Quick Overview

Dashboard Tour

When you sign in at https://www.omnilink-agents.com, you land on the Dashboard. The left sidebar gives you access to every platform feature, organized into logical groups:

• Build — The main agent interface where you chat, test commands, and observe ToolRunner sessions.
• Configuration — System instructions, agent name, persona, and available commands.
• Knowledge — Upload documents, manage chunks, and configure retrieval settings.
• Agent Profiles — Create and switch between multiple agent configurations.
• Appearance — Customize the visual theme for embedded chat widgets.
• Voice — STT and TTS provider selection and voice configuration.
• Modes — Toggle operational modes that change agent behavior.
• AI Engine — Select which model powers your agent.
• Memory — View and manage short-term and long-term memory stores.
• Plans — Subscription tiers and billing management.
• Usage — Credit consumption, API call statistics, and cost breakdown.
• Connection — Agent server configuration and connection status.
• OmniLink API — API key management and endpoint reference.
• Heart — Real-time HTTP message visualization.
• Feedback — Submit and review platform feedback.
• Run Management — Monitor and control active ToolRunner sessions.
• Logs — Browse historical agent interactions and system events.

The Build Page

The Build page is where you interact with your agent in real time. It provides a chat interface that supports:

• Text-based conversation with the selected AI engine
• Voice input via the built-in STT integration
• Audio responses via TTS playback
• Live ToolRunner session output
• Command testing against the local OmniLinkEngine

Think of Build as your agent’s control room. You send messages, watch the AI reason, and see tool calls execute in real time. If a ToolRunner session is active, the Build page streams status updates, memory changes, and final analysis results directly into the conversation.

Configuration

The Configuration pane is where you define who your agent is and what it can do.

System Instructions

Write a system prompt that shapes the agent’s personality, knowledge boundaries, and behavioral rules. This prompt is injected at the start of every conversation.

Agent Name & Persona

Give your agent a name and persona description. These appear in the chat UI and are included in the system context so the model stays in character.

Available Commands

List the commands your agent can invoke. Each command is a template string that the OmniLinkEngine will match against user input:

# Examples of command templates
"turn [state:str] the [device:str]"
"set temperature to [temp:int] degrees"
"navigate to [location:str]"
"play [song:str] on [speaker:str]"

When the AI decides to call a tool, it emits a command string that matches one of these templates. The local runtime parses it, extracts the variables, and calls the registered handler.

Knowledge

The Knowledge page lets you build a retrieval-augmented generation (RAG) pipeline without writing any code.

Upload & Chunking

Upload PDF, TXT, DOCX, or Markdown files. OmniLink automatically splits them into chunks of configurable size. Each chunk is embedded using a high-quality vector model and stored for similarity search.

Retrieval

When your agent receives a question, the platform searches the knowledge base for the most relevant chunks and injects them into the prompt context. The AI answers grounded in your data, reducing hallucinations.

Management

From the Knowledge page you can:

• View all uploaded documents and their chunk counts
• Delete individual documents or clear the entire knowledge base
• Re-embed documents if you change the embedding model
• Preview individual chunks to verify quality

Agent Profiles

Agent Profiles let you maintain multiple distinct agents under a single account. Each profile has its own:

• Name and persona
• System instructions
• Command set
• Knowledge base
• Memory store
• AI engine selection
• Voice configuration

Switch between profiles instantly from the Dashboard sidebar. This is useful when you manage several agents — for example, a customer-support bot, a robotics controller, and a smart-home assistant — all from the same account.

Profiles can also be managed programmatically through the REST API:

from omnilink import OmniLinkClient

client = OmniLinkClient(omni_key="omk_your_key_here")

# Create a new agent profile
client.create_profile(
    name="HomeBot",
    persona="A friendly smart-home assistant",
    system_instructions="You control lights, locks, and thermostats.",
    engine="g1-engine"
)

Voice Settings

The Voice page configures both speech-to-text (STT) and text-to-speech (TTS) for your agent.

Speech-to-Text

Choose an STT provider and configure options such as language, model size, and noise suppression. STT is available through the /api/stt endpoint and the Dashboard’s built-in microphone button.

Text-to-Speech

Select a TTS provider, pick a voice, and adjust speed and pitch. The /api/tts endpoint accepts text and returns audio. The Dashboard plays responses automatically when TTS is enabled.

Voice settings are stored per agent profile, so each agent can have a distinct voice identity.

Modes

Modes modify how your agent processes input and generates responses. They act as behavioral presets that you can toggle on or off:

• Conversational — Standard chat behavior with full context.
• Command-Only — The agent only emits structured commands, no free-form text.
• Streaming — Responses are streamed token-by-token for low-latency UIs.
• Tool Mode — ToolRunner sessions are active and the agent can trigger local tool calls.

Modes can be combined. For example, you might enable both Streaming and Tool Mode for a robotics application that needs real-time feedback.

AI Engine Selection

OmniLink supports four AI engines, each optimized for different workloads:

You can change the engine from the Dashboard or via the API. Engine selection applies per profile and can be overridden per request by including the engine parameter in your API call.

Memory Management

OmniLink provides two tiers of agent memory:

Short-Term Memory

Conversation history within a single session. This is automatically managed and included in the prompt context. It gives the agent awareness of what was said earlier in the conversation.

Long-Term Memory

Persistent facts, preferences, and learned information that survive across sessions. Long-term memory is stored in the platform database and can be read and written through the /api/memory endpoint.

{
  "key": "user_preference_temperature",
  "value": "72 degrees Fahrenheit",
  "category": "preferences",
  "created_at": "2026-03-20T14:30:00Z"
}

The Memory page in the Dashboard lets you browse, search, edit, and delete memory entries. This is useful for debugging agent behavior or manually seeding knowledge.

Plans & Billing

OmniLink offers four subscription tiers:

All plans include access to the full REST API, Dashboard, and both the Python library and Web SDK. The difference is in credit allocation, number of agent profiles, and support tier.

Connection Settings

The Connection page configures the agent HTTP server that your local runtime uses to communicate with the platform.

Agent Server

By default, OmniLink connects to a local HTTP bridge at localhost:8080. You can change the host, port, and authentication credentials from this page.

Endpoints

OmniLink uses three primary HTTP endpoints:

• POST /command — Commands sent from the cloud AI to the local runtime.
• GET /feedback — Status updates and results sent from the local runtime back to the cloud.
• GET /context — Current agent context and state information.

Heart Visualization

The Heart page provides a real-time visualization of HTTP messages flowing through your system. Each message appears as a pulse, color-coded by endpoint. This is invaluable for debugging connectivity issues and verifying that commands are being delivered.

Next Steps

Now that you have a map of the platform, dive deeper:

• Use Cases — Explore real-world applications of OmniLink.
• Quickstart — Build and run your first agent in minutes.
• Architecture — Understand the full system design.