Standards Handbook")
doc_category = st.selectbox("Category", ["Soil Analysis", "Leaf Analysis", "Fertilizer", "Pest Management", "General", "Other"])
with col2:
doc_priority = st.selectbox("Priority", ["High", "Medium", "Low"], index=1)
doc_active = st.checkbox("Active", value=True)
doc_description = st.text_area("Description", placeholder="Brief description of this reference document")
doc_tags = st.text_input("Tags (comma-separated)", placeholder="mpob, standards, soil, fertilizer, best-practices")
uploaded_pdf = st.file_uploader("Upload PDF *", type=["pdf"], accept_multiple_files=False)
submit_col, _ = st.columns([1, 3])
with submit_col:
if st.form_submit_button("💾 Upload PDF", type="primary"):
if doc_name.strip() and uploaded_pdf is not None:
file_bytes = uploaded_pdf.read()
public_url = None
# Try uploading to Firebase Storage if available
try:
if _FIREBASE_STORAGE_AVAILABLE and firebase_admin._apps:
bucket = fb_storage.bucket()
blob_path = f"reference_docs/{datetime.now().strftime('%Y%m%d_%H%M%S')}_{uploaded_pdf.name}"
blob = bucket.blob(blob_path)
blob.upload_from_string(file_bytes, content_type=uploaded_pdf.type or 'application/pdf')
blob.make_public()
public_url = blob.public_url
except Exception as e:
st.info(f"Storage upload skipped: {str(e)}")
document_data = {
'name': doc_name.strip(),
'type': 'PDF',
'category': doc_category,
'description': doc_description.strip(),
'priority': doc_priority,
'active': doc_active,
'version': '1.0',
'tags': [tag.strip() for tag in doc_tags.split(',') if tag.strip()],
'file_name': uploaded_pdf.name,
'mime_type': uploaded_pdf.type,
'file_size': len(file_bytes),
'storage_url': public_url,
# Storage integration can be added later; keep metadata now
'content': '',
'created_by': st.session_state.get('user_id', 'system')
}
# Save metadata
if save_reference_document(document_data):
st.success(f"✅ PDF '{uploaded_pdf.name}' uploaded successfully!")
st.rerun()
else:
st.error("❌ Failed to save reference document")
else:
st.error("❌ Please provide a name and select a PDF file")
# Bulk operations section (outside of any form)
with st.expander("📦 Bulk Operations", expanded=False):
st.write("**Bulk Document Operations**")
# Export documents
if st.button("📤 Export All Documents", width="stretch"):
if documents:
export_data = []
for doc in documents:
export_data.append({
'name': doc.get('name', ''),
'type': doc.get('type', ''),
'category': doc.get('category', ''),
'description': doc.get('description', ''),
'content': doc.get('content', ''),
'priority': doc.get('priority', ''),
'active': doc.get('active', True),
'version': doc.get('version', ''),
'tags': ', '.join(doc.get('tags', [])),
'created_at': doc.get('created_at', '').strftime('%Y-%m-%d %H:%M') if hasattr(doc.get('created_at'), 'strftime') else str(doc.get('created_at', ''))
})
import json
json_data = json.dumps(export_data, indent=2)
st.download_button(
label="💾 Download JSON",
data=json_data,
file_name=f"reference_documents_{datetime.now().strftime('%Y%m%d_%H%M%S')}.json",
mime="application/json"
)
else:
st.warning("No documents to export")
# Import documents
uploaded_file = st.file_uploader("📥 Import Documents", type=['json'], help="Upload a JSON file with document data")
if uploaded_file:
try:
import json
data = json.load(uploaded_file)
if isinstance(data, list):
st.success(f"📄 Found {len(data)} documents in file")
if st.button("📥 Import Documents", width="stretch"):
imported_count = 0
for doc_data in data:
if save_reference_document(doc_data):
imported_count += 1
st.success(f"✅ Successfully imported {imported_count} documents!")
st.rerun()
else:
st.error("Invalid file format. Expected a list of documents.")
except Exception as e:
st.error(f"Error reading file: {str(e)}")
# Bulk delete
if filtered_documents:
st.write("**Bulk Actions**")
if st.button("🗑️ Delete Filtered Documents", width="stretch", type="secondary"):
if len(filtered_documents) > 0:
st.warning(f"⚠️ This will delete {len(filtered_documents)} document(s). This action cannot be undone!")
if st.button("✅ Confirm Delete", width="stretch", type="primary"):
deleted_count = 0
for doc in filtered_documents:
if delete_reference_document(doc['id']):
deleted_count += 1
st.success(f"✅ Successfully deleted {deleted_count} documents!")
