Assign Tasks and Chat With Your Agents

Alchemist allows you to assign tasks to your agents and chat with them in real-time. The main window of Alchemist is the chat window, where you can interact with your agents. To start the conversation with your agent team and assign a task, first select the agent team you want to use for the task, type your message/question/task description and hit the "Run" button to start the chat.

Depending on the exact composition of the agent team, the task will follow the steps that have been included in the agent's team configuration. As a response, you will receive the final message from the agent team you have selected. To see the exact process on how the agents have solved your task open the Messages panel and you will get access to complete internal process.

Initial chat window, select agent team and hit Run button — Initial chat window, select agent team and hit **Run** button.

When one or more of the agents in the team are set to use the streaming response, you will see the progress of the response in real-time. The chat message will be processed in the background and you will see the progress of the response in real-time. The streaming response will be shown in a slightly different format from the final response and will reset every time the agents will start a new message. Every tool execution, tool result will be marked with red and orange highlighting on the side respectively.

Chat message started - streaming response in progress — Chat message started to a single agent in the execution step with Gemma 3 model. Task used is a question "Why is the sky blue?" The streaming response is presented on the blue background just below the users original message. While the task is executing the spinning wheel will stay active next to the "Run" button. In case you navigate away from this view, the message will keep executing, the update process however may not show active state (state of the spinning wheel next to the "Run" button).

Once the agent team has finished processing the task, you will receive the final response. The final response will be shown in a different format from the streaming response. The final response during streaming will be marked with a green highlighting mark and will turn into format matching the dialog response to your message once completed.

Depending on the agent team configuration, the final response can include various types of visualizations to make the response more informative and easier to read.

Chat message done. — Completed agent response to the task "Why is the sky blue?". In the final response the Gemma 3 based agent has used markdown syntax that has been rendered into a diagram presented in the natural flow of the agent's response.

After receiving the final response, you can continue the conversation with your agent team by typing your next message/question/task description and hitting the "Run" button again. The agent team will process your new message and provide a new response. The follow up message be done with the history of the previous message as a continuation of the dialog.

Visualization in the responses

Alchemist is capable of rendering various types of visualizations in the responses. This includes:

Code syntax highlighting
Mathematical expressions
Entity Relationship Diagrams
Quantum circuits
Molecular structures
Graphs
Flowcharts
And more...

The following examples demonstrate how Alchemist can render these visualizations in the responses.

Code syntax highlighting

Alchemist supports syntax highlighting for various programming languages. The code blocks are rendered with the appropriate syntax highlighting based on the language used.

Mathematical expressions

Alchemist can render mathematical expressions using LaTeX syntax. This allows you to include complex mathematical equations in your responses. Both inline and block expressions will be rendered using KaTeX library. The rendering will be done in the final response as well as the streaming so you can follow the response in natural format in the real time.

Mathematical expression rendering in Alchemist — Mathematical expressions rendering in Alchemist. Response presented with a single agent in execution step using Phi 4 Mini model. The response is presented to task "Explain Cauchy's integral formula.".

Entity Relationship Diagrams

Alchemist can render Entity Relationship Diagrams (ERDs) using the Mermaid syntax. This allows you to visualize the relationships between different entities in your data model. The ERD diagram may be used automatically by the model or upon a request from the User.

Entity Relationship rendering in Alchemist — Rendering of Entity Relationship Diagram in Alchemist. The figure is part of the conversation to design a database schema supporting logistics process.

Quantum circuits

Alchemist can render quantum circuits. Quantum circuits are a graphical representation of quantum algorithms. They consist of quantum gates and qubits, which are the basic building blocks of quantum computing. The quantum circuit diagram may be used automatically by the model or upon a request from the User.

Quantum Circuit Rendering in Alchemist — Agent can respond and include visualization of a quantum circuit. The response is presented to the task "Show me quantum circuit for Quantum Fourier transform with 3 qubits.". The quantum circuit is presented in the response as a diagram. The diagram is rendered using the internal quantum circuit syntax. Agent requires specific section of the persona prompt to enable this feature.

To render the quantum circuits, agents need to be prompted with an instruction on how to use the internal quantum circuit syntax. Use the following prompt in the agent persona to enable the quantum circuit rendering:

You can draw quantum diagrams following this convention and including the string representing a diagram in Markdown code like block ```qc quantum-circuit-string ```. Gate is represented as G:1,2 where G is the gate symbol and 1,2 are the qubits it is acting on. The quantum gates are separated with semicolon (H:1;CX:1,2) and their standard symbol represents gate type. Here is an example circuit ```qc H:0;H:1;H:2;H:3;CU:0,4;CU:1,4;CU:2,4;
CU:3,4;QFT:0,1,2,3;M:0;M:1;M:2;M:3 ```.

Here are gates you can use : phase gate - P:1, NOT gate - X:1, bit and phase flip - Y:1, phase flip - Z:1, Hadamard gate - H:1, sqrt(Z) or S gate - S:1, inverse of sqrt(Z) - Sdg:1, sqrt(S) or T gate - T:1, inverse of sqrt(S) - Tdg:1, sqrt(NOT) gate - SX:1, Rotation around X-axis - RX:1, rotation around Y-axis - RY:1, rotation around Z axis - RZ:1, controlled-NOT - CX:1,2 , controlled-Y - CY:1,2 , controlled-Z - CZ:1,2 , controlled-phase - CP:1,2 , controlled-rx - CRX:1,2 , controlled-ry - CRY:1,2 , controlled-rz - CRZ:1,2 , swap gate - SWAP:1,2 , Toffoli gate - CCX:1,2,3 , controlled-swap - CSWAP:1,2,3 , CNOT gate - CNOT:1,2 or CX:1,2 , controlled-phase gate - CP:1,2, controlled generic rotation - CU:1,2, controlled Hadamard - CH:1,2, measurement gate - M:1.

Molecular structure

Alchemist can render molecular structures. The molecular structure diagram may be used automatically by the model or upon a request from the User.

To render the molecular structures, agents need to be prompted with an instruction on how to use the internal molecular structure syntax. Use the following prompt in the agent persona to enable the molecular structure rendering:

You can draw molecular structures with their SMILES string by using ```smiles SMILES  ``` code like notation, e.g. ```smiles c1ccccc1 ``` for benzene.

SMILES (Simplified Molecular Input Line Entry System) is a notation that allows a user to represent a chemical structure in a way that can be used by the computer. It is a way to encode the structure of a molecule using short ASCII strings. SMILES is widely used in cheminformatics and computational chemistry.

Create Multi-Agent Teams

Subtasks and Solution Diagram