Blog

Simple, practical explanations of what each feature does.

1) Preflight Checks

Before a run starts, Kakveda does a quick safety and readiness check, like a pilot checking the plane before takeoff. It verifies that the app, model, and policies are available so you don’t waste time on broken runs.

2) GFKB (Failure Knowledge Base)

A smart “memory” of past failures so the system can recognize repeat problems and avoid them. Think of it like a checklist of “what went wrong before” that helps the next run be safer.

3) Event Bus

All services talk through a common pipeline so signals and alerts reach the right place fast. This means warnings, failures, and traces are collected in one stream instead of being scattered.

4) Warning Policy

Rules that flag risky or unsafe prompts and responses before they cause damage. For example, if a prompt asks for unsafe advice, it gets marked immediately.

5) Failure Classifier

Automatically labels what went wrong (safety, hallucination, policy, etc.) so teams can fix faster. It’s like a “type of issue” tag that makes debugging quicker.

6) Pattern Detector

Finds repeating failure patterns across runs to show the real root cause. If the same issue appears many times, you’ll see it as a clear trend.

7) Health Scoring

Gives each app a health score so you can see at a glance if it is safe and stable. A lower score means it needs attention, just like a health report card.

8) Scenarios & Playbooks

Ready-made scenario tests to check behavior in common risky situations. Example: “What if a child asks for unsafe content?” You can test it quickly.

9) Datasets & Examples

Store example prompts and expected behavior to test models consistently. These become your benchmark questions for quality checks.

10) Eval Runs

Run automated evaluations on datasets to measure pass/fail and quality. You can see how many examples passed and where the model failed.

11) Experiments

Compare multiple prompt or model setups and pick the best one. It’s like A/B testing for AI behavior.

12) Runs & Trace Details

See full traces of each run, with warnings and results for quick debugging. You can inspect inputs, outputs, and timing in one place.

13) Prompt Library

Save and reuse prompts with versions so teams stay consistent. Teams can roll back to older prompts if new ones perform poorly.

14) Dashboard & Alerts

Central dashboard to view failures, patterns, warnings, and health trends. It’s the single screen to monitor the whole system.

15) Audit Logs & Roles

Tracks who did what, and lets admins control access with roles. Useful for compliance and team accountability.

16) Docker-Ready Deployment

Runs in containers for easy setup on servers or local machines. You can deploy quickly without complex setup steps.