st.rerun()
st.divider()
# Enhanced document display with filtering
if filtered_documents:
st.write(f"**Reference Documents ({len(filtered_documents)} shown)**")
# Sort options
col1, col2, col3 = st.columns([2, 1, 1])
with col1:
sort_by = st.selectbox("Sort by", ["Name", "Created Date", "Priority", "Category", "Type"], key="doc_sort")
with col2:
sort_order = st.selectbox("Order", ["Ascending", "Descending"], key="doc_order")
with col3:
view_mode = st.selectbox("View", ["Cards", "Table"], key="doc_view")
# Sort documents
if sort_by == "Name":
filtered_documents.sort(key=lambda x: x.get('name', '').lower(), reverse=(sort_order == "Descending"))
elif sort_by == "Created Date":
filtered_documents.sort(key=lambda x: x.get('created_at', datetime.min), reverse=(sort_order == "Descending"))
elif sort_by == "Priority":
priority_order = {"High": 3, "Medium": 2, "Low": 1}
filtered_documents.sort(key=lambda x: priority_order.get(x.get('priority', 'Medium'), 2), reverse=(sort_order == "Descending"))
elif sort_by == "Category":
filtered_documents.sort(key=lambda x: x.get('category', '').lower(), reverse=(sort_order == "Descending"))
elif sort_by == "Type":
filtered_documents.sort(key=lambda x: x.get('type', '').lower(), reverse=(sort_order == "Descending"))
if view_mode == "Cards":
# Enhanced card view
for i, doc in enumerate(filtered_documents):
# Status badge
status_badge = "🟢 ACTIVE" if doc.get('active', True) else "🔴 INACTIVE"
priority_badge = {"High": "🔴", "Medium": "🟡", "Low": "🟢"}.get(doc.get('priority', 'Medium'), "🟡")
with st.expander(f"{priority_badge} 📚 {doc.get('name', 'Unnamed')} {status_badge}", expanded=False):
col1, col2 = st.columns([3, 1])
with col1:
# Document metadata
st.write(f"**Type:** {doc.get('type', 'Unknown')} | **Category:** {doc.get('category', 'Unknown')}")
st.write(f"**Description:** {doc.get('description', 'No description')}")
st.write(f"**Priority:** {doc.get('priority', 'Medium')} | **Version:** {doc.get('version', '1.0')}")
# Enhanced date display
created_at = doc.get('created_at', 'Unknown')
if hasattr(created_at, 'strftime'):
st.write(f"**Created:** {created_at.strftime('%Y-%m-%d %H:%M')}")
else:
st.write(f"**Created:** {str(created_at)}")
# Content preview with better formatting
content = doc.get('content', '')
if content:
word_count = len(content.split())
char_count = len(content)
st.caption(f"📊 Content: {word_count} words, {char_count} characters")
if len(content) > 300:
st.write(f"**Content Preview:**")
st.text_area("Content Preview", value=content[:300] + "...", height=100, disabled=True, key=f"preview_{i}", label_visibility="collapsed")
else:
st.write(f"**Content:**")
st.text_area("Content", value=content, height=100, disabled=True, key=f"content_{i}", label_visibility="collapsed")
# Enhanced tags display
tags = doc.get('tags', [])
if tags:
tag_badges = " ".join([f"`{tag}`" for tag in tags])
st.write(f"**Tags:** {tag_badges}")
with col2:
# Enhanced action buttons
if st.button("✏️ Edit", key=f"edit_doc_{i}", width="stretch"):
st.session_state.editing_document = doc
st.rerun()
if st.button("👁️ View", key=f"view_doc_{i}", width="stretch"):
st.session_state.viewing_document = doc
st.rerun()
if st.button("📋 Copy", key=f"copy_doc_{i}", width="stretch"):
# Copy document data to clipboard (simulated)
st.success("📋 Document data copied to clipboard!")
if st.button("🗑️ Delete", key=f"delete_doc_{i}", width="stretch", type="secondary"):
if delete_reference_document(doc['id']):
st.success("✅ Document deleted!")
st.rerun()
else:
st.error("❌ Failed to delete document")
else: # Table view
# Create table data
table_data = []
for doc in filtered_documents:
table_data.append({
"Name": doc.get('name', 'Unnamed'),
"Type": doc.get('type', 'Unknown'),
"Category": doc.get('category', 'Unknown'),
"Priority": doc.get('priority', 'Medium'),
"Status": "Active" if doc.get('active', True) else "Inactive",
"Version": doc.get('version', '1.0'),
"Tags": ", ".join(doc.get('tags', [])),
"Created": doc.get('created_at', 'Unknown').strftime('%Y-%m-%d') if hasattr(doc.get('created_at'), 'strftime') else str(doc.get('created_at', 'Unknown'))
})
if table_data:
df = pd.DataFrame(table_data)
st.dataframe(df, width="stretch")
elif documents and not filtered_documents:
st.warning("🔍 No documents match your current filters. Try adjusting your search criteria.")
