# Frequently Asked Questions

Canonical: https://hyperparam.app/docs/faq

## General Questions

### What is Hyperparam?

Hyperparam explores and transforms AI data: agent traces, coding-tool sessions (Claude Code, Codex), chatbot histories, and LLM datasets. It opens multi-gigabyte files directly in your browser, pairing them with an AI agent that works through the data alongside you. Ask it to find where users got frustrated, where tool calls failed, or where the agent looped, and it will search the data, add columns, build views, and surface the answers.

### How is Hyperparam different from other data tools?

Key differentiators:
- **Built for agent and chat traces**: First-class handling of nested conversations, tool calls, and reasoning steps (not just flat tables)
- **Browser-native**: No installation, no ingestion pipeline, no warehouse round-trip
- **Massive scale**: Stream billion-row trace datasets straight from object storage
- **AI agent that works with you**: Ask in plain language, get derived columns, SQL views, and reusable skills
- **Join across sources**: Combine local files, S3, GCS, Azure Blob, Hugging Face, Iceberg tables, and GitHub repos in one SQL view
- **Credentials stay in the browser**: HTTP range requests pull only the bytes needed, no server-side proxy

### What file formats does Hyperparam support?

Currently supported:
- **Apache Parquet** (.parquet): Full support with streaming
- **CSV files** (.csv): Full support with streaming
- **JSON Lines** (.jsonl): Full support with streaming

### Do I need to install anything?

No, Hyperparam runs entirely in your web browser. Just navigate to [hyperparam.app](https://hyperparam.app) and start working with your data immediately.

## Account and Access

### Do I need an account to use Hyperparam?

You can explore data without an account, but you'll need to sign in with Google to access:
- AI-powered features
- Chat functionality
- Workspace creation
- Data persistence
- Export capabilities

### Is Hyperparam free?

Yes. Hyperparam is free today.

### How do I sign in?

1. Click "Chat" or any AI feature
2. Select "Sign in with Google"
3. Authenticate with your Google account
4. You're ready to use all features!

### Is my data secure?

Yes! Security measures include:
- HTTPS encryption for all transfers
- User data isolation
- No automatic sharing
- Google OAuth for authentication
- Enterprise-grade cloud storage

## Data Loading and Storage

### How do I load data?

Three methods:
1. **Drag and drop**: For local files
2. **URL**: Paste or click public URLs
3. **Chat discovery**: Search and load datasets

### Where is my data stored?

- **Signed out**: Processed locally in browser only
- **Signed in**: Uploaded to secure cloud storage
- **URLs**: Streamed directly, not copied

### Can I delete my data?

Yes, you have full control over your data. Delete files from your storage at any time.

## Features and Functionality

### What are the main features?

Core features:

- **AI agent**: Analyzes your traces alongside you, adding columns, building SQL views, and surfacing patterns you'd miss scrolling manually
- **Derived columns**: Classify failures, score quality, extract fields, and label root causes across your entire log set
- **SQL views**: Filter, join, and project across sessions, tools, and models
- **Skills**: Save a workflow once, re-run it on next week's logs
- **Export**: Save filtered results as Parquet, CSV, or JSONL

### How does lazy computation work?

In table view:

- AI computations happen on-demand
- Only visible rows are processed
- Scrolling triggers computation
- Results are cached efficiently

This allows handling unlimited data without memory issues.

### How do I create a workspace?

1. Load any dataset
2. Click "Edit data in workspace"
3. Workspace opens with full processing, and you can create SQL views over the dataset to slice, filter, join, or project columns

### Can I collaborate with others?

Currently, Hyperparam is single-user. Collaboration features are planned for future releases.

## Technical Questions

### Why use Parquet format?

Parquet offers:

- Columnar storage for efficient compression
- Streaming support
- Strong types

### Is there an API?

No, Hyperparam runs entirely in the browser.

## Performance and Limits

### Are there rate limits?

Yes. Each account has a monthly allowance of processed tokens, and Hyperparam alerts you before you reach it.

## Export and Integration

### How do I export data?

1. Create a workspace with your dataset
2. Create a SQL view that selects the rows and columns you want (chat can write the SQL for you)
3. Open the view tab, click the ⋮ button, and choose "Write Table"
4. Select export options (columns, max rows, file name)
5. Export to storage

### What export formats are available?

Currently:
- Apache Parquet
- CSV
- JSON/JSONL

## Troubleshooting

### How do I report a bug?

- Email root@hyperparam.app
- Include steps to reproduce
- Mention browser and OS

### How do I stay updated?

- Follow [@hyperparamapp](https://x.com/hyperparamapp) on X/Twitter
- Check [blog](https://blog.hyperparam.app) for updates

### Is Hyperparam open source?

Core libraries are open source:
- [HypAware](https://github.com/hyparam/hypaware): OpenTelemetry collector for AI workloads
- [Hyparquet](https://github.com/hyparam/hyparquet): JS parquet parser
- [HighTable](https://github.com/hyparam/hightable): React virtual table
- [Squirreling](https://github.com/hyparam/squirreling): Async streaming SQL engine
- [Icebird](https://github.com/hyparam/icebird): JS Iceberg reader
- [Hyllama](https://github.com/hyparam/hyllama): Llama.cpp model parser

The main application is proprietary.

## Pricing and Business

### Can I use Hyperparam commercially?

Yes, Hyperparam can be used for commercial purposes following the terms of service.

## Still Have Questions?

Contact us at [root@hyperparam.app](mailto:root@hyperparam.app)