Phase 3: Add token analytics reports (adapted for AIModelConfig)
This commit is contained in:
@@ -251,3 +251,371 @@ def data_quality_report(request):
|
||||
context.update(admin_context)
|
||||
|
||||
return render(request, 'admin/reports/data_quality.html', context)
|
||||
|
||||
|
||||
@staff_member_required
|
||||
def token_usage_report(request):
|
||||
"""Comprehensive token usage analytics with multi-dimensional insights"""
|
||||
from igny8_core.business.billing.models import CreditUsageLog, AIModelConfig
|
||||
from igny8_core.auth.models import Account
|
||||
from decimal import Decimal
|
||||
|
||||
# Date filter setup
|
||||
days_filter = request.GET.get('days', '30')
|
||||
try:
|
||||
days = int(days_filter)
|
||||
except ValueError:
|
||||
days = 30
|
||||
|
||||
start_date = timezone.now() - timedelta(days=days)
|
||||
|
||||
# Base queryset - include all records (tokens may be 0 for historical data)
|
||||
logs = CreditUsageLog.objects.filter(
|
||||
created_at__gte=start_date
|
||||
)
|
||||
|
||||
# Total statistics
|
||||
total_tokens_input = logs.aggregate(total=Sum('tokens_input'))['total'] or 0
|
||||
total_tokens_output = logs.aggregate(total=Sum('tokens_output'))['total'] or 0
|
||||
total_tokens = total_tokens_input + total_tokens_output
|
||||
total_calls = logs.count()
|
||||
avg_tokens_per_call = total_tokens / total_calls if total_calls > 0 else 0
|
||||
|
||||
# Token usage by model (using model_config FK)
|
||||
token_by_model = logs.filter(model_config__isnull=False).values(
|
||||
'model_config__model_name',
|
||||
'model_config__display_name'
|
||||
).annotate(
|
||||
total_tokens_input=Sum('tokens_input'),
|
||||
total_tokens_output=Sum('tokens_output'),
|
||||
call_count=Count('id'),
|
||||
total_cost_input=Sum('cost_usd_input'),
|
||||
total_cost_output=Sum('cost_usd_output')
|
||||
).order_by('-total_tokens_input')[:10]
|
||||
|
||||
# Add total_tokens and total_cost to each model
|
||||
for model in token_by_model:
|
||||
model['total_tokens'] = (model['total_tokens_input'] or 0) + (model['total_tokens_output'] or 0)
|
||||
model['total_cost'] = (model['total_cost_input'] or 0) + (model['total_cost_output'] or 0)
|
||||
model['avg_tokens'] = model['total_tokens'] / model['call_count'] if model['call_count'] > 0 else 0
|
||||
model['model'] = model['model_config__display_name'] or model['model_config__model_name']
|
||||
token_by_model = sorted(token_by_model, key=lambda x: x['total_tokens'], reverse=True)
|
||||
|
||||
# Token usage by function/operation
|
||||
token_by_function = logs.values('operation_type').annotate(
|
||||
total_tokens_input=Sum('tokens_input'),
|
||||
total_tokens_output=Sum('tokens_output'),
|
||||
call_count=Count('id'),
|
||||
total_cost=Sum('cost_usd_total')
|
||||
).order_by('-total_tokens_input')[:10]
|
||||
|
||||
# Add total_tokens to each function
|
||||
for func in token_by_function:
|
||||
func['total_tokens'] = (func['total_tokens_input'] or 0) + (func['total_tokens_output'] or 0)
|
||||
func['avg_tokens'] = func['total_tokens'] / func['call_count'] if func['call_count'] > 0 else 0
|
||||
func['function'] = func['operation_type'].replace('_', ' ').title() if func['operation_type'] else 'Unknown'
|
||||
token_by_function = sorted(token_by_function, key=lambda x: x['total_tokens'], reverse=True)
|
||||
|
||||
# Token usage by account (top consumers)
|
||||
token_by_account = logs.values('account__name', 'account_id').annotate(
|
||||
total_tokens_input=Sum('tokens_input'),
|
||||
total_tokens_output=Sum('tokens_output'),
|
||||
call_count=Count('id'),
|
||||
total_cost=Sum('cost_usd_total')
|
||||
