bonanza-observe 0.1.0

AI agent cost observability — track spending, latency, and token usage across all your agent sessions

📊 Bonanza Observe

AI agent cost observability — track spending, latency, and token usage across all your agent sessions

Stop guessing how much your AI agents cost. Bonanza Observe gives you real-time cost tracking, token usage analytics, and latency monitoring for every LLM call.

Why?

AI agents make dozens of LLM calls per task. Without observability:

• You don't know what you're spending until the bill arrives
• You can't optimize what you can't measure
• You can't budget without tracking

Bonanza Observe is the missing open() for AI agent economics.

Installation

pip install bonanza-observe

Quick Start

from bonanza_observe import Tracker

# Create a tracker with a budget
tracker = Tracker(name="my-agent", budget_usd=10.00)

# Start tracking
session = tracker.start_session("research-task")

# Track LLM calls
span = tracker.start_span("web-search")
tracker.record_cost(
    provider="openai",
    model="gpt-4o",
    input_tokens=1500,
    output_tokens=800,
    latency_ms=1200,
)
tracker.finish_span()

# Check your budget
print(tracker.check_budget())
# {
#   "total_usd": 0.01175,
#   "budget_usd": 10.00,
#   "remaining_usd": 9.98825,
#   "percentage": 0.12,
#   "over_budget": False,
#   "alert": False
# }

# Get a full summary
print(tracker.summary())
# {
#   "total_cost_usd": 0.01175,
#   "total_tokens": 2300,
#   "by_model": {"gpt-4o": {"cost": 0.01175, "calls": 1}},
#   "by_provider": {"openai": {"cost": 0.01175, "calls": 1}},
#   ...
# }

Features

💰 Cost Tracking

• Per-model, per-provider cost breakdown
• Auto-calculates cost from token counts (supports GPT-4o, Claude 3.5, Gemini, Llama, and more)
• Manual cost override for custom pricing
• Budget alerts at configurable thresholds (default 80%)

📊 Token Usage

• Input/output/total token tracking per call
• Aggregate stats across sessions
• Per-model token breakdown

⏱️ Latency Monitoring

• Per-call latency tracking
• Average latency across sessions
• Nested span timing

📁 Session Management

• Group calls into sessions and spans
• Nested spans for complex agent workflows
• Tags for filtering and grouping

📤 Export

from bonanza_observe import Tracker, JsonExporter, CsvExporter

tracker = Tracker(name="my-agent")
# ... track calls ...

# Export as JSON
json_str = JsonExporter.export(tracker)
JsonExporter.save(tracker, "costs.json")

# Export as CSV
csv_str = CsvExporter.export(tracker)
CsvExporter.save(tracker, "costs.csv")

Supported Models (Pricing)

Model	Input (per 1M tokens)	Output (per 1M tokens)
GPT-4o	$2.50	$10.00
GPT-4o Mini	$0.15	$0.60
GPT-4 Turbo	$10.00	$30.00
Claude 3.5 Sonnet	$3.00	$15.00
Claude 3 Haiku	$0.25	$1.25
Claude 3 Opus	$15.00	$75.00
Gemini 1.5 Pro	$1.25	$5.00
Gemini 1.5 Flash	$0.075	$0.30
Llama 3.1 70B	$0.60	$0.60
Llama 3.1 8B	$0.05	$0.05

Comparison

Feature	Bonanza Observe	Revenium	Helicone	LangSmith
Cost tracking	✅	✅	✅	✅
Token tracking	✅	✅	✅	✅
Latency monitoring	✅	✅	✅	✅
Budget alerts	✅	✅	❌	✅
Session/span grouping	✅	❌	❌	✅
JSON/CSV export	✅	✅	✅	✅
Nested spans	✅	❌	❌	❌
Auto cost calculation	✅	✅	✅	❌
Zero dependencies	✅	❌	❌	❌
Self-hosted	✅	✅	❌	❌
Python-native	✅	❌	✅	✅

Requirements

• Python 3.10+

License

Apache License 2.0 — see LICENSE for details.

Links

• Website: bonanza-labs.com
• x402 Adapter: pypi.org/project/bonanza-x402
• MCP Server: pypi.org/project/bonanza-mcp
• Guard: pypi.org/project/bonanza-guard
• GitHub: github.com/c6zks4gssn-droid/bonanza-labs-website

---

Built by Bonanza Labs 📊