Overview

Outlines GLIK's execution models to help enterprises choose the right app type for scalable automation, policy engines, conversational agents, and developer-controlled orchestration.

GLIK offers four foundational app types, each representing a distinct execution model for building and deploying intelligent systems. App types define how user input is handled, how state is managed, and which blocks and system features are available.

This guide helps teams understand when to use each type, based on enterprise scenarios such as internal approvals, client-facing automation, agent workflows, or memory-aware assistants.

---

🧩 GLIK App Types at a Glance

App Type	Execution Model	Ideal For
Chatbot	Stateless, prompt-response logic	Basic Q&A bots, static form handlers
Agent	Stateful with scoped memory	Policy engines, multi-step enterprise flows
Advanced Chat	Threaded, turn-based interaction	Assistants, advisors, conversation memory
Workflow	Graph-based logic with branching	Conditional processes, escalation, integrations

Use Case Breakdown & Guidance

1. Chatbot

Model: Stateless input/output Best For:

• Simple support bots (FAQs, policy links)

• Front-line triage in internal tools

• Single-turn verification bots (e.g., invoice status check)

Limitations:

• No persistent state or scoped memory

• Cannot use advanced memory or fallback blocks

• Not suitable for multi-step processes or regulated logic flows

---

2. Agent

Model: Stateful with memory and policy enforcement logic Best For:

• Expense approval agents

• HR onboarding logic

• Compliance evaluations

• Vendor risk screening

Features:

• Scoped memory (invoice_data, policy_memory, etc.)

• Full access to Variable Assigners, IF/ELSE, Tool Nodes

• End-to-end workflow continuity

Considerations:

• Does not use threaded chat interface

• LLM blocks must be explicitly invoked as steps

• Memory must be manually written/read using memory blocks

---

3. Advanced Chat

Model: Conversational memory and multi-turn assistant Best For:

• Executive assistants

• Internal helpdesk bots

• SOP navigators

• Contract review advisors

Features:

• Supports conversation_variables natively

• Allows fallback memory injection into prompts

• Multi-turn interaction flows without defining graph logic

Caveats:

• Cannot chain logic using IF/ELSE or Tool blocks

• Not ideal for deterministic policy enforcement

• Suited for reasoning and conversational exploration, not decision routing

---

4. Workflow

Model: Visual flow with branching logic and tool integrations Best For:

• Escalation frameworks (e.g., “if fails → reroute”)

• Multi-agent orchestration

• Cross-system automation (API → plugin → memory)

• Human-in-the-loop decision gates

Features:

• Full block access: LLM, Tool, Data Enrichment, Save Point, Knowledge

• Conditional logic supported at multiple branches

• Integrates with external APIs and internal memory stores

Caveats:

• Requires upfront logic design

• Best suited for technical users or implementation partners

• Works best with deterministic or well-scoped problem domains

---

Memory & Variable Support Summary

Feature	Chatbot	Agent	Advanced Chat	Workflow
Conversation Variables	✖	✖	✅	✖
Scoped Memory (e.g. session)	✖	✅	✅	✅
Tool Node / Plugin Access	✖	✅	✖	✅
Conditional Routing	✖	✅	✖	✅
Visual Flow Design	✖	✖	✖	✅

Choosing the Right App Type

Scenario	Recommended App Type
Automating policy approval workflows	Agent or Workflow
Building an internal compliance chatbot	Advanced Chat
Creating a no-memory Q&A interface	Chatbot
Integrating multiple decision layers with routing	Workflow
Designing a memory-enabled SOP navigator	Advanced Chat
Embedding policy logic into a partner SaaS tool	Agent