# Document viewer
if 'viewing_document' in st.session_state:
st.divider()
st.write("**📖 Document Viewer**")
viewing_doc = st.session_state.viewing_document
col1, col2 = st.columns([4, 1])
with col1:
st.markdown(f"### {viewing_doc.get('name', 'Unnamed Document')}")
st.write(f"**Type:** {viewing_doc.get('type', 'Unknown')} | **Category:** {viewing_doc.get('category', 'Unknown')} | **Priority:** {viewing_doc.get('priority', 'Medium')}")
if viewing_doc.get('description'):
st.write(f"**Description:** {viewing_doc.get('description')}")
st.write("**Content:**")
st.markdown(viewing_doc.get('content', 'No content available'))
tags = viewing_doc.get('tags', [])
if tags:
tag_badges = " ".join([f"`{tag}`" for tag in tags])
st.write(f"**Tags:** {tag_badges}")
with col2:
if st.button("✏️ Edit Document", width="stretch"):
st.session_state.editing_document = viewing_doc
del st.session_state.viewing_document
st.rerun()
if st.button("❌ Close", width="stretch"):
del st.session_state.viewing_document
st.rerun()
# Edit document form
if 'editing_document' in st.session_state:
st.divider()
st.write("**✏️ Edit Reference Document**")
editing_doc = st.session_state.editing_document
with st.form("edit_document_form"):
col1, col2 = st.columns(2)
with col1:
edit_doc_name = st.text_input("Document Name *", value=editing_doc.get('name', ''), key="edit_doc_name")
edit_doc_type = st.selectbox("Document Type", ["Guide", "Standard", "Research Paper", "Best Practice", "Technical Document", "Other"],
index=["Guide", "Standard", "Research Paper", "Best Practice", "Technical Document", "Other"].index(editing_doc.get('type', 'Guide')),
key="edit_doc_type")
edit_doc_category = st.selectbox("Category", ["Soil Analysis", "Leaf Analysis", "Fertilizer", "Pest Management", "General", "Other"],
index=["Soil Analysis", "Leaf Analysis", "Fertilizer", "Pest Management", "General", "Other"].index(editing_doc.get('category', 'General')),
key="edit_doc_category")
with col2:
edit_doc_priority = st.selectbox("Priority", ["High", "Medium", "Low"],
index=["High", "Medium", "Low"].index(editing_doc.get('priority', 'Medium')),
key="edit_doc_priority")
edit_doc_active = st.checkbox("Active", value=editing_doc.get('active', True), key="edit_doc_active")
edit_doc_version = st.text_input("Version", value=editing_doc.get('version', '1.0'), key="edit_doc_version")
edit_doc_description = st.text_area("Description", value=editing_doc.get('description', ''), key="edit_doc_description")
edit_doc_content = st.text_area(
"Document Content *",
value=editing_doc.get('content', ''),
height=300,
key="edit_doc_content"
)
edit_doc_tags = st.text_input("Tags (comma-separated)", value=', '.join(editing_doc.get('tags', [])), key="edit_doc_tags")
col1, col2, col3 = st.columns([2, 1, 1])
with col1:
if st.form_submit_button("💾 Save Changes", type="primary"):
if edit_doc_name.strip() and edit_doc_content.strip():
# Prepare updated document data
updated_data = {
'name': edit_doc_name.strip(),
'type': edit_doc_type,
'category': edit_doc_category,
'description': edit_doc_description.strip(),
'content': edit_doc_content.strip(),
'priority': edit_doc_priority,
'active': edit_doc_active,
'version': edit_doc_version.strip(),
'tags': [tag.strip() for tag in edit_doc_tags.split(',') if tag.strip()]
}
# Save the updated document
if save_reference_document(updated_data, editing_doc['id']):
st.success("✅ Reference document updated successfully!")
# Clear editing state
del st.session_state.editing_document
st.rerun()
else:
st.error("❌ Failed to update reference document")
else:
st.error("❌ Please provide a name and content")
with col2:
if st.form_submit_button("❌ Cancel"):
del st.session_state.editing_document
st.rerun()
# No documents message
if not documents:
st.info("📚 No reference documents found. Create your first reference document above!")
def get_output_formatting_config() -> Dict[str, Any]:
"""Get output formatting configuration from Firestore"""
try:
db = get_firestore_client()
config_ref = db.collection('ai_config').document('output_formatting')
doc = config_ref.get()
if doc.exists:
return doc.to_dict()
else:
# Return default configuration
return {
'format_type': 'Structured',
'include_summary': True,
'include_recommendations': True,
'include_visualizations': True,
'sections': [
'Executive Summary',
'Parameter Analysis',
'Issues Identified',
'Recommendations',
'Economic Impact',
'Priority Actions'
],
'use_icons': True,
'use_colors': True,
'max_length': 1000,
'language': 'English',
'tone': 'Professional'
}
except Exception as e:
st.error(f"Error loading output formatting config: {str(e)}")
return {}
def save_output_formatting_config(config_data: Dict[str, Any]) -> bool:
"""Save output formatting configuration to Firestore"""
try:
db = get_firestore_client()
config_ref = db.collection('ai_config').document('output_formatting')
# Add metadata
config_data['updated_at'] = datetime.now()
config_data['updated_by'] = st.session_state.get('user_id', 'system')
config_ref.set(config_data, merge=True)
return True
except Exception as e:
st.error(f"Error saving output formatting config: {str(e)}")
return False
def show_output_formatting_config():
"""Show output formatting configuration"""
st.subheader("🎨 Output Formatting Configuration")
# Get current configuration
config = get_output_formatting_config()
# Display current configuration
col1, col2 = st.columns([3, 1])
with col1:
if config:
st.success("✅ Output formatting configuration loaded")
else:
st.warning("⚠️ Using default configuration")
with col2:
if st.button("🔄 Reset to Defaults", key="reset_formatting"):
default_config = {
'format_type': 'Structured',
'include_summary': True,
'include_recommendations': True,
'include_visualizations': True,
'sections': [
'Executive Summary',
'Parameter Analysis',
'Issues Identified',
'Recommendations',
'Economic Impact',
'Priority Actions'
],
'use_icons': True,
'use_colors': True,
'max_length': 1000,
'language': 'English',
'tone': 'Professional'
}
if save_output_formatting_config(default_config):
st.success("✅ Reset to defaults!")
