Skip to content

Intel Requests & Workspace

Every analytical task in Aletheia starts with an Intel Request (RFI). The RFI is the container for your work — it holds your question, your chat with the assistant, your files, and everything the system produces along the way: briefings, investigations, case findings, and decisions.

Creating an Intel Request

From the Analyst Workbench, navigate to Intel Requests and create a new one with a title and description. The title is your reference — keep it short and specific ("A320 fleet risk assessment", "Operator X sanctions check"). The description carries the full context of what you need to find out.

Once created, the RFI opens as a workbench page with three main areas:

  • Chat (left panel) — your conversation with the assistant. This is where you ask questions, create briefings, run investigations, and build the case.
  • Workspace (right panel) — a file tree of everything attached to this RFI: briefings, investigation results, case materials, and any files you or the assistant create.
  • Activity (status area) — lifecycle indicators for running investigations and decisions.

The Workspace

The workspace is a hierarchical folder of files scoped to the RFI. It serves two purposes: organizing the artifacts the system generates (briefings, investigation summaries) and storing anything you or the assistant create during analysis (tables, maps, notes, reports).

System-Generated Files

As you work, the system populates the workspace automatically:

Folder What goes here
Questions Briefing documents created from templates
Investigations Investigation summaries and results
Case Case board, key questions, findings
Product Decisions, reports, final deliverables

Files You Create

Ask the assistant to save anything to the workspace:

"Save that table as incidents-by-operator.csv in Analysis." "Write a summary of what we found so far."

The assistant shows you a preview before saving — you approve or reject each file operation. Supported formats include markdown, CSV, JSON, and the special .map and .graph formats for reloadable visualizations.

Uploading Files

Upload external files (PDFs, DOCX, XLSX, images) to the workspace through the file tree. The assistant can read uploaded files:

"Read the report I just uploaded."

Text-based formats are read directly; binary formats (PDF, DOCX, XLSX) are extracted to text automatically.

Profiles

Profiles scope the workbench to a specific operational context — for example, aviation or ciber. When a profile is active:

  • Only templates tagged with that profile (or with no profile) appear in the template picker
  • Only knowledge graphs relevant to the profile are connected
  • The assistant's behavior adapts to the domain context

Profiles are set at launch (ALETHEIA_PROFILE=ciber) and affect template visibility and connector routing. They don't restrict what you can ask — you can always query any connected graph regardless of profile.

The Analyst Workbench has three main areas:

  • Intel Requests (/intel-requests) — the list of all RFIs, with status and creation date
  • Briefings (/briefings) — the template management surface (for authors, not analysts)
  • RFI Workbench (/rfi/<id>) — the working page for a specific Intel Request

Inside the RFI Workbench, the workspace file tree shows lifecycle indicators on briefings and investigations — you can see at a glance whether something is in draft, running, or complete.

Under the Hood

Technical details
  • Each RFI gets its own database records in the workbench SQLite database and, when case-building starts, its own FalkorDB graph (case_rfi_{rfi_id}).
  • The workspace is served through the workbench API; files are stored in the workbench database, not on the filesystem.
  • The chat is powered by a LangGraph agent that routes between protocols (exploratory, elicitation, review, investigation, decision, case) based on the analyst's intent and the current state.
  • File reading caps at 100K characters; the truncated flag signals when content exceeds the limit.
  • File operations (write_file, edit_file, manage_workspace) are interactive — the assistant proposes the change and the UI shows an approval card before it takes effect.
  • Supported text formats for direct reading: .md, .csv, .txt, .json, .log, .tsv, .xml, .html, .graph, .map. Everything else goes through text extraction.

Learn More