).order_by('-total_tokens_input')[:15]
|
||||
|
||||
# Add total_tokens to each account
|
||||
for account in token_by_account:
|
||||
account['total_tokens'] = (account['total_tokens_input'] or 0) + (account['total_tokens_output'] or 0)
|
||||
token_by_account = sorted(token_by_account, key=lambda x: x['total_tokens'], reverse=True)[:15]
|
||||
|
||||
# Daily token trends (time series)
|
||||
daily_data = []
|
||||
daily_labels = []
|
||||
for i in range(days):
|
||||
day = timezone.now().date() - timedelta(days=days-i-1)
|
||||
day_logs = logs.filter(created_at__date=day)
|
||||
day_tokens_input = day_logs.aggregate(total=Sum('tokens_input'))['total'] or 0
|
||||
day_tokens_output = day_logs.aggregate(total=Sum('tokens_output'))['total'] or 0
|
||||
day_tokens = day_tokens_input + day_tokens_output
|
||||
daily_labels.append(day.strftime('%m/%d'))
|
||||
daily_data.append(int(day_tokens))
|
||||
|
||||
# Token efficiency metrics
|
||||
success_rate = 100.0
|
||||
successful_tokens = total_tokens
|
||||
wasted_tokens = 0
|
||||
|
||||
tokens_by_status = [{
|
||||
'error': None,
|
||||
'total_tokens': total_tokens,
|
||||
'call_count': total_calls,
|
||||
'avg_tokens': avg_tokens_per_call
|
||||
}]
|
||||
|
||||
# Peak usage times (hour of day)
|
||||
hourly_usage = logs.extra(
|
||||
select={'hour': "EXTRACT(hour FROM created_at)"}
|
||||
).values('hour').annotate(
|
||||
token_input=Sum('tokens_input'),
|
||||
token_output=Sum('tokens_output'),
|
||||
call_count=Count('id')
|
||||
).order_by('hour')
|
||||
|
||||
# Add total token_count for each hour
|
||||
for hour_data in hourly_usage:
|
||||
hour_data['token_count'] = (hour_data['token_input'] or 0) + (hour_data['token_output'] or 0)
|
||||
|
||||
# Cost efficiency
|
||||
total_cost = logs.aggregate(total=Sum('cost_usd_total'))['total'] or Decimal('0.00')
|
||||
cost_per_1k_tokens = float(total_cost) / (total_tokens / 1000) if total_tokens > 0 else 0.0
|
||||
|
||||
context = {
|
||||
'title': 'Token Usage Report',
|
||||
'days_filter': days,
|
||||
'total_tokens': int(total_tokens),
|
||||
'total_calls': total_calls,
|
||||
'avg_tokens_per_call': round(avg_tokens_per_call, 2),
|
||||
'token_by_model': list(token_by_model),
|
||||
'token_by_function': list(token_by_function),
|
||||
'token_by_account': list(token_by_account),
|
||||
'daily_labels': json.dumps(daily_labels),
|
||||
'daily_data': json.dumps(daily_data),
|
||||
'tokens_by_status': list(tokens_by_status),
|
||||
'success_rate': round(success_rate, 2),
|
||||
'successful_tokens': int(successful_tokens),
|
||||
'wasted_tokens': int(wasted_tokens),
|
||||
'hourly_usage': list(hourly_usage),
|
||||
'total_cost': float(total_cost),
|
||||
'cost_per_1k_tokens': float(cost_per_1k_tokens),
|
||||
'current_app': '_reports',
|
||||
}
|
||||
|
||||
# Merge with admin context
|
||||
from igny8_core.admin.site import admin_site
|
||||
admin_context = admin_site.each_context(request)
|
||||
context.update(admin_context)
|
||||
|
||||
return render(request, 'admin/reports/token_usage.html', context)
|
||||
|
||||
|
||||
@staff_member_required
|
||||
def ai_cost_analysis(request):
|
||||
"""Multi-dimensional AI cost analysis with model pricing, trends, and predictions"""
|
||||
from igny8_core.business.billing.models import CreditUsageLog, AIModelConfig
|
||||
from igny8_core.auth.models import Account
|
||||
from decimal import Decimal
|
||||
|
||||
# Date filter setup
|
||||
days_filter = request.GET.get('days', '30')
|
||||
try:
|
||||
days = int(days_filter)
|
||||
except ValueError:
|
||||
days = 30
|
||||
|
||||
start_date = timezone.now() - timedelta(days=days)
