Getting started with FormulAI

Learn how to navigate FormulAI, create and run governed AI use cases, review outputs, and manage workflows.

Enterprise Agent Platform — Build, run, and manage AI workflows with human-in-the-loop approval.


Table of Contents

  1. Roles
  2. First Steps
  3. Core Workflows
  4. Page Guide
  5. Tips & Tricks
  6. Troubleshooting

Roles

RolePermissionsUse Case
AdminFull access: models, tools, analytics, all workspacesPlatform setup, monitoring, configuration
OperatorCreate and edit use cases, run executions, read memory, export documentsBuild and run AI workflows
ReviewerRun executions, read memory, export documentsRun and check AI outputs

Your administrator assigns roles. Within a workspace, members also hold a workspace role (viewer, member, admin or owner) that controls what they can see and change there.


First Steps

1. Log In

  • Sign in with the account your administrator created for you. FormulAI ships with no default account or password.
  • On a new installation, the first administrator is created when the deployment is configured: the person installing supplies an admin email and a pre-hashed password as deployment secrets. Alternatively, choose Create your account on the sign-in page to register a user, which creates a workspace for that user.
  • Never write passwords in documents or chat.
  • After login, you’ll land on the Home page (Templates)

2. Navigate the Sidebar

The left sidebar provides access to all major sections:

IconSectionPurpose
🏠HomeTemplates overview
🤖Use CasesDefine AI tasks (prompt + model + tools)
⚡RunsRecent executions — retry, delete, view details
📖TemplatesReusable workflow templates
📄DocumentsAI-generated artifacts + uploads
✅ReviewsApproval queue (documents, tools, runs)
🔄WorkflowsMulti-step automated pipelines
📜HistoryFull-text search across all run history
🔗MemoryKnowledge base + embedding search
📊AdminPlatform metrics, models, providers (Admin only)
⚙️SettingsWorkspace config, members, document processing

3. Switch Language / Theme

  • Language: Click “RO” in the header → switch to “EN”
  • Theme: Click “Indigo” in the header → choose from 20+ color themes

Core Workflows

Workflow 1: Create a Use Case

A Use Case defines an AI task (e.g., “Summarize legal document”).

  1. Go to Use Cases → click “Create Use Case”
  2. Fill in:
    • Name and Description
    • System Prompt (instructions for the AI)
    • Model (select from configured models)
    • Tools (optional: API calls, scripts)
    • Input/Output schema (JSON structure)
    • Approval mode (auto, manual, or conditional)
  3. Click Save → the use case is ready to run

Workflow 2: Run a Use Case

  1. Go to Use Cases → find your use case
  2. Click “Run” → fill in input parameters
  3. The run appears in Runs with status: pending → running → completed/failed
  4. Click the run ID to view Run Detail:
    • Inputs and outputs
    • Step-by-step replay (click “Show replay”)
    • Generated document link (if applicable)

Workflow 3: Approve a Document

  1. Go to Reviews → switch to “Documents” tab
  2. Find the document awaiting approval
  3. Review the content → click “Approve” or “Reject”
  4. Approved documents appear in Documents with “approved” status

Workflow 4: Build a Workflow

A Workflow chains multiple steps together (e.g., ingest → transform → approve → export).

  1. Go to Workflows → click “Create Workflow”
  2. Add steps:
    • Run Use Case — execute an existing use case
    • Wait for Approval — pause until human approves
    • Condition — branch based on output values
  3. Set trigger (manual, schedule, or event)
  4. Save and run the workflow

Workflow 5: Search Run History

  1. Go to History
  2. Enter search terms (matches run inputs, outputs, status)
  3. Results show:
    • Run ID (clickable → Run Detail)
    • Status badge
    • Output preview
    • “Summarise” button (AI-generated summary)
    • “Export” button (download run bundle)

Workflow 6: Manage Memory (Knowledge Base)

  1. Go to Memory
  2. Memory entries: persistent knowledge snippets
    • Create new entries with content + metadata
    • Edit existing entries
    • Delete outdated entries
  3. Embedding Workbench: test semantic search
    • Enter a query → see which entries match
    • Adjust embedding parameters

Page Guide

Admin Section (Admin only)

PageWhat it shows
OverviewTotal runs, active workspaces, avg latency, failure rate, estimated cost
ModelsConfigured AI models (provider, endpoint, API key status)
ToolsRegistered tools (API calls, scripts, webhooks)
Run AnalyticsFailure trends, fallback stats, model performance
Provider HealthPer-provider latency, success rate, throughput
RetentionData retention policies
Run TraceDebug individual run execution traces
Cost TrendsCost breakdown by model/provider over time
ErrorsError log with filtering

Note: Metrics like latency and cost require the backend to collect model call data. If you see zeros, run some use cases first.

Settings Section

SectionWhat it does
WorkspaceName, description, members, roles
Document ProcessingInformational — shows supported formats, OCR config, agent tools
MembersAdd/remove workspace members, assign roles

Tips & Tricks

Keyboard Shortcuts

  • Escape: Close any modal dialog or dropdown
  • Tab: Navigate between interactive elements
  • Enter/Space: Activate buttons and links

Quick Actions

  • Retry a failed run: Go to Runs → click the failed run → click “Retry”
  • Export a document: Go to Documents → open the document → click format buttons (PDF, DOCX, etc.)
  • View run timeline: Open Run Detail → click “Show replay” to see step-by-step execution

Mobile

  • Tap the hamburger menu (☰) in the header to toggle the sidebar
  • The sidebar closes automatically when you tap outside it
  • All pages are responsive down to 375px width

Performance

  • The History page uses full-text search across all runs — use it for deep investigation
  • The Runs page shows recent runs only — use it for quick actions
  • Embedding search in Memory is vector-based — phrase queries naturally

Troubleshooting

“No runs found”

  • Runs are created when you execute a Use Case. Go to Use Cases → click “Run” on any use case.

“Avg Latency: 0ms” or “Est. Monthly Cost: $0.00”

  • These metrics require model call data. The backend collects this when runs execute. Run a few use cases and refresh.

AI buttons disabled in Document Editor

  • If the document status is “exported” or “approved”, editing is locked. Hover over the disabled button for details.

“No provider data available”

  • Provider health metrics are collected during run execution. Configure models in Admin → Models and run a use case.

Language not switching

  • The language preference is stored in localStorage. Try clearing cache or switching via the “RO” button in the header.

Can’t see the Admin section

  • The Admin section is only visible to Admins. Contact your workspace admin to elevate your role.

Need More Help?


FormulAI — Enterprise Agent Platform