Skip to content

Speakr is a personal, self-hosted web application designed for transcribing audio recordings

Python

Add token usage tracking and per-user budget enforcement

Murtaza Nasir3d08d7889116
diff --git a/src/api/admin.py b/src/api/admin.py
index 34dfa940..5b05d1df 100644
--- a/src/api/admin.py
+++ b/src/api/admin.py
@@ -17,6 +17,7 @@ from src.models import *
from src.utils import *
from src.services.retention import is_recording_exempt_from_deletion, get_retention_days_for_recording, process_auto_deletion
from src.services.embeddings import EMBEDDINGS_AVAILABLE, process_recording_chunks
+from src.services.token_tracking import token_tracker
from src.config.startup import get_file_monitor_functions
# Create blueprint
@@ -130,7 +131,11 @@ def admin_get_users():
# Get recordings count and storage used
recordings_count = len(user.recordings)
storage_used = sum(r.file_size for r in user.recordings if r.file_size) or 0
-
+
+ # Get current month token usage
+ current_usage = token_tracker.get_monthly_usage(user.id)
+ usage_percentage = (current_usage / user.monthly_token_budget * 100) if user.monthly_token_budget else 0
+
user_data.append({
'id': user.id,
'username': user.username,
@@ -138,7 +143,10 @@ def admin_get_users():
'is_admin': user.is_admin,
'can_share_publicly': user.can_share_publicly,
'recordings_count': recordings_count,
- 'storage_used': storage_used
+ 'storage_used': storage_used,
+ 'monthly_token_budget': user.monthly_token_budget,
+ 'current_token_usage': current_usage,
+ 'token_usage_percentage': round(usage_percentage, 1)
})
return jsonify(user_data)
@@ -175,19 +183,23 @@ def admin_add_user():
username=data['username'],
email=data['email'],
password=hashed_password,
- is_admin=data.get('is_admin', False)
+ is_admin=data.get('is_admin', False),
+ monthly_token_budget=data.get('monthly_token_budget')
)
-
+
db.session.add(new_user)
db.session.commit()
-
+
return jsonify({
'id': new_user.id,
'username': new_user.username,
'email': new_user.email,
'is_admin': new_user.is_admin,
'recordings_count': 0,
- 'storage_used': 0
+ 'storage_used': 0,
+ 'monthly_token_budget': new_user.monthly_token_budget,
+ 'current_token_usage': 0,
+ 'token_usage_percentage': 0
}), 201
@@ -229,12 +241,24 @@ def admin_update_user(user_id):
if 'can_share_publicly' in data:
user.can_share_publicly = data['can_share_publicly']
+ if 'monthly_token_budget' in data:
+ # Allow setting to None (unlimited) or a positive integer
+ budget = data['monthly_token_budget']
+ if budget is None or budget == '' or budget == 0:
+ user.monthly_token_budget = None
+ else:
+ user.monthly_token_budget = int(budget)
+
db.session.commit()
# Get recordings count and storage used
recordings_count = len(user.recordings)
storage_used = sum(r.file_size for r in user.recordings if r.file_size) or 0
+ # Get current month token usage
+ current_usage = token_tracker.get_monthly_usage(user.id)
+ usage_percentage = (current_usage / user.monthly_token_budget * 100) if user.monthly_token_budget else 0
+
return jsonify({
'id': user.id,
'username': user.username,
@@ -242,7 +266,10 @@ def admin_update_user(user_id):
'is_admin': user.is_admin,
'can_share_publicly': user.can_share_publicly,
'recordings_count': recordings_count,
- 'storage_used': storage_used
+ 'storage_used': storage_used,
+ 'monthly_token_budget': user.monthly_token_budget,
+ 'current_token_usage': current_usage,
+ 'token_usage_percentage': round(usage_percentage, 1)
})
@@ -368,6 +395,111 @@ def admin_get_stats():
'total_queries': total_queries
})
+
+# --- Token Usage Stats ---
+
+@admin_bp.route('/admin/token-stats', methods=['GET'])
+@login_required
+def admin_get_token_stats():
+ """Get overall token usage statistics."""
+ if not current_user.is_admin:
+ return jsonify({'error': 'Unauthorized'}), 403
+
+ try:
+ # Get today's usage
+ today_usage = token_tracker.get_today_usage()
+
+ # Get current month usage for all users
+ monthly_stats = token_tracker.get_monthly_stats(months=1)
+ current_month = monthly_stats[-1] if monthly_stats else {'tokens': 0, 'cost': 0}
+
+ # Get per-user stats for current month
+ user_stats = token_tracker.get_user_stats()
+
+ # Calculate totals
+ total_monthly_tokens = current_month.get('tokens', 0)
+ total_monthly_cost = current_month.get('cost', 0)
+
+ return jsonify({
+ 'today': today_usage,
+ 'current_month': {
+ 'tokens': total_monthly_tokens,
+ 'cost': total_monthly_cost
+ },
+ 'user_count_with_usage': len([u for u in user_stats if u['current_usage'] > 0]),
+ 'users_over_80_percent': len([u for u in user_stats if u['percentage'] >= 80]),
+ 'users_at_100_percent': len([u for u in user_stats if u['percentage'] >= 100])
+ })
+
+ except Exception as e:
+ current_app.logger.error(f"Error getting token stats: {e}")
+ return jsonify({'error': str(e)}), 500
+
+
+@admin_bp.route('/admin/token-stats/daily', methods=['GET'])
+@login_required
+def admin_get_daily_token_stats():
+ """Get daily token usage for charts (last 30 days)."""
+ if not current_user.is_admin:
+ return jsonify({'error': 'Unauthorized'}), 403
+
+ try:
+ days = request.args.get('days', 30, type=int)
+ user_id = request.args.get('user_id', type=int)
+
+ daily_stats = token_tracker.get_daily_stats(days=days, user_id=user_id)
+
+ return jsonify({
+ 'stats': daily_stats,
+ 'days': days
+ })
+
+ except Exception as e:
+ current_app.logger.error(f"Error getting daily token stats: {e}")
+ return jsonify({'error': str(e)}), 500
+
+
+@admin_bp.route('/admin/token-stats/monthly', methods=['GET'])
+@login_required
+def admin_get_monthly_token_stats():
+ """Get monthly token usage for charts (last 12 months)."""
+ if not current_user.is_admin:
+ return jsonify({'error': 'Unauthorized'}), 403
+
+ try:
+ months = request.args.get('months', 12, type=int)
+
+ monthly_stats = token_tracker.get_monthly_stats(months=months)
+
+ return jsonify({
+ 'stats': monthly_stats,
+ 'months': months
+ })
+
+ except Exception as e:
+ current_app.logger.error(f"Error getting monthly token stats: {e}")
+ return jsonify({'error': str(e)}), 500
+
+
+@admin_bp.route('/admin/token-stats/users', methods=['GET'])
+@login_required
+def admin_get_user_token_stats():
+ """Get per-user token usage for current month."""
+ if not current_user.is_admin:
+ return jsonify({'error': 'Unauthorized'}), 403
+
+ try:
+ user_stats = token_tracker.get_user_stats()
+
+ return jsonify({
+ 'users': user_stats
+ })
+
+ except Exception as e:
+ current_app.logger.error(f"Error getting user token stats: {e}")
+ return jsonify({'error': str(e)}), 500
+
+
# --- Transcript Template Routes ---
diff --git a/src/api/inquire.py b/src/api/inquire.py
index 643f0a25..9c1b20aa 100644
--- a/src/api/inquire.py
+++ b/src/api/inquire.py
@@ -17,7 +17,7 @@ from src.database import db
from src.models import *
from src.utils import *
from src.services.embeddings import get_accessible_recording_ids, semantic_search_chunks
-from src.services.llm import call_llm_completion, call_chat_completion, process_streaming_with_thinking, client, chat_client
+from src.services.llm import call_llm_completion, call_chat_completion, process_streaming_with_thinking, client, chat_client, TokenBudgetExceeded
# Create blueprint
inquire_bp = Blueprint('inquire', __name__)
@@ -248,7 +248,9 @@ Examples:
{"role": "user", "content": router_prompt}
],
temperature=0.1,
- max_tokens=10
+ max_tokens=10,
+ user_id=user_id,
+ operation_type='query_routing'
)
raw_decision = router_response.choices[0].message.content
@@ -279,11 +281,13 @@ Use proper markdown formatting including headings (##), bold (**text**), bullet
],
temperature=0.7,
max_tokens=int(os.environ.get("CHAT_MAX_TOKENS", "2000")),
- stream=True
+ stream=True,
+ user_id=user_id,
+ operation_type='chat'
)
# Use helper function to process streaming with thinking tag support
- for response in process_streaming_with_thinking(stream):
+ for response in process_streaming_with_thinking(stream, user_id=user_id, operation_type='chat', model_name=os.environ.get('LLM_MODEL')):
yield response
return
@@ -325,7 +329,9 @@ Respond with only a JSON array of strings: ["term1", "term2", "term3", ...]"""
