dif
dropinfinops

The internals, not the pitch

Architecture diagrams, QB detection patterns, full deployment guide. Leave your email — we'll send you updates when new patterns ship.

One email per QB pattern release. No sales sequence.

dropinfinops documentation

Self-hosted multi-cloud cost intelligence. Deploy in minutes. Detect anomalies automatically. Ask questions in plain English.

I want to deploy the platform I want to extend the detection I want to understand the architecture I'm contributing to the pattern library
Last test run Run #11 · PASS ✓
Demo seed seed 99
QB patterns 11 validated
Detection tier Basic (FOCUS billing)
Cloud support AWS · Azure · GCP
Deploy region us-east-1 (POC)

System data flow

Billing Data CUR/Azure/GCP Processor Lambda + S3 Athena Glue + Parquet QB Engine 11 patterns Briefing S3 JSON daily Chat UI CloudFront + Bedrock Answers + Charts click any node for full details →

Documentation

🏗️
Architecture Overview
★ BSides demo asset
Interactive data flow diagram. The 3-layer detection model. Flywheel animation. Terraform module map.
architecture.html
🚀
Deployment Guide
★ BSides demo asset
Step-by-step with progress tracking. CloudFront invalidation. Lambda rebuild. Preflight and postflight verification.
deployment.html
🔍
QB Pattern Guide
★ BSides demo asset
Browse all 11 validated patterns. Filter by tier, cloud, category. View detection SQL. Add a new QB with the 5-step loop.
qb-patterns.html
⚙️
Data Generator Reference
Workload patterns, anomaly pool, threshold contracts, the seed system, and the focus_writer.py split-row mechanism.
datagen.html
💬
Chat LLM UI Guide
Mantle design tokens. Lambda handler and Bedrock invocation. CloudFront distribution. DynamoDB session history.
chat-ui.html
🤝
Contributing & Open Source
What's in the open-source QB pattern repo. Self-hosting quickstart. How to contribute a new detection pattern.
contributing.html

The DropInFinOps flywheel

Why the build sequence is a discipline

Every QB pattern ships as exactly three layers, in this order. The test run is the gate. If the QB does not fire, the article does not get written.

01
Briefing Builder Query
Defines "detected." The SQL IS the pattern definition. Build this first.
02
Data Generator
Synthetic FOCUS data built to match the real-world signature the query expects.
03
Test Run
Detection query must fire. This is the gate. No exceptions.
04
Score Wiring
score_collect.py + ground_truth_html.py. Evidence, not claims.
05
Article
Written ONLY after test passes. A capability claim backed by evidence.