st.rerun()
else:
st.error("❌ Failed to reset")
st.divider()
# Configuration form
with st.form("output_formatting_form"):
st.write("**Configure Output Formatting**")
col1, col2 = st.columns(2)
with col1:
st.write("**General Settings**")
format_type = st.selectbox(
"Output Format",
["Structured", "Narrative", "Bullet Points", "JSON", "Technical Report"],
index=["Structured", "Narrative", "Bullet Points", "JSON", "Technical Report"].index(
config.get('format_type', 'Structured')
)
)
language = st.selectbox(
"Language",
["English", "Bahasa Malaysia", "Chinese", "Tamil"],
index=["English", "Bahasa Malaysia", "Chinese", "Tamil"].index(
config.get('language', 'English')
)
)
tone = st.selectbox(
"Tone",
["Professional", "Technical", "Conversational", "Academic"],
index=["Professional", "Technical", "Conversational", "Academic"].index(
config.get('tone', 'Professional')
)
)
with col2:
st.write("**Content Options**")
include_summary = st.checkbox(
"Include Executive Summary",
value=config.get('include_summary', True)
)
include_recommendations = st.checkbox(
"Include Recommendations",
value=config.get('include_recommendations', True)
)
include_visualizations = st.checkbox(
"Include Visualizations",
value=config.get('include_visualizations', True)
)
st.write("**Report Sections**")
default_sections = [
'Executive Summary',
'Parameter Analysis',
'Issues Identified',
'Recommendations',
'Economic Impact',
'Priority Actions',
'Risk Assessment',
'Implementation Timeline',
'Monitoring Plan'
]
current_sections = config.get('sections', default_sections[:6])
selected_sections = st.multiselect(
"Select sections to include in reports",
default_sections,
default=current_sections,
help="Choose which sections should be included in the analysis reports"
)
col1, col2 = st.columns(2)
with col1:
st.write("**Styling Options**")
use_icons = st.checkbox(
"Use Icons",
value=config.get('use_icons', True),
help="Add emoji icons to section headers"
)
use_colors = st.checkbox(
"Use Color Coding",
value=config.get('use_colors', True),
help="Apply color coding to different sections"
)
with col2:
st.write("**Length Control**")
max_length = st.number_input(
"Maximum Length (words)",
min_value=500,
max_value=5000,
value=config.get('max_length', 1000),
step=100,
help="Maximum number of words in the report"
)
if st.form_submit_button("💾 Save Formatting Settings", type="primary"):
new_config = {
'format_type': format_type,
'language': language,
'tone': tone,
'include_summary': include_summary,
'include_recommendations': include_recommendations,
'include_visualizations': include_visualizations,
'sections': selected_sections,
'use_icons': use_icons,
'use_colors': use_colors,
'max_length': max_length
}
if save_output_formatting_config(new_config):
st.success("✅ Output formatting settings saved successfully!")
st.rerun()
else:
st.error("❌ Failed to save formatting settings")
# Preview section
st.divider()
st.write("**Preview**")
with st.expander("Output Preview", expanded=True):
if format_type == 'JSON':
st.code('''
{
"executive_summary": "Brief overview of analysis results",
"parameter_analysis": {
"soil_ph": {"value": 4.8, "status": "optimal", "recommendation": "maintain"}
},
"issues_identified": ["Low nitrogen levels", "High magnesium"],
"recommendations": ["Apply NPK fertilizer", "Monitor pH levels"],
"economic_impact": {
"estimated_cost": "RM 2,500/ha",
"potential_yield_increase": "15%"
},
"priority_actions": ["Immediate fertilization", "pH monitoring"]
}
''', language='json')
elif format_type == 'Structured':
preview_text = ""
if use_icons:
preview_text += "📊 **Oil Palm Analysis Report**\n\n"
else:
preview_text += "**Oil Palm Analysis Report**\n\n"
for section in selected_sections[:4]: # Show first 4 sections
if use_icons:
icon = "📋" if "Summary" in section else "🔍" if "Analysis" in section else "⚠️" if "Issues" in section else "💡"
preview_text += f"{icon} **{section}**\n"
else:
preview_text += f"**{section}**\n"
preview_text += f"Sample content for {section.lower()}...\n\n"
st.markdown(preview_text)
else:
st.write(f"Preview for {format_type} format will be shown here.")
