Figure template · System & Network · 4:3
Layered Agent Communication from Semantics to Transmission
An architecture figure that splits agent communication into semantic interpretation, interaction protocol and data transmission layers, topped by user-agent, agent-agent and agent-environment scenarios. Use it as the overview figure of a multi-agent systems or protocol paper.
Prompt
Create a scientific figure about: Layered agent communication from semantics to transmission. Visual style: Clean Vector. Requirements: publication-ready scientific quality, clear labels and directional arrows where appropriate, avoid cartoon or stock-photo look, composition optimized for research papers and academic slides.
What this template is
A dashed frame at the top holds three scenario tiles: a user and an agent exchanging messages, several agents talking in a ring, and an agent calling tools, memory and knowledge. Beneath them sit three tilted platforms. L3 turns text, image and voice into task execution through relation extraction, context comprehension, intention identification and semantic parsing, then feeds planning, reasoning and reflection. L2 covers authentication, session management and communication mode definition. L1 shows envelopes travelling across a WAN and a LAN. Upward arrows join the layers, and a separate single-agent panel sits below. Readers should see that agent messaging, like networking, needs distinct transmission, protocol and meaning layers.
Elements to label
Scenario tiles
Pink, blue and green tiles for user, agent-to-agent and environment exchanges; rename them after your own application domain.
L3 semantic interpretation layer
Four processing steps such as intention identification; replace them with the semantic modules your system really has.
Planning, reasoning and reflection
Three icons at the right end of L3 that follow task execution; name the component that implements each one.
L2 interaction protocol layer
Authentication, session management and mode definition; name the actual protocol or message schema you use.
L1 data transmission layer
Envelopes cross a WAN and a LAN along a dashed path; state the transport or network type in the label.
WAN and LAN segments
Remote communication runs on the left of L1 and private communication on the right; name the protocol each segment uses.
Upward arrows
All three arrow groups point up; add a downward encapsulation path if the figure should show sending too.
Single-agent panel
Perception, memory, reasoning, tools and action in a bottom box; say how it maps onto the three layers above.
Where it is used
- Paper mechanism figure
- Defense slides
- Course handout
- Grant proposal
What reviewers look for
Basis for the layers
Reviewers will ask how the three layers relate to established network models, what interface sits between each adjacent pair, and why separate presentation or security layers are not needed in your design.
Both directions
Every inter-layer arrow points upward, which depicts only the receiving side. A complete link also needs the sender's top-down encapsulation; otherwise the stack reads as a one-way pipe. Add the reverse arrows.
The bottom panel
The single-agent panel has no connection to the stack. Expect a question about how perception and memory hand content to the semantic layer, or readers will treat it as an unrelated second figure.
Security
The protocol layer shows a single authentication icon. If agents talk over open networks, reviewers usually ask about encryption, permission control and protection against impersonation. One more icon may not be enough.
How to use it in Yila
- 1Pick a figure
- 2Describe your subject and the steps to stress
- 3Refine it in chat, then download the PNG
Tips
Do
- Name the modules and protocols of each layer in your prompt so the platforms carry your design, not generic terms.
- Ask for a top-down sending path so both sender and receiver appear in the stack.
- If space is short, ask Yila to drop the bottom panel and keep only the three layers and the scenarios.
Avoid
- Leaving the single-agent panel isolated; at least one arrow should show how it relates to the layers.
- Packing each platform with icons; labels longer than one line become hard to read at column width.
- Mixing the scenario colours into the layers below, which makes readers think each scenario owns a layer.
Questions
Is this layered architecture ready for my paper?
It serves as a draft overview. It is an AI-generated illustration whose layer names and module split come from a prompt, not from your system design. Substitute your own modules and protocols, add the sending direction, and have co-authors check that the terminology matches the rest of the manuscript.
Can I change the modules or icons in one layer?
Icons and text boxes cannot be dragged or rewritten individually, and the image is not a vector graphic. Ask in a Yila chat, for instance to add an encryption module to the protocol layer or a satellite link to the transmission layer, and the whole figure is regenerated with your change.
What format do I download?
An AI-generated PNG in landscape with the three platforms stacked from top to bottom. For vector artwork or editable text, use it as a layout guide and redraw it in illustration software, keeping the tilt and spacing of the platforms. Fonts can then match your other figures.
Does this template cost anything?
Browsing the page and sample costs nothing. Generating and revising in Yila consumes credits, and new accounts are given 2,500 credits at sign-up, which is plenty for trying out different layer designs before you commit to one. Viewing samples never uses credits.
My system has four layers. Can the figure show four?
Yes. Give each layer's name and responsibility and Yila rearranges the platforms to match. The more layers you add, the shorter each label should be, so move the detail into the caption and keep one or two words per module. Keep the scenario tiles at the top.
Can one communication scenario be emphasised?
Yes. If your paper only covers agent-to-agent collaboration, ask for the other two tiles to be removed and for the message types between agents, such as request, reply and broadcast, to be drawn in a close-up with one arrow style each. The layers below stay unchanged.