{"role": "user", "content": enrichment_prompt}
],
temperature=0.3,
- max_tokens=200
+ max_tokens=200,
+ user_id=user_id,
+ operation_type='query_enrichment'
)
raw_content = enrichment_response.choices[0].message.content
@@ -635,7 +641,9 @@ Order your response with notes from the most recent meetings first. Always use p
messages=messages,
temperature=0.7,
max_tokens=int(os.environ.get("CHAT_MAX_TOKENS", "2000")),
- stream=True
+ stream=True,
+ user_id=user_id,
+ operation_type='chat'
)
# Buffer content to detect full transcript requests
@@ -697,11 +705,13 @@ Order your response with notes from the most recent meetings first. Always use p
messages=updated_messages,
temperature=0.7,
max_tokens=int(os.environ.get("CHAT_MAX_TOKENS", "2000")),
- stream=True
+ stream=True,
+ user_id=user_id,
+ operation_type='chat'
)
-
+
# Use helper function to process streaming with thinking tag support
- for response in process_streaming_with_thinking(new_stream):
+ for response in process_streaming_with_thinking(new_stream, user_id=user_id, operation_type='chat', model_name=os.environ.get('CHAT_MODEL')):
yield response
return
else:
@@ -768,6 +778,9 @@ Order your response with notes from the most recent meetings first. Always use p
yield f"data: {json.dumps({'end_of_stream': True})}\n\n"
+ except TokenBudgetExceeded as e:
+ app.logger.warning(f"Token budget exceeded for user {user_id}: {e}")
+ yield f"data: {json.dumps({'error': str(e), 'budget_exceeded': True})}\n\n"
except Exception as e:
app.logger.error(f"Error in enhanced chat generation: {e}")
yield f"data: {json.dumps({'error': str(e)})}\n\n"
diff --git a/src/api/recordings.py b/src/api/recordings.py
index 32702c07..9923bdc4 100644
--- a/src/api/recordings.py
+++ b/src/api/recordings.py
@@ -29,7 +29,7 @@ from src.services.speaker import update_speaker_usage, identify_unidentified_spe
from src.services.speaker_embedding_matcher import update_speaker_embedding
from src.services.speaker_snippets import create_speaker_snippets
from src.services.document import process_markdown_to_docx
-from src.services.llm import client, chat_client, call_llm_completion, call_chat_completion, process_streaming_with_thinking
+from src.services.llm import client, chat_client, call_llm_completion, call_chat_completion, process_streaming_with_thinking, TokenBudgetExceeded
from src.services.embeddings import process_recording_chunks
from src.file_exporter import export_recording, mark_export_as_deleted
from src.utils.ffprobe import get_codec_info, get_creation_date, FFProbeError
@@ -982,7 +982,9 @@ JSON Response:
{"role": "user", "content": prompt}
],
temperature=0.2,
- response_format={"type": "json_object"}
+ response_format={"type": "json_object"},
+ user_id=current_user.id,
+ operation_type='speaker_identification'
)
response_content = completion.choices[0].message.content
current_app.logger.info(f"[Auto-Identify] LLM Raw Response: {response_content}")
@@ -2773,22 +2775,31 @@ Additional context and notes about the meeting:
messages.extend(message_history)
messages.append({"role": "user", "content": user_message})
+ # Capture context before generator starts (app context may not be available inside generator)
+ user_id = current_user.id
+ app = current_app._get_current_object()
+
def generate():
try:
- # Enable streaming
+ # Enable streaming with user_id for budget enforcement
stream = call_chat_completion(
messages=messages,
temperature=0.7,
max_tokens=int(os.environ.get("CHAT_MAX_TOKENS", "2000")),
- stream=True
+ stream=True,
+ user_id=user_id,
+ operation_type='chat'
)
# Use helper function to process streaming with thinking tag support
- for response in process_streaming_with_thinking(stream):
+ for response in process_streaming_with_thinking(stream, user_id=user_id, operation_type='chat', app=app):
yield response
+ except TokenBudgetExceeded as e:
+ app.logger.warning(f"Token budget exceeded for user {user_id}: {e}")
+ yield f"data: {json.dumps({'error': str(e), 'budget_exceeded': True})}\n\n"
except Exception as e:
- current_app.logger.error(f"Error during chat stream generation: {str(e)}")
+ app.logger.error(f"Error during chat stream generation: {str(e)}")
# Yield an error message in SSE format
yield f"data: {json.dumps({'error': str(e)})}\n\n"
diff --git a/src/api/system.py b/src/api/system.py
index e1db2ea0..9bd537d6 100644
--- a/src/api/system.py
+++ b/src/api/system.py
@@ -17,6 +17,7 @@ from src.utils import *
from src.config.version import get_version
from src.services.llm import TEXT_MODEL_BASE_URL, TEXT_MODEL_NAME
from src.config.app_config import ASR_BASE_URL
+from src.services.token_tracking import token_tracker
# Create blueprint
system_bp = Blueprint('system', __name__)
@@ -79,6 +80,38 @@ def save_user_preferences():
'ui_language': current_user.ui_language
})
+
+@system_bp.route('/api/user/token-budget', methods=['GET'])
+@login_required
+def get_user_token_budget():
+ """Get current user's token budget status."""
+ try:
+ user = current_user
+
+ # If user has no budget, return null to indicate unlimited
+ if not user.monthly_token_budget:
+ return jsonify({
+ 'has_budget': False,
+ 'budget': None,
+ 'usage': 0,
+ 'percentage': 0
+ })
+
+ # Get current usage
+ current_usage = token_tracker.get_monthly_usage(user.id)
+ percentage = (current_usage / user.monthly_token_budget) * 100
+
+ return jsonify({
+ 'has_budget': True,
+ 'budget': user.monthly_token_budget,
+ 'usage': current_usage,
+ 'percentage': round(percentage, 1)
+ })
+ except Exception as e:
+ current_app.logger.error(f"Error getting token budget for user {current_user.id}: {e}")
+ return jsonify({'error': str(e)}), 500
+
+
# --- System Info API Endpoint ---
diff --git a/src/init_db.py b/src/init_db.py
index 9d5c42bb..34293989 100644
--- a/src/init_db.py
+++ b/src/init_db.py
@@ -230,6 +230,10 @@ def initialize_database(app):
if add_column_if_not_exists(engine, 'user', 'can_share_publicly', 'BOOLEAN DEFAULT 1'):
app.logger.info("Added can_share_publicly column to user table")
+ # Token budget for rate limiting
+ if add_column_if_not_exists(engine, 'user', 'monthly_token_budget', 'INTEGER'):
+ app.logger.info("Added monthly_token_budget column to user table")
+
# Add source tracking columns to internal_share table
if add_column_if_not_exists(engine, 'internal_share', 'source_type', 'VARCHAR(20) DEFAULT "manual"'):
app.logger.info("Added source_type column to internal_share table")
diff --git a/src/models/__init__.py b/src/models/__init__.py
index 7312d268..3149db09 100644
--- a/src/models/__init__.py
+++ b/src/models/__init__.py
@@ -27,6 +27,7 @@ from .system import SystemSetting
from .audit import ShareAuditLog
from .push_subscription import PushSubscription
from .processing_job import ProcessingJob
+from .token_usage import TokenUsage
# Export all models
__all__ = [
@@ -57,4 +58,5 @@ __all__ = [
'SystemSetting',
'PushSubscription',
'ProcessingJob',
+ 'TokenUsage',
]
diff --git a/src/models/token_usage.py b/src/models/token_usage.py
new file mode 100644
index 00000000..0868fff9
--- /dev/null
+++ b/src/models/token_usage.py
@@ -0,0 +1,44 @@
+"""
+Token usage tracking model for monitoring LLM API consumption.
+"""
+
+from datetime import datetime, date
+from src.database import db
+
+
+class TokenUsage(db.Model):
+ """Daily token usage aggregates per user per operation type."""
+ __tablename__ = 'token_usage'
+
+ id = db.Column(db.Integer, primary_key=True)
+ user_id = db.Column(db.Integer, db.ForeignKey('user.id'), nullable=False)
+ date = db.Column(db.Date, nullable=False, default=date.today)
+ operation_type = db.Column(db.String(50), nullable=False)
+
+ # Token counts (from API response.usage)
+ prompt_tokens = db.Column(db.Integer, default=0)
+ completion_tokens = db.Column(db.Integer, default=0)
+ total_tokens = db.Column(db.Integer, default=0)
+
+ # Cost tracking (OpenRouter provides this)
+ cost = db.Column(db.Float, default=0.0)
+
+ # Request count for this day/operation
+ request_count = db.Column(db.Integer, default=0)
+
+ # Model info
+ model_name = db.Column(db.String(100), nullable=True)
+
+ created_at = db.Column(db.DateTime, default=datetime.utcnow)
+ updated_at = db.Column(db.DateTime, default=datetime.utcnow, onupdate=datetime.utcnow)
+
+ # Relationships
+ user = db.relationship('User', backref=db.backref('token_usage', lazy='dynamic'))
+
+ __table_args__ = (
+ db.UniqueConstraint('user_id', 'date', 'operation_type', name='uq_user_date_op'),
+ db.Index('idx_token_user_date', 'user_id', 'date'),
+ )
+
+ def __repr__(self):
+ return f'<TokenUsage {self.user_id} {self.date} {self.operation_type}: {self.total_tokens}>'
diff --git a/src/models/user.py b/src/models/user.py
index 7ac35519..fc05ebf3 100644
--- a/src/models/user.py
+++ b/src/models/user.py
@@ -32,6 +32,9 @@ class User(db.Model, UserMixin):
company = db.Column(db.String(100), nullable=True)
diarize = db.Column(db.Boolean, default=False)
+ # Token budget (None = unlimited)
+ monthly_token_budget = db.Column(db.Integer, nullable=True)
+
def __repr__(self):
return f"User('{self.username}', '{self.email}')"
diff --git a/src/services/llm.py b/src/services/llm.py
index d654c08c..e473b38c 100644
--- a/src/services/llm.py
+++ b/src/services/llm.py
@@ -12,6 +12,14 @@ from openai import OpenAI
# Use standard logging instead of current_app.logger for context independence
logger = logging.getLogger(__name__)
+
+class TokenBudgetExceeded(Exception):
+ """Raised when user exceeds their token budget."""
