Research Agent

Use natural language to explore, analyze, and synthesize insights from your document universe.

The Research Agent is Ragnerock’s conversational AI assistant, purpose-built for document research. Unlike generic AI assistants, the Research Agent has deep access to your document library and annotations, with full provenance for every insight.

How It Works

The Research Agent combines:

  • Semantic search across your entire document library
  • Access to annotations and structured data you’ve extracted
  • Multi-step reasoning to synthesize complex insights
  • Full citations for every claim it makes

Using the Research Agent

In the Web Application

Navigate to your project in the Ragnerock web app and click “Research Agent” to start a conversation. The agent has full access to all documents and annotations in your project.

The conversational interface lets you:

  • Ask questions in natural language
  • Receive answers with citations to source documents
  • Follow up with clarifying questions
  • Export insights and data for further analysis

Capabilities

Ask the agent to find relevant documents across your library:

“Find all earnings calls from Q4 2024 where management discussed AI initiatives”

The agent searches semantically, understanding the meaning behind your query rather than just matching keywords.

Data Analysis

Leverage your annotations for quantitative insights:

“What’s the average sentiment score across tech sector earnings calls this quarter?”

The agent can query your annotation data and perform calculations.

Synthesis

Combine information from multiple sources:

“Compare Apple and Microsoft’s risk disclosures around supply chain issues”

The agent reads across documents, identifies relevant passages, and synthesizes a coherent response.

Code Generation

Request Python code to work with your data:

“Write a script to calculate year-over-year revenue growth for each company in our database”

The agent can generate code that uses the Ragnerock SDK and your annotation data.

Citations and Provenance

Every response from the Research Agent includes citations back to source documents. This is critical for:

  • Verification — Check the original source for any claim
  • Compliance — Maintain audit trails for investment decisions
  • Context — Understand the broader context around extracted insights

Citations include:

  • Document name
  • Page number (when applicable)
  • Relevant text excerpt

Notebook Integration

The Research Agent integrates with JupyterLab through the Ragnerock extension, allowing you to:

  1. Chat with the agent in a notebook sidebar
  2. Export responses as Python objects
  3. Continue analysis programmatically

This bridges the gap between natural language exploration and quantitative analysis.

Context Management

The agent maintains context throughout a conversation:

  1. First message — Establishes the topic or scope
  2. Follow-ups — Build on previous responses
  3. Context switching — Explicitly change topics when needed

For example:

  • “I’m analyzing Apple’s 10-K filing”
  • “What does it say about services revenue?” (uses context)
  • “Now let’s look at Microsoft instead” (switches context)

Best Practices

  1. Be specific — Precise questions get better answers
  2. Use follow-ups — Build on previous responses for deeper analysis
  3. Request citations — Ask “with citations” if provenance is important
  4. Iterate on complex queries — Break down multi-part questions into steps
  5. Export for analysis — Use the notebook integration for quantitative follow-up

Limitations

  • The Research Agent is available through the Ragnerock web application
  • Responses are based on documents and annotations in your project
  • Complex calculations may require exporting data and using Python directly

Next Steps

  • Learn about Documents to build your document library
  • Explore Annotations to extract structured data
  • Use the Python SDK for programmatic access to your data