|
||||
|
||||
# Base queryset - filter for records with cost data
|
||||
logs = CreditUsageLog.objects.filter(
|
||||
created_at__gte=start_date,
|
||||
cost_usd_total__isnull=False
|
||||
)
|
||||
|
||||
# Overall cost metrics
|
||||
total_cost = logs.aggregate(total=Sum('cost_usd_total'))['total'] or Decimal('0.00')
|
||||
total_calls = logs.count()
|
||||
avg_cost_per_call = logs.aggregate(avg=Avg('cost_usd_total'))['avg'] or Decimal('0.00')
|
||||
total_tokens_input = logs.aggregate(total=Sum('tokens_input'))['total'] or 0
|
||||
total_tokens_output = logs.aggregate(total=Sum('tokens_output'))['total'] or 0
|
||||
total_tokens = total_tokens_input + total_tokens_output
|
||||
|
||||
# Revenue & Margin calculation
|
||||
total_credits_charged = logs.aggregate(total=Sum('credits_used'))['total'] or 0
|
||||
# Average credit price (simplified - in reality would vary by plan)
|
||||
avg_credit_price = Decimal('0.01') # $0.01 per credit default
|
||||
total_revenue = Decimal(total_credits_charged) * avg_credit_price
|
||||
total_margin = total_revenue - total_cost
|
||||
margin_percentage = float((total_margin / total_revenue * 100) if total_revenue > 0 else 0)
|
||||
|
||||
# Per-unit margins
|
||||
margin_per_1m_tokens = float(total_margin) / (total_tokens / 1_000_000) if total_tokens > 0 else 0
|
||||
margin_per_1k_credits = float(total_margin) / (total_credits_charged / 1000) if total_credits_charged > 0 else 0
|
||||
|
||||
# Cost by model with efficiency metrics (using model_config FK)
|
||||
cost_by_model = logs.filter(model_config__isnull=False).values(
|
||||
'model_config__model_name',
|
||||
'model_config__display_name'
|
||||
).annotate(
|
||||
total_cost=Sum('cost_usd_total'),
|
||||
call_count=Count('id'),
|
||||
avg_cost=Avg('cost_usd_total'),
|
||||
total_tokens_input=Sum('tokens_input'),
|
||||
total_tokens_output=Sum('tokens_output'),
|
||||
total_credits=Sum('credits_used')
|
||||
).order_by('-total_cost')
|
||||
|
||||
# Add cost efficiency and margin for each model
|
||||
for model in cost_by_model:
|
||||
model['total_tokens'] = (model['total_tokens_input'] or 0) + (model['total_tokens_output'] or 0)
|
||||
model['avg_tokens'] = model['total_tokens'] / model['call_count'] if model['call_count'] > 0 else 0
|
||||
model['model'] = model['model_config__display_name'] or model['model_config__model_name']
|
||||
|
||||
if model['total_tokens'] and model['total_tokens'] > 0:
|
||||
model['cost_per_1k_tokens'] = float(model['total_cost']) / (model['total_tokens'] / 1000)
|
||||
else:
|
||||
model['cost_per_1k_tokens'] = 0
|
||||
|
||||
# Calculate margin for this model
|
||||
model_revenue = Decimal(model['total_credits'] or 0) * avg_credit_price
|
||||
model_margin = model_revenue - model['total_cost']
|
||||
model['revenue'] = float(model_revenue)
|
||||
model['margin'] = float(model_margin)
|
||||
model['margin_percentage'] = float((model_margin / model_revenue * 100) if model_revenue > 0 else 0)
|
||||
|
||||
# Cost by account (top spenders)
|
||||
cost_by_account = logs.values('account__name', 'account_id').annotate(
|
||||
total_cost=Sum('cost_usd_total'),
|
||||
call_count=Count('id'),
|
||||
total_tokens_input=Sum('tokens_input'),
|
||||
total_tokens_output=Sum('tokens_output'),
|
||||
avg_cost=Avg('cost_usd_total')
|
||||
).order_by('-total_cost')[:15]
|
||||
|
||||
# Add total_tokens to each account
|
||||
for account in cost_by_account:
|
||||
account['total_tokens'] = (account['total_tokens_input'] or 0) + (account['total_tokens_output'] or 0)
|
||||
|
||||
# Cost by function/operation
|
||||
cost_by_function = logs.values('operation_type').annotate(