+ def __init__(self, message, usage_percentage=100):
+ self.message = message
+ self.usage_percentage = usage_percentage
+ super().__init__(message)
+
from src.utils import safe_json_loads, extract_json_object
# Configuration - use TEXT_MODEL_* variables for LLM
@@ -128,7 +136,8 @@ def is_using_openai_api():
-def call_llm_completion(messages, temperature=0.7, response_format=None, stream=False, max_tokens=None):
+def call_llm_completion(messages, temperature=0.7, response_format=None, stream=False, max_tokens=None,
+ user_id=None, operation_type=None):
"""
Centralized function for LLM API calls with proper error handling and logging.
@@ -138,6 +147,8 @@ def call_llm_completion(messages, temperature=0.7, response_format=None, stream=
response_format: Optional response format dict (e.g., {"type": "json_object"})
stream: Whether to stream the response
max_tokens: Optional maximum tokens to generate
+ user_id: Optional user ID for token tracking and budget enforcement
+ operation_type: Optional operation type for token tracking (e.g., 'summarization', 'chat')
Returns:
OpenAI completion object or generator (if streaming)
@@ -148,6 +159,21 @@ def call_llm_completion(messages, temperature=0.7, response_format=None, stream=
if not TEXT_MODEL_API_KEY:
raise ValueError("TEXT_MODEL_API_KEY not configured")
+ # Check budget before making the call
+ if user_id and operation_type:
+ try:
+ from src.services.token_tracking import token_tracker
+ can_proceed, usage_pct, msg = token_tracker.check_budget(user_id)
+ if not can_proceed:
+ raise TokenBudgetExceeded(msg, usage_pct)
+ if usage_pct >= 80:
+ logger.warning(f"User {user_id} at {usage_pct:.1f}% of token budget")
+ except TokenBudgetExceeded:
+ raise
+ except Exception as e:
+ # Log but don't block on budget check errors
+ logger.warning(f"Budget check failed for user {user_id}: {e}")
+
try:
# Check if we're using GPT-5 with OpenAI API
using_gpt5 = is_gpt5_model(TEXT_MODEL_NAME) and is_using_openai_api()
@@ -158,6 +184,10 @@ def call_llm_completion(messages, temperature=0.7, response_format=None, stream=
"stream": stream
}
+ # Add stream_options to get usage in final chunk for streaming
+ if stream:
+ completion_args["stream_options"] = {"include_usage": True}
+
if using_gpt5:
# GPT-5 models don't support temperature, top_p, or logprobs
# They use reasoning_effort and verbosity instead
@@ -186,6 +216,22 @@ def call_llm_completion(messages, temperature=0.7, response_format=None, stream=
response = client.chat.completions.create(**completion_args)
+ # Track usage for non-streaming calls
+ if user_id and operation_type and not stream and response.usage:
+ try:
+ from src.services.token_tracking import token_tracker
+ token_tracker.record_usage(
+ user_id=user_id,
+ operation_type=operation_type,
+ prompt_tokens=response.usage.prompt_tokens,
+ completion_tokens=response.usage.completion_tokens,
+ total_tokens=response.usage.total_tokens,
+ model_name=TEXT_MODEL_NAME,
+ cost=getattr(response.usage, 'cost', None)
+ )
+ except Exception as e:
+ logger.warning(f"Failed to record token usage: {e}")
+
# Debug log for empty responses
if not stream and response.choices:
content = response.choices[0].message.content
@@ -199,12 +245,15 @@ def call_llm_completion(messages, temperature=0.7, response_format=None, stream=
return response
+ except TokenBudgetExceeded:
+ raise
except Exception as e:
logger.error(f"LLM API call failed: {e}")
raise
-def call_chat_completion(messages, temperature=0.7, response_format=None, stream=False, max_tokens=None):
+def call_chat_completion(messages, temperature=0.7, response_format=None, stream=False, max_tokens=None,
+ user_id=None, operation_type=None):
"""
Chat-specific LLM completion function. Uses dedicated chat model if configured,
otherwise falls back to standard TEXT_MODEL configuration.
@@ -215,6 +264,8 @@ def call_chat_completion(messages, temperature=0.7, response_format=None, stream
response_format: Optional response format dict (e.g., {"type": "json_object"})
stream: Whether to stream the response
max_tokens: Optional maximum tokens to generate
+ user_id: Optional user ID for token tracking and budget enforcement
+ operation_type: Optional operation type for token tracking (e.g., 'chat')
Returns:
OpenAI completion object or generator (if streaming)
@@ -228,6 +279,21 @@ def call_chat_completion(messages, temperature=0.7, response_format=None, stream
if not chat_config['api_key']:
raise ValueError("Chat model API key not configured")
+ # Check budget before making the call
+ if user_id and operation_type:
+ try:
+ from src.services.token_tracking import token_tracker
+ can_proceed, usage_pct, msg = token_tracker.check_budget(user_id)
+ if not can_proceed:
+ raise TokenBudgetExceeded(msg, usage_pct)
+ if usage_pct >= 80:
+ logger.warning(f"User {user_id} at {usage_pct:.1f}% of token budget")
+ except TokenBudgetExceeded:
+ raise
+ except Exception as e:
+ # Log but don't block on budget check errors
+ logger.warning(f"Budget check failed for user {user_id}: {e}")
+
try:
model_name = chat_config['model_name']
base_url = chat_config['base_url'] or ''
@@ -241,6 +307,10 @@ def call_chat_completion(messages, temperature=0.7, response_format=None, stream
"stream": stream
}
+ # Add stream_options to get usage in final chunk for streaming
+ if stream:
+ completion_args["stream_options"] = {"include_usage": True}
+
if using_gpt5:
logger.debug(f"Using GPT-5 chat model: {model_name}")
# Use chat-specific GPT-5 settings from config
@@ -259,6 +329,22 @@ def call_chat_completion(messages, temperature=0.7, response_format=None, stream
response = effective_client.chat.completions.create(**completion_args)
+ # Track usage for non-streaming calls
+ if user_id and operation_type and not stream and response.usage:
+ try:
+ from src.services.token_tracking import token_tracker
+ token_tracker.record_usage(
+ user_id=user_id,
+ operation_type=operation_type,
+ prompt_tokens=response.usage.prompt_tokens,
+ completion_tokens=response.usage.completion_tokens,
+ total_tokens=response.usage.total_tokens,
+ model_name=model_name,
+ cost=getattr(response.usage, 'cost', None)
+ )
+ except Exception as e:
+ logger.warning(f"Failed to record token usage: {e}")
+
# Debug log for empty responses
if not stream and response.choices:
content = response.choices[0].message.content
@@ -267,6 +353,8 @@ def call_chat_completion(messages, temperature=0.7, response_format=None, stream
return response
+ except TokenBudgetExceeded:
+ raise
except Exception as e:
logger.error(f"Chat LLM API call failed: {e}")
raise
@@ -297,18 +385,55 @@ def format_api_error_message(error_str):
return f"[Summary generation failed: {error_str}]"
-def process_streaming_with_thinking(stream):
+def process_streaming_with_thinking(stream, user_id=None, operation_type=None, model_name=None, app=None):
"""
Generator that processes a streaming response and separates thinking content.
Yields SSE-formatted data with 'delta' for regular content and 'thinking' for thinking content.
