Skip to main content
SYMI is currently in beta. Features, behavior, and configuration options may change as we continue to refine the platform. We recommend testing SYMI workflows in non-production environments and sharing feedback with your 4MINDS contact.

​
Overview

Screenshot 2026 05 28 At 2 44 13 PM
SYMI is a separate product layer built on top of the 4MINDS model platform. While the model platform handles knowledge ingestion, training, and inference, SYMI handles action — triggering automatically on incoming events, routing requests across your models, running scheduled jobs, and pushing outputs to external systems without manual intervention. SYMI runs on its own base model (Qwen 3.6) and can route requests to any 4MINDS model you have trained. It is designed for teams that want to automate multi-step workflows — not just answer questions, but take action based on what it finds.
SYMI is a separate product layer. You must have at least one trained 4MINDS model before configuring SYMI workflows. SYMI can query your models but cannot replace them.

​
Key components

SYMI is composed of six core components that work together to power agentic workflows:

​
Model Router

The Model Router controls which 4MINDS model SYMI uses to answer a query or complete a task. By default, SYMI uses its own Qwen 3.6 base model. You can configure the router to direct requests to any trained model in your workspace. SYMI can intelligently query across multiple models when instructed — for example, routing a security question to your SOC model while routing a product question to your general knowledge model. To configure the Model Router:
  1. Open SYMI from the main navigation.
  2. Navigate to Model Router settings.
  3. Select the default model for general queries.
  4. Configure per-sub-agent model routing for specialized workflows.

​
Connections

Connections link SYMI to external systems so it can receive triggers, read data, and push outputs.

​
Supported connections

Gmail is currently the only connection that supports automatic session triggering on inbound email. When a new email arrives in a Gmail inbox connected to SYMI, it can instantly trigger a workflow — no manual action needed.Outlook and Office 365 do not support inbound auto-trigger at this time. Teams using enterprise email via Outlook can work around this by configuring the Scheduler to poll the inbox at regular intervals (e.g. every 15 minutes).

​
Setting up a Gmail connection

Gmail auto-trigger is the recommended approach for event-driven agentic workflows. Example use case: a ServiceNow ticket is created → an automated email is sent to a SYMI-monitored Gmail inbox → SYMI reads the email, queries your knowledge graph, and responds with recommended next steps. Steps:
  1. In SYMI, navigate to Connections.
  2. Select Gmail and authenticate with the account you want SYMI to monitor.
  3. Configure the trigger condition (e.g. all inbound mail, or filtered by subject/sender).
  4. Link the trigger to a sub-agent or workflow.
Use a dedicated Gmail account for SYMI integrations — do not connect a personal or primary production inbox.

​
Sub-agents

Sub-agents are specialized agents you deploy within SYMI for specific tasks. Each sub-agent has:
  • A label — a name identifying its purpose (e.g. “RFP Responder”, “SOC Triage Agent”).
  • A task prompt — the instructions that define what the sub-agent does and how it should behave.
  • A model routing assignment — which 4MINDS model the sub-agent queries.
Sub-agent task prompts are where you define role-specific context for SYMI — for example, “You are a cybersecurity analyst working in a SOC. When given an alert, identify the affected system, classify the severity, and suggest a remediation step.” This is different from Personas, which control tone only. For system-level role definitions, always use the sub-agent task prompt.
To create a sub-agent:
  1. In SYMI, navigate to Sub-Agents.
  2. Click + New Sub-Agent.
  3. Enter a label, task prompt, and select the model to route to.
  4. Save and activate.

​
Example: RFP response workflow

A pre-built example of what a sub-agent workflow looks like end-to-end:
  1. SYMI monitors a Gmail inbox for emails with “RFP” in the subject line.
  2. On trigger, the sub-agent extracts the scope and requirements from the email body.
  3. SYMI queries the connected 4MINDS model (trained on your proposals and org knowledge) to generate a draft response.
  4. The draft is pushed to HubSpot for internal review and approval before any external send.

​
Scheduler

The Scheduler lets you run SYMI workflows on a recurring cadence without needing an inbound trigger. Jobs are configured using cron-style timing — hourly, daily, weekly, or a custom interval. Example use cases:
  • Run a threat intelligence summary every morning at 7 AM using your SOC model.
  • Poll an S3 bucket every few hours for new log events and produce an incident report.
  • Send a weekly digest of new HubSpot activity to a Slack channel.
To configure a scheduled job:
  1. In SYMI, navigate to Scheduler.
  2. Click + New Job.
  3. Define the task (either a sub-agent or a direct prompt).
  4. Set the schedule interval.
  5. Activate the job.

​
Reasoning Log (Constellation)

Every SYMI session generates a Reasoning Log — an audit trail of the steps SYMI took to arrive at its output. This is powered by Constellation, 4MINDS’ structured cognitive reasoning system. The Reasoning Log shows a summary-level view of SYMI’s thinking steps. It does not expose the full context window. What the Reasoning Log captures:
  • Which models were queried.
  • What sub-agents were invoked.
  • Key reasoning steps and decision points.
  • What was sent to external systems.
This log is useful for debugging workflows, verifying that the right model was used, and demonstrating reasoning transparency to stakeholders.

​
Slash commands

SYMI supports a library of slash commands in the Chat Window for quick access to platform controls during a session. The slash command toolkit covers: Type / in the SYMI Chat Window to browse available commands.

​
Example architecture: email-triggered agentic workflow

The following illustrates a complete event-driven workflow using SYMI’s Gmail auto-trigger:
This workflow runs entirely without human intervention once configured.

​
Frequently asked questions

Can SYMI query multiple models in a single session? Yes. You can configure sub-agents to route to different models, allowing a single SYMI workflow to pull from multiple knowledge sources — for example, your SOC model for security context and your General KB model for company policy. Can I use SYMI without a Gmail account? Yes, but auto-triggering on inbound email requires Gmail at this time. Outlook users can configure the Scheduler to poll their inbox at regular intervals as a workaround. Broader auto-trigger support for Outlook and other providers is on the roadmap. Is SYMI the right place to define role-specific context for my AI? Yes — for system-level role definitions like “You are a cybersecurity expert in a SOC”, use the sub-agent task prompt in SYMI. The Persona feature at the model level controls tone and communication style only, not role context. Does SYMI have access to my knowledge graph directly? SYMI queries your 4MINDS models, and those models are connected to your knowledge graph. SYMI does not access the Synthesis Graph™ directly — all knowledge retrieval happens through model inference endpoints.

​
Next steps

Add Integrations & Data Sources

Connect external data sources to the models SYMI will query

Explore the Model Dashboard

View and manage the models available for SYMI to route to

Personas

Understand the difference between Personas and SYMI sub-agent task prompts

Key Terms & Concepts

Explore definitions for SYMI, sub-agents, and agentic workflows