|
||||
total_cost=Sum('cost_usd_total'),
|
||||
call_count=Count('id'),
|
||||
avg_cost=Avg('cost_usd_total'),
|
||||
total_tokens_input=Sum('tokens_input'),
|
||||
total_tokens_output=Sum('tokens_output'),
|
||||
total_credits=Sum('credits_used')
|
||||
).order_by('-total_cost')[:10]
|
||||
|
||||
# Add total_tokens, function alias, and margin
|
||||
for func in cost_by_function:
|
||||
func['total_tokens'] = (func['total_tokens_input'] or 0) + (func['total_tokens_output'] or 0)
|
||||
func['function'] = func['operation_type'].replace('_', ' ').title() if func['operation_type'] else 'Unknown'
|
||||
|
||||
# Calculate margin for this operation
|
||||
func_revenue = Decimal(func['total_credits'] or 0) * avg_credit_price
|
||||
func_margin = func_revenue - func['total_cost']
|
||||
func['revenue'] = float(func_revenue)
|
||||
func['margin'] = float(func_margin)
|
||||
func['margin_percentage'] = float((func_margin / func_revenue * 100) if func_revenue > 0 else 0)
|
||||
|
||||
# Daily cost trends (time series)
|
||||
daily_cost_data = []
|
||||
daily_cost_labels = []
|
||||
daily_call_data = []
|
||||
|
||||
for i in range(days):
|
||||
day = timezone.now().date() - timedelta(days=days-i-1)
|
||||
day_logs = logs.filter(created_at__date=day)
|
||||
day_cost = day_logs.aggregate(total=Sum('cost_usd_total'))['total'] or Decimal('0.00')
|
||||
day_calls = day_logs.count()
|
||||
|
||||
daily_cost_labels.append(day.strftime('%m/%d'))
|
||||
daily_cost_data.append(float(day_cost))
|
||||
daily_call_data.append(day_calls)
|
||||
|
||||
# Cost prediction (simple linear extrapolation)
|
||||
if len(daily_cost_data) > 7:
|
||||
recent_avg_daily = sum(daily_cost_data[-7:]) / 7
|
||||
projected_monthly = recent_avg_daily * 30
|
||||
else:
|
||||
projected_monthly = 0
|
||||
|
||||
# Cost anomalies (calls costing > 3x average)
|
||||
failed_cost = Decimal('0.00')
|
||||
if avg_cost_per_call > 0:
|
||||
anomaly_threshold = float(avg_cost_per_call) * 3
|
||||
anomalies = logs.filter(cost_usd_total__gt=anomaly_threshold).select_related('model_config').values(
|
||||
'model_config__model_name',
|
||||
'model_config__display_name',
|
||||
'operation_type',
|
||||
'account__name',
|
||||
'cost_usd_total',
|
||||
'tokens_input',
|
||||
'tokens_output',
|
||||
'created_at'
|
||||
).order_by('-cost_usd_total')[:10]
|
||||
|
||||
# Add aliases for template
|
||||
for anomaly in anomalies:
|
||||
anomaly['model'] = anomaly['model_config__display_name'] or anomaly['model_config__model_name'] or 'Unknown'
|
||||
anomaly['function'] = anomaly['operation_type'].replace('_', ' ').title() if anomaly['operation_type'] else 'Unknown'
|
||||
anomaly['cost'] = anomaly['cost_usd_total']
|
||||
anomaly['tokens'] = (anomaly['tokens_input'] or 0) + (anomaly['tokens_output'] or 0)
|
||||
else:
|
||||
anomalies = []
|
||||
|
||||
# Model comparison matrix
|
||||
model_comparison = []
|
||||
for model_data in cost_by_model:
|
||||
model_comparison.append({
|
||||
'model': model_data['model'],
|
||||
'total_cost': float(model_data['total_cost']),
|
||||
'calls': model_data['call_count'],
|
||||
'avg_cost': float(model_data['avg_cost']),
|
||||
'total_tokens': model_data['total_tokens'],
|
||||
'cost_per_1k': model_data['cost_per_1k_tokens'],
|
||||
})
|
||||
|
||||
# Cost distribution percentages
|
||||
if total_cost > 0:
|
||||
for item in cost_by_model:
|
||||
item['cost_percentage'] = float((item['total_cost'] / total_cost) * 100)
|
||||
|
||||
# Peak cost hours
|
||||
hourly_cost = logs.extra(
|
||||
select={'hour': "EXTRACT(hour FROM created_at)"}
|
||||
).values('hour').annotate(
|
||||