+
+ Args:
+ stream: The streaming response from the LLM API
+ user_id: Optional user ID for token tracking
+ operation_type: Optional operation type for token tracking
+ model_name: Optional model name for token tracking
+ app: Optional Flask app instance for database context in generators
"""
content_buffer = ""
in_thinking = False
thinking_buffer = ""
for chunk in stream:
- content = chunk.choices[0].delta.content
- if content:
+ # Check for usage in final chunk (from stream_options={'include_usage': True})
+ if hasattr(chunk, 'usage') and chunk.usage and user_id and operation_type:
+ try:
+ from src.services.token_tracking import token_tracker
+ # Use app context if provided (needed for generators where context may be lost)
+ if app:
+ with app.app_context():
+ token_tracker.record_usage(
+ user_id=user_id,
+ operation_type=operation_type,
+ prompt_tokens=chunk.usage.prompt_tokens,
+ completion_tokens=chunk.usage.completion_tokens,
+ total_tokens=chunk.usage.total_tokens,
+ model_name=model_name or TEXT_MODEL_NAME,
+ cost=getattr(chunk.usage, 'cost', None)
+ )
+ else:
+ token_tracker.record_usage(
+ user_id=user_id,
+ operation_type=operation_type,
+ prompt_tokens=chunk.usage.prompt_tokens,
+ completion_tokens=chunk.usage.completion_tokens,
+ total_tokens=chunk.usage.total_tokens,
+ model_name=model_name or TEXT_MODEL_NAME,
+ cost=getattr(chunk.usage, 'cost', None)
+ )
+ except Exception as e:
+ logger.warning(f"Failed to record streaming token usage: {e}")
+
+ # Process content delta
+ if chunk.choices and chunk.choices[0].delta.content:
+ content = chunk.choices[0].delta.content
content_buffer += content
# Process the buffer to detect and handle thinking tags
diff --git a/src/services/token_tracking.py b/src/services/token_tracking.py
new file mode 100644
index 00000000..d05390da
--- /dev/null
+++ b/src/services/token_tracking.py
@@ -0,0 +1,270 @@
+"""
+Token usage tracking service for monitoring LLM API consumption and budget enforcement.
+"""
+
+import logging
+from datetime import date, datetime, timedelta
+from typing import Tuple, Optional, Dict, List
+
+from sqlalchemy import func, extract
+
+from src.database import db
+from src.models.token_usage import TokenUsage
+from src.models.user import User
+
+logger = logging.getLogger(__name__)
+
+
+class TokenTracker:
+ """Service for recording and checking token usage."""
+
+ OPERATION_TYPES = [
+ 'summarization',
+ 'chat',
+ 'title_generation',
+ 'event_extraction',
+ 'query_routing',
+ 'query_enrichment'
+ ]
+
+ def record_usage(
+ self,
+ user_id: int,
+ operation_type: str,
+ prompt_tokens: int,
+ completion_tokens: int,
+ total_tokens: int,
+ model_name: str = None,
+ cost: float = None
+ ):
+ """
+ Record token usage - upserts into daily aggregate.
+
+ Args:
+ user_id: User ID who made the request
+ operation_type: Type of operation (summarization, chat, etc.)
+ prompt_tokens: Number of input tokens
+ completion_tokens: Number of output tokens
+ total_tokens: Total tokens (prompt + completion)
+ model_name: Name of the model used
+ cost: API cost if available (e.g., from OpenRouter)
+ """
+ try:
+ today = date.today()
+
+ # Find or create today's record for this user/operation
+ usage = TokenUsage.query.filter_by(
+ user_id=user_id,
+ date=today,
+ operation_type=operation_type
+ ).first()
+
+ if usage:
+ # Update existing record
+ usage.prompt_tokens += prompt_tokens
+ usage.completion_tokens += completion_tokens
+ usage.total_tokens += total_tokens
+ usage.request_count += 1
+ if cost:
+ usage.cost += cost
+ if model_name:
+ usage.model_name = model_name # Update to latest model used
+ else:
+ # Create new record
+ usage = TokenUsage(
+ user_id=user_id,
+ date=today,
+ operation_type=operation_type,
+ prompt_tokens=prompt_tokens,
+ completion_tokens=completion_tokens,
+ total_tokens=total_tokens,
+ request_count=1,
+ model_name=model_name,
+ cost=cost or 0.0
+ )
+ db.session.add(usage)
+
+ db.session.commit()
+ logger.debug(f"Recorded {total_tokens} tokens for user {user_id}, operation {operation_type}")
+ return usage
+
+ except Exception as e:
+ logger.error(f"Failed to record token usage: {e}")
+ db.session.rollback()
+ return None
+
+ def get_monthly_usage(self, user_id: int, year: int = None, month: int = None) -> int:
+ """Get total tokens used by a user in a given month."""
+ if year is None:
+ year = date.today().year
+ if month is None:
+ month = date.today().month
+
+ result = db.session.query(func.sum(TokenUsage.total_tokens)).filter(
+ TokenUsage.user_id == user_id,
+ extract('year', TokenUsage.date) == year,
+ extract('month', TokenUsage.date) == month
+ ).scalar()
+
+ return result or 0
+
+ def get_monthly_cost(self, user_id: int, year: int = None, month: int = None) -> float:
+ """Get total cost for a user in a given month."""
+ if year is None:
+ year = date.today().year
+ if month is None:
+ month = date.today().month
+
+ result = db.session.query(func.sum(TokenUsage.cost)).filter(
+ TokenUsage.user_id == user_id,
+ extract('year', TokenUsage.date) == year,
+ extract('month', TokenUsage.date) == month
+ ).scalar()
+
+ return result or 0.0
+
+ def check_budget(self, user_id: int) -> Tuple[bool, float, Optional[str]]:
+ """
+ Check if user is within budget.
+
+ Returns:
+ (can_proceed, usage_percentage, message)
+ - can_proceed: False if hard cap (100%) reached
+ - usage_percentage: 0-100+
+ - message: Warning/error message if applicable
+ """
+ try:
+ user = db.session.get(User, user_id)
+ if not user or not user.monthly_token_budget:
+ return (True, 0, None) # No budget = unlimited
+
+ current_usage = self.get_monthly_usage(user_id)
+ budget = user.monthly_token_budget
+ percentage = (current_usage / budget) * 100
+
+ if percentage >= 100:
+ return (False, percentage,
+ f"Monthly token budget exceeded ({percentage:.1f}%). Contact admin for more tokens.")
+ elif percentage >= 80:
+ return (True, percentage,
+ f"Warning: {percentage:.1f}% of monthly token budget used.")
+ else:
+ return (True, percentage, None)
+
+ except Exception as e:
+ logger.error(f"Failed to check budget for user {user_id}: {e}")
+ # Fail open - allow the request if we can't check
+ return (True, 0, None)
+
+ def get_daily_stats(self, days: int = 30, user_id: int = None) -> List[Dict]:
+ """Get daily token usage for charts."""
+ start_date = date.today() - timedelta(days=days - 1)
+
+ query = db.session.query(
+ TokenUsage.date,
+ TokenUsage.operation_type,
+ func.sum(TokenUsage.total_tokens).label('tokens'),
+ func.sum(TokenUsage.cost).label('cost')
+ ).filter(TokenUsage.date >= start_date)
+
+ if user_id:
+ query = query.filter(TokenUsage.user_id == user_id)
+
+ results = query.group_by(TokenUsage.date, TokenUsage.operation_type).all()
+
+ # Organize by date
+ stats = {}
+ for r in results:
+ date_str = r.date.isoformat()
+ if date_str not in stats:
+ stats[date_str] = {'date': date_str, 'total': 0, 'cost': 0.0, 'by_operation': {}}
+ stats[date_str]['total'] += r.tokens or 0
+ stats[date_str]['cost'] += r.cost or 0
+ stats[date_str]['by_operation'][r.operation_type] = r.tokens or 0
+
+ # Fill in missing dates with zeros
+ all_dates = []
+ current = start_date
+ while current <= date.today():
+ date_str = current.isoformat()
+ if date_str not in stats:
+ stats[date_str] = {'date': date_str, 'total': 0, 'cost': 0.0, 'by_operation': {}}
+ all_dates.append(date_str)
+ current += timedelta(days=1)
+
+ return [stats[d] for d in sorted(all_dates)]
+
+ def get_monthly_stats(self, months: int = 12) -> List[Dict]:
+ """Get monthly token usage for charts."""
+ results = db.session.query(
+ extract('year', TokenUsage.date).label('year'),
+ extract('month', TokenUsage.date).label('month'),
+ func.sum(TokenUsage.total_tokens).label('tokens'),
+ func.sum(TokenUsage.cost).label('cost')
+ ).group_by('year', 'month').order_by('year', 'month').all()
+
+ # Get last N months
+ monthly_data = [
+ {
+ 'year': int(r.year),
+ 'month': int(r.month),
+ 'tokens': r.tokens or 0,
+ 'cost': r.cost or 0
+ }
+ for r in results
+ ]
+
+ return monthly_data[-months:] if len(monthly_data) > months else monthly_data
+
+ def get_user_stats(self) -> List[Dict]:
+ """Get per-user token usage breakdown for current month."""
+ today = date.today()
+
+ results = db.session.query(
+ User.id,
+ User.username,
+ User.monthly_token_budget,
+ func.sum(TokenUsage.total_tokens).label('usage'),
+ func.sum(TokenUsage.cost).label('cost')
+ ).outerjoin(
+ TokenUsage,
+ (User.id == TokenUsage.user_id) &
+ (extract('year', TokenUsage.date) == today.year) &
+ (extract('month', TokenUsage.date) == today.month)
+ ).group_by(User.id).all()
+
+ return [
+ {
+ 'user_id': r.id,
+ 'username': r.username,
+ 'monthly_budget': r.monthly_token_budget,
+ 'current_usage': r.usage or 0,
+ 'cost': r.cost or 0,
+ 'percentage': ((r.usage or 0) / r.monthly_token_budget * 100)
+ if r.monthly_token_budget else 0
+ }
+ for r in results
+ ]
+
+ def get_today_usage(self, user_id: int = None) -> Dict:
+ """Get today's token usage."""
