MoongraphMoongraph

Tutorial: Research Investigation

A complete walkthrough of using Moongraph for document research — from upload to insight.

Tutorial: Research Investigation

This tutorial walks you through a complete research workflow in Moongraph. By the end, you'll know how to:

  • Upload documents and organize them
  • Build a knowledge graph from your documents
  • Explore entities and relationships visually
  • Ask questions and get cited answers
  • Trace claims back to source documents

Time: 30-45 minutes
Prerequisites: A Moongraph account with available credits

What you'll need

For this tutorial, gather 5-15 documents on a topic you want to investigate. Good examples:

  • Research papers on a specific subject
  • Company reports and filings
  • Legal documents or case files
  • Historical records
  • News articles about a topic

The documents should share some common entities (people, organizations, places) so the knowledge graph has connections to explore.

If you don't have documents ready, you can still follow along to learn the workflow. The screenshots and examples will help you understand what to expect.


Part 1: Upload your documents

Step 1: Navigate to Files

Click Files in the sidebar. You'll see the files table, which may be empty if you're just getting started.

Step 2: Create a folder

Organization helps later. Create a folder for this investigation:

  1. Click + Add in the top-right
  2. Select Create Folder
  3. Name it something descriptive (e.g., "Climate Research" or "Project Alpha Investigation")
  4. Click Create

Step 3: Upload documents

  1. Click into your new folder (or stay at the top level)
  2. Click + AddUpload Files
  3. Drag and drop your documents onto the dropzone, or click to select files
  4. Leave Smart Analysis enabled for best results
  5. Click Upload

Step 4: Wait for processing

Documents appear in the table with status indicators. Processing takes a few minutes depending on document size and complexity.

StatusMeaning
QueuedWaiting to start
ParsingExtracting text
EmbeddingGenerating search vectors
CompletedReady to use

You can continue to the next part once at least a few documents show Completed.


Part 2: Build a knowledge graph

Step 5: Navigate to Graphs

Click Graphs in the sidebar to see your knowledge graphs. If this is your first, the list will be empty.

Step 6: Create a new graph

  1. Click Create Graph in the top-right
  2. Enter a name (e.g., "Climate Research Graph")
  3. Optionally add a description

Step 7: Select documents

In the document selection panel:

  1. Click Select Folder and choose your folder, OR
  2. Click Select Documents and check individual documents

For this tutorial, include all documents from your investigation folder.

Step 8: Start graph creation

  1. Review your selection (document count shown)
  2. Click Create Graph

You'll be taken to the graph detail page.

Step 9: Monitor extraction

Graph creation has several stages:

StageWhat's happening
InitializingSetting up the extraction job
ExtractionFinding entities and relationships
ResolutionMerging duplicate entities
StorageSaving the graph

This can take 5-30 minutes depending on document count and size. You can leave this page and come back — processing continues in the background.

Step 10: Review the completed graph

Once complete, the graph detail page shows:

  • Overview: Entity and relationship counts, top connected entities
  • Entities tab: Browse all extracted entities
  • Relationships tab: Browse all relationships
  • Source Data tab: Documents included in the graph

Take a moment to browse. Notice what entity types were extracted and which entities appear most connected.


Part 3: Explore visually in Workbench

Now the real investigation begins. Workbench combines graph visualization with AI chat.

Step 11: Open in Workbench

From the graph detail page, click Open in Workbench (or navigate to Workbench in the sidebar and select your graph).

You'll see:

  • Graph canvas: Your entities as nodes, relationships as edges
  • Agent panel: Floating chat interface on the side

Step 12: Orient yourself

Spend a minute getting your bearings:

  1. Zoom out (scroll down) to see the whole graph
  2. Look for clusters — tightly connected groups often represent related entities
  3. Find hubs — nodes with many connections are often important

Step 13: Click a node

Click any node that looks interesting — perhaps a hub or one with a familiar name.

The Entity Drawer slides in, showing:

  • Entity name and type
  • Properties extracted from documents
  • Connected entities (click to navigate)
  • Source documents with page numbers

Step 14: Trace to source

In the Entity Drawer, under Source Documents:

  1. Click a document link
  2. The document opens at the page where this entity was mentioned
  3. Verify the information in context

This is provenance — the ability to trace any extracted information back to its source.


Part 4: Ask questions with the Agent

The Agent panel lets you ask questions while exploring.

Step 15: Ask your first question

In the Agent panel, type a question about your documents. Examples:

  • "Who are the main people mentioned in these documents?"
  • "What organizations are discussed?"
  • "What is the relationship between [Entity A] and [Entity B]?"

Press Enter or click Send.

Step 16: Watch tool execution

As the Agent works, you'll see tool pills appear:

  • retrieve_chunks — Searching documents
  • entity_search — Querying the graph
  • knowledge_graph_query — Finding connections

Click any pill to see what the Agent searched for and what it found.

Step 17: Read the answer and check citations

The Agent's response includes:

  • A synthesized answer based on your documents
  • Citation markers (e.g., [1], [2])

Click a citation to see the exact source chunk. Verify that the claim is supported by the source.

Step 18: Ask follow-up questions

Good investigations are iterative. Based on what you learned, ask follow-ups:

  • "Tell me more about [Entity X]"
  • "What documents mention [Topic Y]?"
  • "How is [Person A] connected to [Organization B]?"

Each answer builds on the context of your conversation.


Part 5: Investigate a connection

Let's put it all together with a focused investigation.

Step 19: Find two entities you want to connect

From your exploration, identify two entities you're curious about. For example:

  • A person and an organization
  • Two people
  • An organization and a location

Step 20: Search the graph

Use the search bar at the top of the graph canvas:

  1. Type the first entity's name
  2. Click the result to zoom to it
  3. Note its connections

Step 21: Ask about the connection

In the Agent panel, ask:

"What is the relationship between [Entity A] and [Entity B]? How are they connected?"

The Agent will use knowledge_graph_query to find paths between the entities and retrieve_chunks to provide supporting evidence.

Step 22: Explore the path

If the Agent finds a connection path, it might be:

  • Direct: A → B (they're directly connected)
  • Indirect: A → C → B (connected through intermediate entities)

Click on intermediate entities in the graph to understand the full chain.

Step 23: Verify with sources

For each claim in the Agent's answer:

  1. Click the citation
  2. Read the source chunk
  3. Click to open the full document if needed
  4. Confirm the information is accurate

Part 6: Generate a visualization

The Agent can create visual summaries of what you've learned.

Step 24: Request a diagram

Ask the Agent to visualize something you discovered:

  • "Create a flowchart showing the relationship between [Organization] and its subsidiaries"
  • "Make a timeline of events mentioned in these documents"
  • "Draw a diagram showing how [Person A] connects to [Person B]"

Step 25: Review the visualization

The Agent generates a diagram that renders directly in the chat. You can:

  • Zoom and pan the diagram
  • Toggle between rendered view and source code
  • Ask for modifications ("Add more detail to the timeline")

What you've learned

Congratulations! You've completed a full research investigation workflow:

SkillWhat you did
Document managementUploaded documents and organized them into folders
Graph creationBuilt a knowledge graph with entity extraction
Visual explorationNavigated the graph to find patterns and connections
AI-powered researchAsked questions and got cited answers
Provenance checkingTraced claims back to source documents
VisualizationGenerated diagrams to summarize findings

Next steps

Go deeper with your investigation

  • Add more documents to your graph (Graphs → Update)
  • Merge entities that should be the same (if you notice duplicates)
  • Export extraction results for analysis

Learn more

Get help

If something didn't work as expected:

On this page