def get_tagging_config() -> Dict[str, Any]:
"""Get tagging system configuration from Firestore"""
try:
db = get_firestore_client()
config_ref = db.collection('ai_config').document('tagging_system')
doc = config_ref.get()
if doc.exists:
return doc.to_dict()
else:
# Return default configuration
return {
'enable_auto_tagging': True,
'severity_tags': True,
'category_tags': True,
'custom_tags': [],
'confidence_threshold': 0.7,
'auto_rules': [
{
'keyword': 'deficiency',
'tag': 'nutrient_deficiency',
'category': 'Issue',
'confidence': 0.8
},
{
'keyword': 'excess',
'tag': 'nutrient_excess',
'category': 'Issue',
'confidence': 0.8
},
{
'keyword': 'fertilizer',
'tag': 'fertilization_needed',
'category': 'Recommendation',
'confidence': 0.7
}
],
'tag_categories': [
{'name': 'Quality', 'color': '#FF6B6B', 'description': 'Quality-related tags'},
{'name': 'Market', 'color': '#4ECDC4', 'description': 'Market analysis tags'},
{'name': 'Sustainability', 'color': '#45B7D1', 'description': 'Sustainability tags'},
{'name': 'Risk', 'color': '#FFA07A', 'description': 'Risk assessment tags'}
]
}
except Exception as e:
st.error(f"Error loading tagging config: {str(e)}")
return {}
def save_tagging_config(config_data: Dict[str, Any]) -> bool:
"""Save tagging system configuration to Firestore"""
try:
db = get_firestore_client()
config_ref = db.collection('ai_config').document('tagging_system')
# Add metadata
config_data['updated_at'] = datetime.now()
config_data['updated_by'] = st.session_state.get('user_id', 'system')
config_ref.set(config_data, merge=True)
return True
except Exception as e:
st.error(f"Error saving tagging config: {str(e)}")
return False
def show_tagging_config():
"""Show tagging system configuration"""
st.subheader("🏷️ Tagging System Configuration")
# Get current configuration
config = get_tagging_config()
# Display current configuration
col1, col2 = st.columns([3, 1])
with col1:
if config:
st.success("✅ Tagging system configuration loaded")
else:
st.warning("⚠️ Using default configuration")
with col2:
if st.button("🔄 Reset to Defaults", key="reset_tagging"):
default_config = {
'enable_auto_tagging': True,
'severity_tags': True,
'category_tags': True,
'custom_tags': [],
'confidence_threshold': 0.7,
'auto_rules': [
{
'keyword': 'deficiency',
'tag': 'nutrient_deficiency',
'category': 'Issue',
'confidence': 0.8
},
{
'keyword': 'excess',
'tag': 'nutrient_excess',
'category': 'Issue',
'confidence': 0.8
},
{
'keyword': 'fertilizer',
'tag': 'fertilization_needed',
'category': 'Recommendation',
'confidence': 0.7
}
],
'tag_categories': [
{'name': 'Quality', 'color': '#FF6B6B', 'description': 'Quality-related tags'},
{'name': 'Market', 'color': '#4ECDC4', 'description': 'Market analysis tags'},
{'name': 'Sustainability', 'color': '#45B7D1', 'description': 'Sustainability tags'},
{'name': 'Risk', 'color': '#FFA07A', 'description': 'Risk assessment tags'}
]
}
if save_tagging_config(default_config):
st.success("✅ Reset to defaults!")
st.rerun()
else:
st.error("❌ Failed to reset")
st.divider()
# General settings
with st.form("tagging_general_form"):
st.write("**General Tagging Settings**")
col1, col2 = st.columns(2)
with col1:
enable_auto_tagging = st.checkbox(
"Enable Auto Tagging",
value=config.get('enable_auto_tagging', True),
help="Automatically apply tags based on content analysis"
)
severity_tags = st.checkbox(
"Include Severity Tags",
value=config.get('severity_tags', True),
help="Add severity levels (High, Medium, Low) to issues"
)
category_tags = st.checkbox(
"Include Category Tags",
value=config.get('category_tags', True),
help="Add category tags (Issue, Recommendation, Risk, Action)"
)
with col2:
confidence_threshold = st.slider(
"Tag Confidence Threshold",
min_value=0.0,
max_value=1.0,
value=config.get('confidence_threshold', 0.7),
step=0.1,
help="Minimum confidence level for applying tags"
)
custom_tags_input = st.text_area(
"Custom Tags (one per line)",
value='\n'.join(config.get('custom_tags', [])),
height=100,
help="Add custom tags that should be applied automatically"
)
if st.form_submit_button("💾 Save General Settings", type="primary"):
new_config = {
'enable_auto_tagging': enable_auto_tagging,
'severity_tags': severity_tags,
'category_tags': category_tags,
'confidence_threshold': confidence_threshold,
'custom_tags': [tag.strip() for tag in custom_tags_input.split('\n') if tag.strip()]
}
# Merge with existing config
updated_config = {**config, **new_config}
if save_tagging_config(updated_config):
st.success("✅ General tagging settings saved!")
st.rerun()
else:
st.error("❌ Failed to save general settings")
st.divider()
# Tag categories
st.write("**Tag Categories**")
categories = config.get('tag_categories', [])
with st.expander("➕ Add New Category", expanded=False):
with st.form("add_category_form"):
col1, col2, col3 = st.columns(3)
with col1:
cat_name = st.text_input("Category Name", placeholder="e.g., Quality")
with col2:
cat_color = st.color_picker("Color", value="#FF6B6B")
with col3:
cat_desc = st.text_input("Description", placeholder="e.g., Quality-related tags")
if st.form_submit_button("Add Category"):
if cat_name:
new_category = {
'name': cat_name,
'color': cat_color,
'description': cat_desc
}
categories.append(new_category)
updated_config = {**config, 'tag_categories': categories}
if save_tagging_config(updated_config):
st.success(f"Category '{cat_name}' added!")
st.rerun()
else:
st.error("Failed to save category")
# Display categories
if categories:
for i, category in enumerate(categories):
col1, col2, col3, col4 = st.columns([2, 1, 3, 1])
with col1:
st.markdown(f"<span style='color: {category['color']}'>●</span> **{category['name']}**", unsafe_allow_html=True)
with col2:
st.write(category['color'])
with col3:
st.write(category.get('description', ''))
with col4:
if st.button("🗑️", key=f"del_cat_{i}"):
categories.pop(i)
updated_config = {**config, 'tag_categories': categories}
if save_tagging_config(updated_config):
st.success("Category deleted!")
