In Hindu tradition, Chitragupta is the deity who maintains a complete record of every being's actions; the divine accountant himself. Fitting name for a system that tracks exactly who used what and how much it cost.
Multi-ecosystem infrastructure cost chargeback engine. Allocates costs to teams and service accounts across Confluent Cloud, self-managed Kafka, and any Prometheus-instrumented system. The goal is to support multiple ecosystems and custom cost allocation strategies. This was originally built for Confluent Cloud but has been extended to support other ecosystems.
Important
The v2 version is a complete rewrite from ground up for a full plugin architecture, multi-tenancy, FastAPI, proper storage layer with mitigations, emitter framework, docs site. Essentially an entirely new system with a lot more features and a much better performance profile. The goal is to keep adding more features and improvements as I go along and as more requests come in.
- 4x-10x faster performance for chargeback calculations and persistence compared to V1.
- Custom Tags support to allow ease of filters and chargeback grouping/aggregation.
- Full documentation website for ease of use.
- Pulls billing data from APIs or YAML cost models
- Discovers resources and identities using plugin specific implementations.
- Allocation strategies are now pluggable and can be customized for each SKU type.
- Evolving REST API for querying chargeback data and triggering pipeline runs.
- New UI (still in progress) that does need Grafana or external viewers.
- Multi emitter support for different output formats and more coming as needed/requested in the future.
- Topic attribution overlay — breaks Kafka cluster costs down to individual topics using Prometheus metrics (CCloud-only, requires a configured metrics source).
- Nascent support for Self Managed Kafka styled cost models.
- Config YAML format has changed substantially to support multiple ecosystems and custom cost allocation strategies.
- Plugin based architecture for adding new ecosystems, cost allocation strategies, emitters and more.
- Code now has internal persistence layer using SQLite(default) instead of in-memory cache.
- Prometheus metrics have been removed in favor of a database-backed retention.
- Grafana directly queries the database instead of Prometheus. No prometheus instance or script to write are needed anymore, yay!
| Ecosystem | Plugin | Billing Source |
|---|---|---|
| Confluent Cloud | confluent_cloud |
CCloud Billing API |
| Self-managed Kafka | self_managed_kafka |
YAML cost model + Prometheus |
| Generic metrics | generic_metrics_only |
YAML cost model + Prometheus |
git clone https://github.com/waliaabhishek/chitragupta.git
cd chitragupta/examples/ccloud-full
# Fill in your CCloud API credentials
cp .env.example .env
vim .env
# Start the full stack (API + Grafana + UI)
docker compose up -d- API: http://localhost:8080
- Grafana dashboards: http://localhost:3000 (admin / password)
- Frontend UI: http://localhost:8081
The Quickstart guide covers everything end-to-end: service account creation, permissions, API key setup, and running with Docker Compose. Three self-contained examples are available in examples/ — see ccloud-grafana/, ccloud-full/, or self-managed-full/.
AppSettings → WorkflowRunner → ChargebackOrchestrator
├── EcosystemPlugin
│ ├── ServiceHandler×N → CostAllocator
│ ├── CostInput
│ └── MetricsSource
├── StorageBackend
└── Emitter×N
Each tenant maps to one ecosystem plugin. The orchestrator runs a per-tenant, per-date pipeline: gather resources → resolve identities → fetch costs → allocate → store → emit. An optional topic attribution overlay stage (CCloud + Prometheus only) runs after chargeback calculation to attribute Kafka cluster costs to individual topics.
Full documentation is available here.
# Install with dev dependencies
uv sync --group dev
# Run tests
uv run pytest
# Lint and type check
uv run ruff check src tests
uv run mypy src- Python 3.14+
- uv package manager