total_cost=Sum('cost_usd_total'),
|
||||
call_count=Count('id')
|
||||
).order_by('hour')
|
||||
|
||||
# Cost efficiency score
|
||||
successful_cost = total_cost
|
||||
efficiency_score = 100.0
|
||||
|
||||
context = {
|
||||
'title': 'AI Cost & Margin Analysis',
|
||||
'days_filter': days,
|
||||
'total_cost': float(total_cost),
|
||||
'total_revenue': float(total_revenue),
|
||||
'total_margin': float(total_margin),
|
||||
'margin_percentage': round(margin_percentage, 2),
|
||||
'margin_per_1m_tokens': round(margin_per_1m_tokens, 4),
|
||||
'margin_per_1k_credits': round(margin_per_1k_credits, 4),
|
||||
'total_credits_charged': total_credits_charged,
|
||||
'credit_price': float(avg_credit_price),
|
||||
'total_calls': total_calls,
|
||||
'avg_cost_per_call': float(avg_cost_per_call),
|
||||
'total_tokens': int(total_tokens),
|
||||
'cost_by_model': list(cost_by_model),
|
||||
'cost_by_account': list(cost_by_account),
|
||||
'cost_by_function': list(cost_by_function),
|
||||
'daily_cost_labels': json.dumps(daily_cost_labels),
|
||||
'daily_cost_data': json.dumps(daily_cost_data),
|
||||
'daily_call_data': json.dumps(daily_call_data),
|
||||
'projected_monthly': round(projected_monthly, 2),
|
||||
'failed_cost': float(failed_cost),
|
||||
'wasted_percentage': float((failed_cost / total_cost * 100) if total_cost > 0 else 0),
|
||||
'anomalies': list(anomalies),
|
||||
'model_comparison': model_comparison,
|
||||
'hourly_cost': list(hourly_cost),
|
||||
'efficiency_score': round(efficiency_score, 2),
|
||||
'successful_cost': float(successful_cost),
|
||||
'current_app': '_reports',
|
||||
}
|
||||
|
||||
# Merge with admin context
|
||||
from igny8_core.admin.site import admin_site
|
||||
admin_context = admin_site.each_context(request)
|
||||
context.update(admin_context)
|
||||
|
||||
return render(request, 'admin/reports/ai_cost_analysis.html', context)
|
||||
|
||||
@@ -24,7 +24,10 @@ class Igny8AdminSite(UnfoldAdminSite):
|
||||
"""Get admin URLs with dashboard and reports available"""
|
||||
from django.urls import path
|
||||
from .dashboard import admin_dashboard
|
||||
from .reports import revenue_report, usage_report, content_report, data_quality_report
|
||||
from .reports import (
|
||||
revenue_report, usage_report, content_report, data_quality_report,
|
||||
token_usage_report, ai_cost_analysis
|
||||
)
|
||||
|
||||
urls = super().get_urls()
|
||||
custom_urls = [
|
||||
@@ -33,6 +36,8 @@ class Igny8AdminSite(UnfoldAdminSite):
|
||||
path('reports/usage/', self.admin_view(usage_report), name='report_usage'),
|
||||
path('reports/content/', self.admin_view(content_report), name='report_content'),
|
||||
path('reports/data-quality/', self.admin_view(data_quality_report), name='report_data_quality'),
|
||||
path('reports/token-usage/', self.admin_view(token_usage_report), name='report_token_usage'),
|
||||
path('reports/ai-cost-analysis/', self.admin_view(ai_cost_analysis), name='report_ai_cost'),
|
||||
]
|
||||
return custom_urls + urls
|
||||
|
||||
@@ -292,6 +297,20 @@ class Igny8AdminSite(UnfoldAdminSite):
|
||||
'view_only': True,
|
||||
'perms': {'view': True},
|
||||
},
|
||||
{
|
||||
'name': 'Token Usage Report',
|
||||
'object_name': 'TokenUsageReport',
|
||||
'admin_url': '/admin/reports/token-usage/',
|
||||
'view_only': True,
|
||||
'perms': {'view': True},
|
||||
},
|
||||
{
|
||||
'name': 'AI Cost Analysis',
|
||||
'object_name': 'AICostAnalysis',
|
||||
'admin_url': '/admin/reports/ai-cost-analysis/',
|
||||
'view_only': True,
|
||||
'perms': {'view': True},
|
||||
},
|
||||
],
|
||||
})
|
||||
|
||||
|
||||
Reference in New Issue
Block a user