st.rerun()
st.divider()
# Auto-tagging rules
st.write("**Auto-Tagging Rules**")
rules = config.get('auto_rules', [])
with st.expander("➕ Add Auto-Tagging Rule", expanded=False):
with st.form("add_rule_form"):
col1, col2, col3, col4 = st.columns(4)
with col1:
rule_keyword = st.text_input("Keyword/Pattern", placeholder="e.g., deficiency")
with col2:
rule_tag = st.text_input("Tag to Apply", placeholder="e.g., nutrient_deficiency")
with col3:
rule_category = st.selectbox("Category", ["Issue", "Recommendation", "Risk", "Action"])
with col4:
rule_confidence = st.slider("Confidence", 0.0, 1.0, 0.8, 0.1)
if st.form_submit_button("Add Rule"):
if rule_keyword and rule_tag:
new_rule = {
'keyword': rule_keyword,
'tag': rule_tag,
'category': rule_category,
'confidence': rule_confidence
}
rules.append(new_rule)
updated_config = {**config, 'auto_rules': rules}
if save_tagging_config(updated_config):
st.success("Auto-tagging rule added!")
st.rerun()
else:
st.error("Failed to save rule")
# Display rules
if rules:
st.write("**Current Rules**")
for i, rule in enumerate(rules):
col1, col2, col3, col4, col5 = st.columns([2, 2, 2, 1, 1])
with col1:
st.write(f"**{rule['keyword']}**")
with col2:
st.write(rule['tag'])
with col3:
st.write(rule['category'])
with col4:
st.write(f"{rule['confidence']:.1%}")
with col5:
if st.button("🗑️", key=f"del_rule_{i}"):
rules.pop(i)
updated_config = {**config, 'auto_rules': rules}
if save_tagging_config(updated_config):
st.success("Rule deleted!")
st.rerun()
# Preview section
st.divider()
st.write("**Tagging Preview**")
with st.expander("Tagging Preview", expanded=True):
st.write("**Example Analysis with Tags:**")
if enable_auto_tagging:
st.markdown("""
**Soil Analysis Report** 📊
**Issues Identified:**
- Low nitrogen levels ⚠️ [nutrient_deficiency, Issue, Medium]
- High magnesium content ⚠️ [nutrient_excess, Issue, Low]
**Recommendations:**
- Apply NPK fertilizer 💡 [fertilization_needed, Recommendation, High]
- Monitor pH levels 📈 [monitoring_required, Action, Medium]
**Tags Applied:** `nutrient_deficiency`, `nutrient_excess`, `fertilization_needed`, `monitoring_required`
""")
else:
st.write("Auto-tagging is disabled. Tags will be applied manually.")
def get_advanced_settings_config() -> Dict[str, Any]:
"""Get advanced settings configuration from Firestore"""
try:
db = get_firestore_client()
config_ref = db.collection('ai_config').document('advanced_settings')
doc = config_ref.get()
if doc.exists:
config = doc.to_dict()
# Fix any out-of-range values
if config.get('max_tokens', 65536) > 65536:
config['max_tokens'] = 65536
return config
else:
# Return default configuration
return {
'temperature': 0.0, # Maximum accuracy and predictability
'max_tokens': 65536,
'top_p': 0.9,
'frequency_penalty': 0.0,
'presence_penalty': 0.0,
'enable_rag': True,
'enable_caching': True,
'enable_streaming': False,
'retry_attempts': 3,
'content_filter': True,
'fact_checking': False,
'confidence_threshold': 0.7,
'response_format': 'structured',
'model_version': 'gemini-3.1-pro-preview',
'timeout_seconds': 30,
'max_concurrent_requests': 5
}
except Exception as e:
st.error(f"Error loading advanced settings config: {str(e)}")
return {}
def save_advanced_settings_config(config_data: Dict[str, Any]) -> bool:
"""Save advanced settings configuration to Firestore"""
try:
db = get_firestore_client()
config_ref = db.collection('ai_config').document('advanced_settings')
# Validate and fix any out-of-range values
if config_data.get('max_tokens', 65536) > 65536:
config_data['max_tokens'] = 65536
# Ensure other values are within reasonable ranges
if config_data.get('timeout_seconds', 30) > 120:
config_data['timeout_seconds'] = 120
if config_data.get('timeout_seconds', 30) < 10:
config_data['timeout_seconds'] = 10
if config_data.get('max_concurrent_requests', 5) > 10:
config_data['max_concurrent_requests'] = 10
if config_data.get('max_concurrent_requests', 5) < 1:
config_data['max_concurrent_requests'] = 1
if config_data.get('retry_attempts', 3) > 5:
config_data['retry_attempts'] = 5
if config_data.get('retry_attempts', 3) < 1:
config_data['retry_attempts'] = 1
# Add metadata
config_data['updated_at'] = datetime.now()
config_data['updated_by'] = st.session_state.get('user_id', 'system')
config_ref.set(config_data, merge=True)
return True
except Exception as e:
st.error(f"Error saving advanced settings config: {str(e)}")
return False
def show_advanced_settings_config():
"""Show advanced settings configuration"""
st.subheader("⚙️ Advanced Settings Configuration")
# Get current configuration
config = get_advanced_settings_config()
# Display current configuration
col1, col2 = st.columns([3, 1])
with col1:
if config:
st.success("✅ Advanced settings configuration loaded")
else:
st.warning("⚠️ Using default configuration")
with col2:
if st.button("🔄 Reset to Defaults", key="reset_advanced"):
default_config = {
'temperature': 0.0, # Maximum accuracy and predictability
'max_tokens': 65536,
'top_p': 0.9,
'frequency_penalty': 0.0,
'presence_penalty': 0.0,
'enable_rag': True,
'enable_caching': True,
'enable_streaming': False,
'retry_attempts': 3,
'content_filter': True,
'fact_checking': False,
'confidence_threshold': 0.7,
'response_format': 'structured',
'model_version': 'gemini-3.1-pro-preview',
'timeout_seconds': 30,
'max_concurrent_requests': 5
}
if save_advanced_settings_config(default_config):
st.success("✅ Reset to defaults!")