+ today = date.today()
+
+ query = db.session.query(
+ func.sum(TokenUsage.total_tokens).label('tokens'),
+ func.sum(TokenUsage.cost).label('cost')
+ ).filter(TokenUsage.date == today)
+
+ if user_id:
+ query = query.filter(TokenUsage.user_id == user_id)
+
+ result = query.first()
+
+ return {
+ 'tokens': result.tokens or 0,
+ 'cost': result.cost or 0
+ }
+
+
+# Singleton instance
+token_tracker = TokenTracker()
diff --git a/src/tasks/processing.py b/src/tasks/processing.py
index 6fe8a092..83ba1803 100644
--- a/src/tasks/processing.py
+++ b/src/tasks/processing.py
@@ -261,7 +261,9 @@ Title:"""
{"role": "user", "content": prompt_text}
],
temperature=0.7,
- max_tokens=5000
+ max_tokens=5000,
+ user_id=recording.user_id,
+ operation_type='title_generation'
)
raw_response = completion.choices[0].message.content
@@ -499,7 +501,9 @@ Summarization Instructions:
{"role": "user", "content": prompt_text}
],
temperature=0.5,
- max_tokens=int(os.environ.get("SUMMARY_MAX_TOKENS", "3000"))
+ max_tokens=int(os.environ.get("SUMMARY_MAX_TOKENS", "3000")),
+ user_id=recording.user_id,
+ operation_type='summarization'
)
raw_response = completion.choices[0].message.content
@@ -742,7 +746,9 @@ You must respond with valid JSON format only."""
],
temperature=0.2,
response_format={"type": "json_object"},
- max_tokens=3000
+ max_tokens=3000,
+ user_id=recording.user_id,
+ operation_type='event_extraction'
)
response_content = completion.choices[0].message.content
diff --git a/static/js/app.modular.js b/static/js/app.modular.js
index 79d5332b..1c25b435 100644
--- a/static/js/app.modular.js
+++ b/static/js/app.modular.js
@@ -116,6 +116,12 @@ document.addEventListener('DOMContentLoaded', async () => {
const isSidebarCollapsed = ref(false);
const searchTipsExpanded = ref(false);
const isUserMenuOpen = ref(false);
+ const tokenBudget = ref({
+ has_budget: false,
+ budget: null,
+ usage: 0,
+ percentage: 0
+ });
const isDarkMode = ref(false);
const currentColorScheme = ref('blue');
const showColorSchemeModal = ref(false);
@@ -439,7 +445,7 @@ document.addEventListener('DOMContentLoaded', async () => {
isLoadingMore, searchDebounceTimer,
// UI
- browser, isSidebarCollapsed, searchTipsExpanded, isUserMenuOpen, isDarkMode,
+ browser, isSidebarCollapsed, searchTipsExpanded, isUserMenuOpen, tokenBudget, isDarkMode,
currentColorScheme, showColorSchemeModal, windowWidth, mobileTab, isMetadataExpanded,
showSortOptions, currentLanguage, currentLanguageName, availableLanguages, showLanguageMenu,
colorSchemes, isMobileScreen, isMobileDevice,
@@ -558,6 +564,17 @@ document.addEventListener('DOMContentLoaded', async () => {
showToast(message, 'fa-exclamation-circle', duration, 'error');
};
+ const loadTokenBudget = async () => {
+ try {
+ const response = await fetch('/api/user/token-budget');
+ if (response.ok) {
+ tokenBudget.value = await response.json();
+ }
+ } catch (error) {
+ console.error('Error loading token budget:', error);
+ }
+ };
+
// Helper function to calculate global segment index in bubble view
const getBubbleGlobalIndex = (rowIndex, bubbleIndex) => {
if (!processedTranscription.value.bubbleRows) return 0;
@@ -782,7 +799,8 @@ document.addEventListener('DOMContentLoaded', async () => {
t, tc, setGlobalError, showToast, formatFileSize, formatDisplayDate, formatShortDate,
formatStatus, getStatusClass, formatTime, formatDuration, formatEventDateTime,
getDateForSorting, isToday, isYesterday, isThisWeek, isLastWeek, isThisMonth, isLastMonth, isSameDay,
- nextTick
+ nextTick,
+ onChatComplete: loadTokenBudget // Refresh token budget after chat
};
// =========================================================================
@@ -1851,7 +1869,8 @@ document.addEventListener('DOMContentLoaded', async () => {
await Promise.all([
recordingsComposable.loadRecordings(),
recordingsComposable.loadTags(),
- recordingsComposable.loadSpeakers()
+ recordingsComposable.loadSpeakers(),
+ loadTokenBudget()
]);
// Load config
@@ -2046,6 +2065,7 @@ document.addEventListener('DOMContentLoaded', async () => {
formatDateTime: formatEventDateTime, // Alias for recovery modal
setGlobalError,
showToast,
+ loadTokenBudget,
getContrastTextColor,
getBubbleGlobalIndex,
formatRecordingMode,
diff --git a/static/js/modules/composables/chat.js b/static/js/modules/composables/chat.js
index 33bef7a2..ad21fb43 100644
--- a/static/js/modules/composables/chat.js
+++ b/static/js/modules/composables/chat.js
@@ -9,7 +9,7 @@ export function useChat(state, utils) {
isChatLoading, chatMessagesRef, selectedRecording, csrfToken
} = state;
- const { showToast, setGlobalError } = utils;
+ const { showToast, setGlobalError, onChatComplete } = utils;
// Helper function to check if chat is scrolled to bottom (within bottom 5%)
const isChatScrolledToBottom = () => {
@@ -158,6 +158,9 @@ export function useChat(state, utils) {
return;
}
if (data.error) {
+ if (data.budget_exceeded) {
+ throw new Error('Monthly token budget exceeded. Please contact your administrator.');
+ }
throw new Error(data.error);
}
} catch (e) {
@@ -189,6 +192,10 @@ export function useChat(state, utils) {
if (isChatScrolledToBottom()) {
scrollChatToBottom();
}
+ // Refresh token budget after chat completion
+ if (onChatComplete) {
+ onChatComplete();
+ }
}
};
diff --git a/static/js/modules/composables/reprocess.js b/static/js/modules/composables/reprocess.js
index fa541d33..8b2e58ef 100644
--- a/static/js/modules/composables/reprocess.js
+++ b/static/js/modules/composables/reprocess.js
@@ -15,7 +15,7 @@ export function useReprocess(state, utils) {
currentlyProcessingFile, uploadQueue
} = state;
- const { showToast, setGlobalError } = utils;
+ const { showToast, setGlobalError, onChatComplete } = utils;
// Store for active polling intervals
const reprocessingPolls = new Map();
@@ -340,6 +340,8 @@ export function useReprocess(state, utils) {
if (statusData.status === 'COMPLETED') {
showToast('Processing completed!', 'fa-check-circle');
+ // Refresh token budget after LLM operations complete
+ if (onChatComplete) onChatComplete();
}
}
} else if (statusData.status === 'FAILED') {
diff --git a/static/js/modules/composables/speakers.js b/static/js/modules/composables/speakers.js
index a37ddece..5c6c0df7 100644
--- a/static/js/modules/composables/speakers.js
+++ b/static/js/modules/composables/speakers.js
@@ -19,7 +19,7 @@ export function useSpeakers(state, utils, processedTranscription) {
voiceSuggestions, loadingVoiceSuggestions
} = state;
- const { showToast, setGlobalError } = utils;
+ const { showToast, setGlobalError, onChatComplete } = utils;
// Current speaker highlight state
let currentSpeakerId = null;
@@ -401,6 +401,8 @@ export function useSpeakers(state, utils, processedTranscription) {
}
showToast('Summary updated!', 'fa-check-circle');
+ // Refresh token budget after LLM operation
+ if (onChatComplete) onChatComplete();
} else if (statusData.status === 'FAILED' || statusData.status === 'ERROR') {
// Stop polling if it failed
clearInterval(pollInterval);
@@ -765,6 +767,9 @@ export function useSpeakers(state, utils, processedTranscription) {
showToast('No speakers could be identified from the context.', 'fa-info-circle');
}
+ // Refresh token budget after LLM operation
+ if (onChatComplete) onChatComplete();
+
} catch (error) {
console.error('Auto Identify Speakers Error:', error);
showToast(`Error: ${error.message}`, 'fa-exclamation-circle', 5000, 'error');
diff --git a/static/js/modules/composables/upload.js b/static/js/modules/composables/upload.js
index e1696395..069210b8 100644
--- a/static/js/modules/composables/upload.js
+++ b/static/js/modules/composables/upload.js
