What is Ragnerock?

Make every piece of raw data queryable and analyzable.

Ragnerock is a research intelligence platform that makes raw data (documents, spreadsheets, transcripts, filings) queryable and analyzable. Upload your data, define how it should be processed, and query the results with SQL.

Core Capabilities

Data Processing

Ragnerock ingests a wide range of file formats and processes them into searchable, analyzable units:

  1. Ingestion: Upload PDFs, Word documents, Excel spreadsheets, PowerPoint presentations, plain text, HTML, or Markdown
  2. Extraction: Text, tables, page boundaries, and document structure are extracted automatically
  3. Chunking: Content is split into semantic chunks optimized for search and analysis
  4. Embedding: Vector embeddings enable semantic search across your entire data library

Workflows & Operators

If you have a repeatable analysis process (conducting a risk assessment, evaluating an investment, checking contract terms, classifying documents), Ragnerock lets you operationalize that methodology with AI-powered workflows.

Operators define the outputs of your workflows. Each operator specifies:

  • JSON Schema: The exact table structure to produce (columns, types, constraints)
  • Generation Prompt: AI instructions for how to fill in each row
  • Scope: What unit of data to process

Operators come in two flavors:

  • Document-like: Operate on documents, pages, paragraphs, or sentences
  • Tabular: Operate on sheets or rows (for spreadsheet data)

Chain multiple operators into Workflows using the visual graph editor. Add conditional logic, error handling, and multi-stage processing. When a workflow runs, its outputs become durable, persisted tables.

SQL Query Interface

Every workflow output becomes a persistent, queryable table. The operator name is the table name, and the schema fields become columns. Use the Query Explorer to write SQL against your results:

SELECT document_name, sentiment, confidence, key_themes
FROM sentiment_analysis
WHERE confidence > 0.8
ORDER BY confidence DESC

Because workflow outputs are persistent tables, you can analyze across all of your processed data with standard SQL. Group by document, filter by confidence, join across operators, aggregate metrics.

Notebooks & Research Agent

Notebooks are Ragnerock’s interactive workspace, combining four cell types:

  • Chat cells: Conversations with the Research Agent
  • Query cells: SQL queries against your operator tables
  • Code cells: Python (executed by a Jupyter kernel)
  • Markdown cells: Documentation and notes

The Research Agent lives inside notebooks. It understands how your data is organized (from your workflows), what data is present in the system, and how it should be queried. It can search, analyze, write code, run queries, and synthesize insights, all with full citations back to source documents.

Provenance

Every data point maintains full provenance:

  • Source: The original document or data file
  • Location: The specific page, paragraph, or row
  • Operator: Which operator produced the result

This enables complete audit trails and verification of any result.

Integration Points

JupyterLab Integration

The Ragnerock JupyterLab extension provides a Research Agent sidebar alongside your notebook, with the ability to export agent responses as Python variables directly into your kernel.

Getting Started

  1. Quick Start: Upload data and start exploring in minutes
  2. Tutorials: Guided walkthroughs for common workflows
  3. Core Concepts: Understand data sources, annotations, and notebooks