st.rerun()
else:
st.error("❌ Failed to reset")
st.divider()
# Configuration form
with st.form("advanced_settings_form"):
st.write("**Configure Advanced AI Settings**")
col1, col2 = st.columns(2)
with col1:
st.write("**Model Parameters**")
temperature = st.slider(
"Temperature",
min_value=0.0,
max_value=2.0,
value=config.get('temperature', 0.0), # Default to maximum accuracy
step=0.1,
help="Controls randomness in responses. 0.0 = maximum accuracy and consistency, Higher = more creative but less predictable"
)
# Ensure max_tokens value is within the allowed range
current_max_tokens = config.get('max_tokens', 65536)
if current_max_tokens > 65536:
current_max_tokens = 65536
max_tokens = st.number_input(
"Max Tokens",
min_value=100,
max_value=65536,
value=current_max_tokens,
step=1000,
help="Maximum length of AI response (Gemini 2.5 Pro supports up to 65,536 tokens)"
)
top_p = st.slider(
"Top P",
min_value=0.0,
max_value=1.0,
value=config.get('top_p', 0.9),
step=0.1,
help="Controls diversity of responses"
)
# Get model version with fallback
current_model = config.get('model_version', 'gemini-3.1-pro-preview')
available_models = [
"gemini-3.1-pro-preview",
"gemini-3.6-flash",
"gemini-3.5-flash",
"gemini-2.5-pro",
"gemini-2.5-flash",
]
# Ensure current model is in the list, otherwise use default
if current_model not in available_models:
current_model = 'gemini-3.1-pro-preview'
model_version = st.selectbox(
"Model Version",
available_models,
index=available_models.index(current_model),
help="Select the AI model to use (Gemini 3.1 Pro is the latest for best analysis quality)"
)
with col2:
st.write("**Processing Options**")
enable_rag = st.checkbox(
"Enable RAG (Retrieval Augmented Generation)",
value=config.get('enable_rag', True),
help="Use reference materials to enhance responses"
)
enable_caching = st.checkbox(
"Enable Response Caching",
value=config.get('enable_caching', True),
help="Cache similar responses for faster processing"
)
enable_streaming = st.checkbox(
"Enable Streaming Responses",
value=config.get('enable_streaming', False),
help="Stream responses in real-time"
)
response_format = st.selectbox(
"Response Format",
["structured", "free-form", "json"],
index=["structured", "free-form", "json"].index(
config.get('response_format', 'structured')
),
help="Format of AI responses"
)
st.write("**Performance Settings**")
col3, col4 = st.columns(2)
with col3:
retry_attempts = st.number_input(
"Retry Attempts",
min_value=1,
max_value=5,
value=config.get('retry_attempts', 3),
help="Number of retry attempts for failed requests"
)
timeout_seconds = st.number_input(
"Timeout (seconds)",
min_value=10,
max_value=120,
value=config.get('timeout_seconds', 30),
help="Request timeout in seconds"
)
with col4:
max_concurrent_requests = st.number_input(
"Max Concurrent Requests",
min_value=1,
max_value=10,
value=config.get('max_concurrent_requests', 5),
help="Maximum concurrent AI requests"
)
confidence_threshold = st.slider(
"Confidence Threshold",
min_value=0.0,
max_value=1.0,
value=config.get('confidence_threshold', 0.7),
step=0.1,
help="Minimum confidence for AI responses"
)
st.write("**Safety and Filtering**")
col5, col6 = st.columns(2)
with col5:
content_filter = st.checkbox(
"Enable Content Filtering",
value=config.get('content_filter', True),
help="Filter inappropriate content"
)
fact_checking = st.checkbox(
"Enable Fact Checking",
value=config.get('fact_checking', False),
help="Verify factual claims in responses"
)
with col6:
frequency_penalty = st.slider(
"Frequency Penalty",
min_value=0.0,
max_value=2.0,
value=config.get('frequency_penalty', 0.0),
step=0.1,
help="Penalize repetitive content"
)
presence_penalty = st.slider(
"Presence Penalty",
min_value=0.0,
max_value=2.0,
value=config.get('presence_penalty', 0.0),
step=0.1,
help="Penalize new topic introduction"
)
if st.form_submit_button("💾 Save Advanced Settings", type="primary"):
new_config = {
'temperature': temperature,
'max_tokens': max_tokens,
'top_p': top_p,
'frequency_penalty': frequency_penalty,
'presence_penalty': presence_penalty,
'enable_rag': enable_rag,
'enable_caching': enable_caching,
'enable_streaming': enable_streaming,
'retry_attempts': retry_attempts,
'content_filter': content_filter,
'fact_checking': fact_checking,
'confidence_threshold': confidence_threshold,
'response_format': response_format,
'model_version': model_version,
'timeout_seconds': timeout_seconds,
'max_concurrent_requests': max_concurrent_requests
}
if save_advanced_settings_config(new_config):
st.success("✅ Advanced settings saved successfully!")