@@ -18,7 +18,7 @@ export function useUpload(state, utils) {
const { computed, nextTick } = Vue;
- const { setGlobalError, showToast, formatFileSize } = utils;
+ const { setGlobalError, showToast, formatFileSize, onChatComplete } = utils;
// Compute selected tags from IDs
const selectedTags = computed(() => {
@@ -441,6 +441,8 @@ export function useUpload(state, utils) {
pollInterval.value = null;
resetCurrentFileProcessingState();
isProcessingActive.value = false;
+ // Refresh token budget after LLM operations
+ if (onChatComplete) onChatComplete();
startProcessingQueue();
return;
@@ -465,6 +467,8 @@ export function useUpload(state, utils) {
pollInterval.value = null;
resetCurrentFileProcessingState();
isProcessingActive.value = false;
+ // Refresh token budget after LLM operations
+ if (onChatComplete) onChatComplete();
startProcessingQueue();
return;
}
@@ -479,6 +483,8 @@ export function useUpload(state, utils) {
pollInterval.value = null;
resetCurrentFileProcessingState();
isProcessingActive.value = false;
+ // Refresh token budget after LLM operations
+ if (onChatComplete) onChatComplete();
startProcessingQueue();
return;
}
diff --git a/templates/admin.html b/templates/admin.html
index 6fed0422..a99bce22 100644
--- a/templates/admin.html
+++ b/templates/admin.html
@@ -15,6 +15,7 @@
<script src="{{ url_for('static', filename='vendor/js/tailwind.min.js') }}"></script>
<!-- All dependencies bundled locally for offline support -->
<script src="{{ url_for('static', filename='vendor/js/vue.global.js') }}"></script>
+<script src="https://cdn.jsdelivr.net/npm/[email protected]/dist/chart.umd.min.js"></script>
<!-- All dependencies bundled locally for offline support -->
<link rel="stylesheet" href="{{ url_for('static', filename='vendor/css/fontawesome.min.css') }}">
<link rel="stylesheet" href="{{ url_for('static', filename='css/styles.css') }}">
@@ -399,8 +400,124 @@
</div>
</div>
</div>
+
+ <!-- Token Usage Section -->
+ <div class="mt-6">
+ <h4 class="text-lg font-medium text-[var(--text-secondary)] mb-4">
+ <i class="fas fa-coins mr-2"></i> Token Usage Statistics
+ </h4>
+
+ <!-- Token Usage Summary Cards -->
+ <div class="grid grid-cols-1 md:grid-cols-2 lg:grid-cols-4 gap-4 mb-6">
+ <div class="bg-[var(--bg-tertiary)] p-4 rounded-lg border border-[var(--border-primary)] shadow-sm">
+ <div class="flex items-center">
+ <div class="p-3 rounded-full bg-[var(--bg-info-light)] text-[var(--text-info-strong)]">
+ <i class="fas fa-calendar-day text-xl"></i>
+ </div>
+ <div class="ml-4">
+ <p class="text-sm font-medium text-[var(--text-muted)]">Today's Tokens</p>
+ <p class="text-2xl font-semibold text-[var(--text-primary)]">${ tokenStats.today?.tokens?.toLocaleString() || 0 }</p>
+ </div>
+ </div>
+ </div>
+
+ <div class="bg-[var(--bg-tertiary)] p-4 rounded-lg border border-[var(--border-primary)] shadow-sm">
+ <div class="flex items-center">
+ <div class="p-3 rounded-full bg-[var(--bg-success-light)] text-[var(--text-success-strong)]">
+ <i class="fas fa-calendar-alt text-xl"></i>
+ </div>
+ <div class="ml-4">
+ <p class="text-sm font-medium text-[var(--text-muted)]">This Month</p>
+ <p class="text-2xl font-semibold text-[var(--text-primary)]">${ tokenStats.current_month?.tokens?.toLocaleString() || 0 }</p>
+ </div>
+ </div>
+ </div>
+
+ <div class="bg-[var(--bg-tertiary)] p-4 rounded-lg border border-[var(--border-primary)] shadow-sm">
+ <div class="flex items-center">
+ <div class="p-3 rounded-full bg-[var(--bg-warn-light)] text-[var(--text-warn-strong)]">
+ <i class="fas fa-dollar-sign text-xl"></i>
+ </div>
+ <div class="ml-4">
+ <p class="text-sm font-medium text-[var(--text-muted)]">Monthly Cost</p>
+ <p class="text-2xl font-semibold text-[var(--text-primary)]">$${ (tokenStats.current_month?.cost || 0).toFixed(4) }</p>
+ </div>
+ </div>
+ </div>
+
+ <div class="bg-[var(--bg-tertiary)] p-4 rounded-lg border border-[var(--border-primary)] shadow-sm">
+ <div class="flex items-center">
+ <div class="p-3 rounded-full" :class="tokenStats.users_at_100_percent > 0 ? 'bg-[var(--bg-danger-light)] text-[var(--text-danger-strong)]' : 'bg-[var(--bg-accent)] text-[var(--text-accent)]'">
+ <i class="fas fa-exclamation-triangle text-xl"></i>
+ </div>
+ <div class="ml-4">
+ <p class="text-sm font-medium text-[var(--text-muted)]">Budget Warnings</p>
+ <p class="text-2xl font-semibold text-[var(--text-primary)]">
+ <span v-if="tokenStats.users_at_100_percent > 0" class="text-[var(--text-danger-strong)]">${ tokenStats.users_at_100_percent } blocked</span>
+ <span v-else-if="tokenStats.users_over_80_percent > 0" class="text-[var(--text-warn-strong)]">${ tokenStats.users_over_80_percent } warning</span>
+ <span v-else class="text-[var(--text-success-strong)]">None</span>
+ </p>
+ </div>
+ </div>
+ </div>
+ </div>
+
+ <!-- Token Usage Charts -->
+ <div class="grid grid-cols-1 lg:grid-cols-2 gap-6">
+ <!-- Daily Usage Chart -->
+ <div class="bg-[var(--bg-tertiary)] p-4 rounded-lg border border-[var(--border-primary)] shadow-sm">
+ <h5 class="text-md font-medium text-[var(--text-secondary)] mb-3">Daily Token Usage (Last 30 Days)</h5>
+ <div class="h-64">
+ <canvas id="dailyTokenChart"></canvas>
+ </div>
+ </div>
+
+ <!-- Monthly Usage Chart -->
+ <div class="bg-[var(--bg-tertiary)] p-4 rounded-lg border border-[var(--border-primary)] shadow-sm">
+ <h5 class="text-md font-medium text-[var(--text-secondary)] mb-3">Monthly Token Usage (Last 12 Months)</h5>
+ <div class="h-64">
+ <canvas id="monthlyTokenChart"></canvas>
+ </div>
+ </div>
+ </div>
+
+ <!-- Per-User Token Usage -->
+ <div class="mt-6 bg-[var(--bg-tertiary)] p-4 rounded-lg border border-[var(--border-primary)] shadow-sm">
+ <h5 class="text-md font-medium text-[var(--text-secondary)] mb-3">User Token Usage (This Month)</h5>
+ <div class="space-y-3 max-h-64 overflow-y-auto">
+ <div v-for="user in tokenUserStats" :key="user.user_id" class="flex justify-between items-center p-2 bg-[var(--bg-secondary)] rounded-lg border border-[var(--border-primary)]">
+ <div class="flex items-center flex-1">
+ <i class="fas fa-user text-[var(--text-accent)] mr-2"></i>
+ <span class="text-sm font-medium text-[var(--text-primary)]">${ user.username }</span>
+ </div>
+ <div class="flex items-center space-x-4 flex-1">
+ <div class="flex-1">
+ <div class="w-full bg-[var(--bg-tertiary)] rounded-full h-2">
+ <div class="h-2 rounded-full"
+ :class="{
+ 'bg-[var(--text-danger-strong)]': user.percentage >= 100,
+ 'bg-[var(--text-warn-strong)]': user.percentage >= 80 && user.percentage < 100,
+ 'bg-[var(--text-success-strong)]': user.percentage < 80
+ }"
+ :style="{width: Math.min(user.percentage || 0, 100) + '%'}">
+ </div>
+ </div>
+ </div>
+ <div class="text-sm text-[var(--text-secondary)] text-right min-w-[120px]">
+ <span class="font-medium">${ user.current_usage?.toLocaleString() || 0 }</span>
+ <span v-if="user.monthly_budget" class="text-[var(--text-muted)]"> / ${ user.monthly_budget?.toLocaleString() }</span>
+ <span v-else class="text-[var(--text-muted)]"> (unlimited)</span>
+ </div>
+ </div>
+ </div>
+ <div v-if="tokenUserStats.length === 0" class="text-center text-sm text-[var(--text-muted)] p-2">
+ No token usage data available
+ </div>
+ </div>
+ </div>
+ </div>
</div>
-
+
<!-- System Settings Tab -->
<div v-show="activeTab === 'settings'">
<div class="flex justify-between items-center mb-4">
@@ -862,6 +979,11 @@ Summarization Instructions:
<span class="ml-3 text-sm text-[var(--text-secondary)]">${t('adminDashboard.adminUser')}</span>
</label>
</div>
+ <div class="mb-4">
+ <label class="block text-sm font-medium text-[var(--text-secondary)] mb-1">Monthly Token Budget</label>
