> ## Documentation Index
> Fetch the complete documentation index at: https://docs.nebuly.com/llms.txt
> Use this file to discover all available pages before exploring further.

# What is Nebuly?

> Nebuly is the user analytics platform for Generative AI products, turning every user-AI interaction into structured, actionable insights.

Nebuly gives teams user analytics built specifically for Generative AI products. As enterprises deploy chatbots, copilots, and AI agents across customer-facing and internal tools, conversational interfaces create a critical visibility gap. Unlike traditional web or mobile apps where behavior is tracked through clicks and page views, unstructured dialogue is largely opaque. As a result, companies struggle to measure the ROI of their AI initiatives and pinpoint where their models are falling short.

Nebuly closes this gap by transforming every user-AI interaction into structured, actionable insights.

## Why Nebuly exists

Most teams shipping GenAI products operate in the dark. They know their AI is being used, but cannot see *how*, *why*, or whether it is delivering value. Traditional feedback mechanisms (thumbs up/down, surveys) are notoriously unreliable in conversational UX:

* **Coverage is tiny.** Less than **1% of interactions** receive explicit feedback.
* **Bias is extreme.** Users only rate when they are very satisfied or very frustrated, missing the vast moderate middle.
* **Click-stream analytics don't fit.** Tools like Amplitude or Pendo were built for buttons and pages, not natural language.

Nebuly treats **every interaction as feedback**. By analyzing 100% of conversations through a combination of explicit and implicit signals, the platform delivers a representative view of user experience without requiring active user participation.

## What Nebuly does

With Nebuly, product and business teams can answer questions like:

* What are users actually asking my AI agent?
* Which topics are growing in volume week over week?
* Are users satisfied with the responses they get?
* Is my AI driving real value — and how do I measure it?

Ingest conversation logs through a vendor-agnostic [Interaction API](https://docs.nebuly.com/tracking/api-reference/events/post-events-interaction-with-trace-v2) and have Nebuly automatically enrich them with structured analytics:

* **User intents and user actions**: what users are actually trying to accomplish.{/* Claude (per your request): renamed "tasks" entity to "user actions" */}
* **Topics**: themes and subjects users care about, automatically clustered.
* **Sentiment and emotion**: six sentiment levels and 29 distinct emotional states detected from psychological literature.
* **Implicit and explicit feedback**: signals of frustration, rephrasing, abandonment, or satisfaction.
* **Failure detection**: silent failures, hallucinations, off-topic requests, language gaps, unhandled queries.
* **Business risks**: interactions that need attention, from safety and compliance issues (PII exposure, toxic content, policy violations) to user complaints or dissatisfaction directed at your company, plus custom-defined risks.

Architecturally, the product is split into two complementary analytical layers, both accessible from the sidebar:

* **User Intelligence**: the human side. Who your users are, what they want, how they feel, how they segment.
* **Failure Intelligence**: the diagnostic side. Where the AI fails to deliver value, why, and how often.

## Who Nebuly is for

Nebuly is built for AI, product, and revenue teams operating GenAI products at enterprise scale. Typical users include:

* **Product managers** who need to know which features get adopted and which user goals go unmet.
* **AI / ML engineers** who need to debug silent failures and prioritize model improvements.
* **Customer success and revenue teams** who need to measure ROI per department, role, or region.
* **Compliance and security teams** who need to monitor PII leakage, toxic outputs, and policy violations.

## How Nebuly is different from observability tools

Nebuly is sometimes considered part of the broader LLM monitoring stack, but it is not a traditional observability or evaluation tool:

|                    | Traditional LLM Observability | Nebuly                                                  |
| ------------------ | ----------------------------- | ------------------------------------------------------- |
| **Data**           | Model logs and traces         | Real conversations and behavioral signals               |
| **Feedback model** | Explicit ratings + benchmarks | Implicit + explicit signals across 100% of interactions |

Nebuly does not provide or replace LLMs, it complements them with a user-centric control plane that turns conversations into product, growth, and risk decisions.

## Deployment

Nebuly is available in two deployment options from day one, **Cloud (SaaS)** and **Self-Hosted**, both built to enterprise-grade security and compliance standards. See [Nebuly deployment modes](/guides/saas-vs-self-hosted) for the full comparison.

## Where to go next

* Deploying Nebuly? Start with [Setting up Nebuly](/guides/getting-started).
* Already have an environment? Jump to [How to send data to Nebuly](/guides/data-ingest) or learn how to [navigate the platform](/guides/navigating-nebuly).
* Looking for analytical depth? Explore [Failure Intelligence](/guides/failure-intelligence) and [Reports](/guides/reports-guide).