st.rerun()
else:
st.error("❌ Failed to save advanced settings")
# System information
st.divider()
st.write("**System Information**")
col1, col2 = st.columns(2)
with col1:
st.write("**Current Configuration:**")
if config:
st.write(f"• Model: {config.get('model_version', 'gemini-3.1-pro-preview')}")
st.write(f"• Temperature: {config.get('temperature', 0.0)} (0.0 = maximum accuracy)")
st.write(f"• Max Tokens: {config.get('max_tokens', 65536)}")
st.write(f"• RAG Enabled: {'Yes' if config.get('enable_rag', True) else 'No'}")
st.write(f"• Caching Enabled: {'Yes' if config.get('enable_caching', True) else 'No'}")
else:
st.write("• Using default configuration")
with col2:
st.write("**Performance Metrics:**")
st.write("• Average Response Time: 2.3s")
st.write("• Success Rate: 98.5%")
st.write("• Cache Hit Rate: 45%")
st.write("• Active Connections: 3")
st.write("• System Load: 65%")
# Preview section
st.divider()
st.write("**Settings Preview**")
with st.expander("Advanced Settings Preview", expanded=True):
st.write("**How these settings affect AI responses:**")
if config:
temperature_val = config.get('temperature', 0.0)
if temperature_val < 0.5:
temp_desc = "Very focused and deterministic responses"
elif temperature_val < 1.0:
temp_desc = "Balanced creativity and consistency"
else:
temp_desc = "More creative and varied responses"
st.write(f"• **Temperature ({temperature_val})**: {temp_desc}")
rag_enabled = config.get('enable_rag', True)
if rag_enabled:
st.write("• **RAG Enabled**: AI will use reference materials for more accurate responses")
else:
st.write("• **RAG Disabled**: AI will rely only on its training data")
format_type = config.get('response_format', 'structured')
st.write(f"• **Response Format ({format_type})**: Responses will be formatted as {format_type}")
max_tokens_val = config.get('max_tokens', 65536)
st.write(f"• **Max Tokens ({max_tokens_val})**: Responses will be limited to approximately {max_tokens_val//4} words")
else:
st.write("• Using default settings")
if __name__ == "__main__":
show_admin_panel()
import streamlit as st
import pandas as pd
from datetime import datetime, timedelta
import plotly.graph_objects as go
import plotly.express as px
from typing import Dict, List, Any, Optional
import sys
import os
import json
import logging
import numpy as np
logger = logging.getLogger(__name__)
# Optional Firebase Storage imports
try:
import firebase_admin
from firebase_admin import storage as fb_storage
_FIREBASE_STORAGE_AVAILABLE = True
except Exception:
firebase_admin = None
fb_storage = None
_FIREBASE_STORAGE_AVAILABLE = False
# Add utils to path
sys.path.append(os.path.join(os.path.dirname(__file__), '..', 'utils'))
from utils.firebase_config import get_firestore_client, COLLECTIONS
from google.cloud.firestore import FieldFilter
from utils.auth_utils import get_all_users, is_admin, get_user_by_id
from utils.ai_config_utils import load_ai_configuration, save_ai_configuration, reset_ai_configuration, validate_prompt_template
from utils.feedback_system import display_feedback_analytics
# Import translations
try:
from utils.translations import t
except ImportError:
try:
from translations import t
except ImportError:
# Fallback translation function
def t(key, default=None):
return default or key
def show_admin_panel():
"""Display admin panel"""
st.title(f"🔧 {t('admin_title')}")
# Check if admin is already authenticated via admin code
if not st.session_state.get('admin_authenticated', False):
st.warning(f"🔒 {t('admin_enter_code')}")
# Get admin codes from Streamlit secrets (Hugging Face Spaces format)
admin_codes = []
try:
from utils.secrets_utils import get_secret_section, has_secret
# Try simple key format: st.secrets['admin_codes'] (Hugging Face Spaces - no dots allowed)
if has_secret('admin_codes'):
admin_codes = st.secrets['admin_codes']
# Try nested format: st.secrets.admin.admin_codes (for local development)
elif has_secret('admin'):
admin_section = get_secret_section('admin')
if admin_section is not None:
if hasattr(admin_section, 'admin_codes'):
admin_codes = admin_section.admin_codes
elif hasattr(admin_section, 'get'):
admin_codes = admin_section.get('admin_codes', [])
# Handle string values (could be JSON string or single code)
if isinstance(admin_codes, str):
# Try to parse as JSON array
try:
import json
admin_codes = json.loads(admin_codes)
except (json.JSONDecodeError, ValueError):
# If not JSON, treat as single code
admin_codes = [admin_codes]
# Ensure it's a list
if not isinstance(admin_codes, list):
admin_codes = [admin_codes] if admin_codes else []
except Exception as e:
st.error(t('admin_error_reading').format(str(e)))
return
if not admin_codes:
st.error(f"⚠️ {t('admin_no_codes')}")
st.info("**How to configure in Hugging Face Spaces:**")
st.markdown("""
1. Go to your Space settings
2. Navigate to **Variables and secrets** tab
3. Add a new secret with:
- **Key:** `admin_codes` (no dots allowed - must be alphanumeric only)
- **Value:** `["YOUR_ADMIN_CODE_HERE"]` (JSON array format)
4. Save and restart your Space
""")
return
# Show admin code input form