+ <input v-model.number="newUser.monthlyTokenBudget" type="number" min="5000" step="5000" placeholder="Leave empty for unlimited" class="w-full px-3 py-2 border border-[var(--border-secondary)] rounded-md focus:outline-none focus:ring-2 focus:ring-[var(--border-focus)] bg-[var(--bg-input)] text-[var(--text-primary)]">
+ <p class="text-xs text-[var(--text-muted)] mt-1">Set a monthly token limit for AI features. Leave empty for unlimited.</p>
+ </div>
<div class="flex justify-end space-x-3">
<button type="button" @click="showAddUserModal = false" class="px-4 py-2 border border-[var(--border-secondary)] text-[var(--text-secondary)] rounded-md hover:bg-[var(--bg-tertiary)]">${t('buttons.cancel')}</button>
<button type="submit" class="px-4 py-2 bg-[var(--bg-button)] text-[var(--text-button)] rounded-md hover:bg-[var(--bg-button-hover)]">${t('adminDashboard.addUser')}</button>
@@ -869,7 +991,7 @@ Summarization Instructions:
</form>
</div>
</div>
-
+
<!-- Edit User Modal -->
<div v-if="showEditUserModal" @click.self="showEditUserModal = false" class="fixed inset-0 bg-black bg-opacity-70 backdrop-blur-sm flex items-center justify-center z-50 p-4">
<div class="bg-[var(--bg-secondary)] p-6 rounded-lg shadow-xl w-full max-w-md">
@@ -897,6 +1019,20 @@ Summarization Instructions:
<span class="ml-3 text-sm text-[var(--text-secondary)]">${t('adminDashboard.adminUser')}</span>
</label>
</div>
+ <div class="mb-4">
+ <label class="block text-sm font-medium text-[var(--text-secondary)] mb-1">Monthly Token Budget</label>
+ <input v-model.number="editingUser.monthly_token_budget" type="number" min="5000" step="5000" placeholder="Leave empty for unlimited" class="w-full px-3 py-2 border border-[var(--border-secondary)] rounded-md focus:outline-none focus:ring-2 focus:ring-[var(--border-focus)] bg-[var(--bg-input)] text-[var(--text-primary)]">
+ <p class="text-xs text-[var(--text-muted)] mt-1">Set a monthly token limit for AI features. Leave empty for unlimited.</p>
+ <div v-if="editingUser.monthly_token_budget && editingUser.current_token_usage !== undefined" class="mt-2">
+ <div class="flex justify-between text-xs text-[var(--text-muted)] mb-1">
+ <span>Current usage: ${ editingUser.current_token_usage?.toLocaleString() || 0 } tokens</span>
+ <span :class="{'text-[var(--text-danger-strong)]': editingUser.token_usage_percentage >= 100, 'text-[var(--text-warn-strong)]': editingUser.token_usage_percentage >= 80 && editingUser.token_usage_percentage < 100}">${ editingUser.token_usage_percentage || 0 }%</span>
+ </div>
+ <div class="w-full bg-[var(--bg-tertiary)] rounded-full h-2">
+ <div class="h-2 rounded-full" :class="{'bg-[var(--text-danger-strong)]': editingUser.token_usage_percentage >= 100, 'bg-[var(--text-warn-strong)]': editingUser.token_usage_percentage >= 80 && editingUser.token_usage_percentage < 100, 'bg-[var(--text-success-strong)]': editingUser.token_usage_percentage < 80}" :style="{width: Math.min(editingUser.token_usage_percentage || 0, 100) + '%'}"></div>
+ </div>
+ </div>
+ </div>
<div class="flex justify-end space-x-3">
<button type="button" @click="showEditUserModal = false" class="px-4 py-2 border border-[var(--border-secondary)] text-[var(--text-secondary)] rounded-md hover:bg-[var(--bg-tertiary)]">Cancel</button>
<button type="submit" class="px-4 py-2 bg-[var(--bg-button)] text-[var(--text-button)] rounded-md hover:bg-[var(--bg-button-hover)]">${t('adminDashboard.updateUser')}</button>
@@ -904,7 +1040,7 @@ Summarization Instructions:
</form>
</div>
</div>
-
+
<!-- Delete User Confirmation Modal -->
<div v-if="showDeleteUserModal" @click.self="showDeleteUserModal = false" class="fixed inset-0 bg-black bg-opacity-70 backdrop-blur-sm flex items-center justify-center z-50 p-4">
<div class="bg-[var(--bg-secondary)] p-6 rounded-lg shadow-xl w-full max-w-md">
@@ -1456,6 +1592,16 @@ Summarization Instructions:
failed_recordings: 0,
top_users: []
});
+ const tokenStats = ref({
+ today: { tokens: 0, cost: 0 },
+ current_month: { tokens: 0, cost: 0 },
+ user_count_with_usage: 0,
+ users_over_80_percent: 0,
+ users_at_100_percent: 0
+ });
+ const tokenUserStats = ref([]);
+ const dailyTokenChart = ref(null);
+ const monthlyTokenChart = ref(null);
const settings = ref([]);
const isLoadingUsers = ref(true);
const isLoadingSettings = ref(false);
@@ -1472,7 +1618,8 @@ Summarization Instructions:
email: '',
password: '',
confirmPassword: '',
- isAdmin: false
+ isAdmin: false,
+ monthlyTokenBudget: null
});
const editingUser = ref(null);
const userToDelete = ref(null);
@@ -1593,7 +1740,7 @@ Summarization Instructions:
throw new Error('Rate limit exceeded. Please wait a moment and try again.');
}
if (!response.ok) throw new Error('Failed to load settings');
-
+
const data = await response.json();
settings.value = data;
} catch (error) {
@@ -1603,7 +1750,153 @@ Summarization Instructions:
isLoadingSettings.value = false;
}
};
-
+
+ // Token Stats Functions
+ const loadTokenStats = async () => {
+ try {
+ // Load summary stats
+ const statsResponse = await fetch('/admin/token-stats');
+ if (statsResponse.ok) {
+ tokenStats.value = await statsResponse.json();
+ }
+
+ // Load user stats
+ const userStatsResponse = await fetch('/admin/token-stats/users');
+ if (userStatsResponse.ok) {
+ const data = await userStatsResponse.json();
+ tokenUserStats.value = data.users || [];
+ }
+
+ // Load and render charts
+ await loadTokenCharts();
+ } catch (error) {
+ console.error('Error loading token stats:', error);
+ }
+ };
+
+ const loadTokenCharts = async () => {
+ try {
+ // Load daily stats
+ const dailyResponse = await fetch('/admin/token-stats/daily?days=30');
+ if (dailyResponse.ok) {
+ const dailyData = await dailyResponse.json();
+ renderDailyChart(dailyData.stats);
+ }
+
+ // Load monthly stats
+ const monthlyResponse = await fetch('/admin/token-stats/monthly?months=12');
+ if (monthlyResponse.ok) {
+ const monthlyData = await monthlyResponse.json();
+ renderMonthlyChart(monthlyData.stats);
+ }
+ } catch (error) {
+ console.error('Error loading token charts:', error);
+ }
+ };
+
+ const renderDailyChart = (data) => {
+ const ctx = document.getElementById('dailyTokenChart');
+ if (!ctx) return;
+
+ // Destroy existing chart if any
+ if (dailyTokenChart.value) {
+ dailyTokenChart.value.destroy();
+ }
+
+ const labels = data.map(d => d.date);
+ const tokens = data.map(d => d.total);
+
+ // Detect dark mode
+ const isDark = document.documentElement.classList.contains('dark');
+ const textColor = isDark ? '#9ca3af' : '#6b7280';
+ const gridColor = isDark ? '#374151' : '#e5e7eb';
+
+ dailyTokenChart.value = new Chart(ctx, {
+ type: 'bar',
+ data: {
+ labels: labels,
+ datasets: [{
+ label: 'Tokens',
+ data: tokens,
+ backgroundColor: isDark ? 'rgba(99, 102, 241, 0.7)' : 'rgba(79, 70, 229, 0.7)',
+ borderColor: isDark ? 'rgba(99, 102, 241, 1)' : 'rgba(79, 70, 229, 1)',
+ borderWidth: 1
+ }]
+ },
+ options: {
+ responsive: true,
+ maintainAspectRatio: false,
+ plugins: {
+ legend: { display: false }
+ },
+ scales: {
+ x: {
+ ticks: { color: textColor, maxRotation: 45, minRotation: 45 },
+ grid: { color: gridColor }
+ },
+ y: {
+ beginAtZero: true,
+ ticks: { color: textColor },
+ grid: { color: gridColor }
+ }
+ }
+ }
+ });
+ };
+
+ const renderMonthlyChart = (data) => {
+ const ctx = document.getElementById('monthlyTokenChart');
+ if (!ctx) return;
+
+ // Destroy existing chart if any
+ if (monthlyTokenChart.value) {
+ monthlyTokenChart.value.destroy();
+ }
+
+ const monthNames = ['Jan', 'Feb', 'Mar', 'Apr', 'May', 'Jun', 'Jul', 'Aug', 'Sep', 'Oct', 'Nov', 'Dec'];
+ const labels = data.map(d => `${monthNames[d.month - 1]} ${d.year}`);
+ const tokens = data.map(d => d.tokens);
+
+ // Detect dark mode
+ const isDark = document.documentElement.classList.contains('dark');
+ const textColor = isDark ? '#9ca3af' : '#6b7280';
+ const gridColor = isDark ? '#374151' : '#e5e7eb';
+
+ monthlyTokenChart.value = new Chart(ctx, {
+ type: 'line',
+ data: {
+ labels: labels,
+ datasets: [{
+ label: 'Tokens',
+ data: tokens,
+ fill: true,
+ backgroundColor: isDark ? 'rgba(34, 197, 94, 0.2)' : 'rgba(22, 163, 74, 0.2)',
+ borderColor: isDark ? 'rgba(34, 197, 94, 1)' : 'rgba(22, 163, 74, 1)',
+ borderWidth: 2,
+ tension: 0.3
+ }]
+ },
+ options: {
+ responsive: true,
+ maintainAspectRatio: false,
+ plugins: {
+ legend: { display: false }
+ },
+ scales: {
+ x: {
+ ticks: { color: textColor },
+ grid: { color: gridColor }
+ },
+ y: {
+ beginAtZero: true,
+ ticks: { color: textColor },
+ grid: { color: gridColor }
+ }
+ }
+ }
+ });
+ };
+
const showEditSettingModal = ref(false);
const editingSetting = ref(null);
const settingNewValue = ref('');
@@ -1785,7 +2078,8 @@ Summarization Instructions:
username: newUser.value.username,
email: newUser.value.email,
password: newUser.value.password,
- is_admin: newUser.value.isAdmin
+ is_admin: newUser.value.isAdmin,
+ monthly_token_budget: newUser.value.monthlyTokenBudget || null
})
});
@@ -1800,7 +2094,8 @@ Summarization Instructions:
email: '',
password: '',
confirmPassword: '',
- isAdmin: false
+ isAdmin: false,
+ monthlyTokenBudget: null
};
showAddUserModal.value = false;
@@ -1824,9 +2119,10 @@ Summarization Instructions:
const payload = {
username: editingUser.value.username,
email: editingUser.value.email,
- is_admin: editingUser.value.is_admin
+ is_admin: editingUser.value.is_admin,
+ monthly_token_budget: editingUser.value.monthly_token_budget || null
};
-
+
// Only include password if it was changed
if (editingUser.value.password) {
payload.password = editingUser.value.password;
@@ -2598,6 +2894,7 @@ Summarization Instructions:
if (!isTeamAdminOnly.value) {
loadUsers();
loadStats();
+ loadTokenStats();
// Load settings first, then default prompt can use that data
if (activeTab.value === 'settings' || activeTab.value === 'prompts') {
@@ -2695,6 +2992,8 @@ Summarization Instructions:
activeTab,
users,
stats,
+ tokenStats,
+ tokenUserStats,
settings,
isLoadingUsers,
isLoadingSettings,
@@ -2768,6 +3067,7 @@ Summarization Instructions:
loadUsers,
loadStats,
loadSettings,
+ loadTokenStats,
editSetting,
saveSettingValue,
updateSetting,
diff --git a/templates/components/header.html b/templates/components/header.html
index 3b0d2e64..00555960 100644
--- a/templates/components/header.html
+++ b/templates/components/header.html
@@ -16,7 +16,9 @@
</div>
<!-- Right side: User menu and controls -->
- <div class="flex items-center gap-3">
+ <div class="flex items-center gap-2 sm:gap-3">
+ {% include 'components/token_budget_indicator.html' %}
+
<!-- Inquire Mode Button -->
{% if inquire_mode_enabled %}
<a href="/inquire"
diff --git a/templates/components/token_budget_indicator.html b/templates/components/token_budget_indicator.html
new file mode 100644
index 00000000..1f450a40
--- /dev/null
+++ b/templates/components/token_budget_indicator.html
@@ -0,0 +1,17 @@
+<!-- Token Budget Indicator Component -->
+<!-- Only shows if user has a budget limit set -->
+<div v-if="tokenBudget && tokenBudget.has_budget"
+ class="hidden sm:flex items-center gap-1.5 px-2 py-1 rounded text-xs text-[var(--text-muted)] hover:text-[var(--text-secondary)] transition-colors cursor-default"
+ :title="'Token usage: ' + (tokenBudget.usage || 0).toLocaleString() + ' / ' + (tokenBudget.budget || 0).toLocaleString()">
+ <i class="fas fa-coins text-[10px]"
+ :style="tokenBudget.percentage >= 100 ? {color: '#ef4444'} : tokenBudget.percentage >= 80 ? {color: '#f59e0b'} : {}"></i>
+ <span :style="tokenBudget.percentage >= 100 ? {color: '#ef4444'} : tokenBudget.percentage >= 80 ? {color: '#f59e0b'} : {}">${ tokenBudget.percentage }%</span>
+ <div class="w-12 h-1 bg-[var(--bg-tertiary)] rounded-full overflow-hidden">
+ <div class="h-full rounded-full transition-all duration-300 bg-[var(--text-accent)]"
+ :style="{
+ width: Math.min(tokenBudget.percentage, 100) + '%',
+ backgroundColor: tokenBudget.percentage >= 100 ? '#ef4444' : tokenBudget.percentage >= 80 ? '#f59e0b' : null
+ }">
+ </div>
+ </div>
+</div>
diff --git a/templates/inquire.html b/templates/inquire.html
index cdfbe1db..6d1ff9ae 100644
--- a/templates/inquire.html
+++ b/templates/inquire.html
@@ -149,9 +149,11 @@
</div>
<!-- Right side: User menu -->
- <div class="flex items-center gap-3">
+ <div class="flex items-center gap-2 sm:gap-3">
+ {% include 'components/token_budget_indicator.html' %}
+
<!-- New Recording Button -->
- <a href="/"
+ <a href="/"
class="px-3 py-1.5 bg-[var(--bg-button)] text-[var(--text-button)] rounded-lg hover:bg-[var(--bg-button-hover)] transition-colors text-sm flex items-center gap-2">
<i class="fas fa-plus"></i>
<span class="hidden sm:inline">${t('nav.newRecording')}</span>
@@ -434,6 +436,19 @@
<!-- Toast Container -->
<div id="toastContainer" class="fixed top-4 right-4 z-50 space-y-2"></div>
+
+ <!-- Global Error Banner -->
+ <div v-if="globalError" class="fixed top-16 left-1/2 transform -translate-x-1/2 z-50 max-w-lg w-full mx-4">
+ <div class="bg-red-600 text-white px-4 py-3 rounded-lg shadow-lg flex items-center justify-between">
+ <div class="flex items-center">
+ <i class="fas fa-exclamation-circle mr-3"></i>
+ <span class="text-sm font-medium">${ globalError }</span>
+ </div>
+ <button @click="globalError = null" class="ml-4 text-white hover:text-gray-200">
+ <i class="fas fa-times"></i>
+ </button>
+ </div>
+ </div>
<!-- Color Scheme Modal -->
<div v-if="showColorSchemeModal" class="color-scheme-modal" @click.self="closeColorSchemeModal">
<div class="color-scheme-modal-content">
@@ -604,6 +619,12 @@
const inquireMessagesRef = ref(null);
const globalError = ref(null);
const shouldAutoScroll = ref(true); // Track if we should auto-scroll
+ const tokenBudget = ref({
+ has_budget: false,
+ budget: null,
+ usage: 0,
+ percentage: 0
+ });
// --- Computed ---
const hasActiveFilters = computed(() => {
@@ -636,6 +657,17 @@
}
};
+ const loadTokenBudget = async () => {
+ try {
+ const response = await fetch('/api/user/token-budget');
+ if (response.ok) {
+ tokenBudget.value = await response.json();
+ }
+ } catch (error) {
+ console.error('Error loading token budget:', error);
+ }
+ };
+
const loadAvailableFilters = async () => {
try {
const response = await fetch('/api/inquire/available_filters');
@@ -723,6 +755,9 @@
} else if (data.end_of_stream) {
return;
} else if (data.error) {
+ if (data.budget_exceeded) {
+ throw new Error('Monthly token budget exceeded. Please contact your administrator for more tokens.');
+ }
throw new Error(data.error);
}
} catch (parseError) {
@@ -733,10 +768,11 @@
}
} catch (error) {
console.error('Error in inquire chat:', error);
- setGlobalError('Chat failed. Please try again.');
+ setGlobalError(error.message || 'Chat failed. Please try again.', 10000);
} finally {
isInquireChatLoading.value = false;
showProcessingStatus.value = false;
+ loadTokenBudget(); // Refresh token usage after chat
await nextTick();
scrollToBottom(); // Use smart scrolling
}
@@ -846,7 +882,8 @@
applyThemeFromStorage();
loadAvailableFilters();
-
+ loadTokenBudget();
+
// Add scroll listener for smart auto-scroll after DOM is ready
await nextTick();
if (inquireMessagesRef.value) {
@@ -872,6 +909,7 @@
isInquireChatLoading,
inquireMessagesRef,
globalError,
+ tokenBudget,
isUserMenuOpen,
isMobileSidebarOpen,
chatProcessingStatus,