Welcome to Agentspace
Agentspace is your dedicated area for managing, building, and exploring AI agents on the UBOS platform.
Here you can:
- Discover new agent features
- Access guides and documentation
- Start building your own agents
Stay tuned for more updates and resources!
Why Agentspace?
Challenge Today | Impact | Agentspace Remedy |
---|---|---|
Ad‑hoc prompts + scripts | Slow iteration, hard to share | Structured templates & onboarding |
Re‑inventing capabilities | Wasted cycles | Reusable Topics & Tools library |
Hard to teach or evaluate agents | Unpredictable quality | Built-in educational + assessment tools (Teacher) |
Prototype → Production gap | Experiments stall | Roadmap to APIs, orchestration, metering |
Agentspace collapses scattered workflows into a single lifecycle hub.
Core Concepts
Concept | Description | Analogy |
---|---|---|
Agent | Configured AI persona with identity + abilities | A staffed role |
Topic | A job domain that groups related tools & instructions | Department / function |
Tool | Executable action (explain, summarize, quiz, search) | A skill or instrument |
Template | Reusable blueprint for a class of agents | Playbook / starter kit |
Reasoning Loop (forthcoming) | Plan → select topic → invoke tools → refine | Workflow engine |
How It Works (High Level Flow)
- You start in the Onboarding Flow: name, purpose, audience, tone.
- Select or adapt a Template (e.g., Teacher Agent).
- Attach Topics (Teaching, Knowledge Search, Support).
- Each Topic exposes Tools the agent can call.
- Interact via chat; refine prompts and metadata.
- (Upcoming) Track usage, quality signals, and promote to production endpoints.
User Prompt ─▶ Agent Persona ─▶ Topic Selection ─▶ Tool Invocations ─▶ Response
▲ │ │
└──── Template / Instructions ───┘
The Teacher Agent (Flagship Template)
A pedagogically structured template showcasing the learning loop:
Tool | Purpose | Example Input | Example Output Key |
---|---|---|---|
explain_concept | Adaptive explanation by audience level | {"topic":"Photosynthesis","audience":"10-year-old"} | explanation |
summarize_text | Condense long passages into bullets / paragraph | {"text":"...","format":"bullet points"} | summary |
quiz_me | Generate formative assessment | {"topic":"Roman History","difficulty":"hard","num_questions":3} | quiz[] |
Loop: Explain → Reinforce (summarize / restate) → Assess (quiz) → Iterate.
Future extensions: spaced repetition, misconception detection, mastery scoring.
Other Early Use Cases
Use Case | Description | Starter Topic |
---|---|---|
Knowledge Navigator | Semantic + structured KB search | Knowledge Search |
Support Assistant | FAQ + case triage + summarization | Support Ops |
Research Synthesizer | Multi-source compare & merge summaries | Research |
Domain Coach | Role-specific micro-coaching (sales, compliance) | Coaching |
Standard vs. Custom
Layer | Standard (Built-In) | Customizable |
---|---|---|
Templates | Teacher (baseline), Support (planned) | Create & version your own |
Topics | Teaching, Knowledge Search (planned) | Define domain-specific jobs |
Tools | Explain, Summarize, Quiz (initial) | Add new tool contracts & handlers |
Storage (now) | Local/mock | DB / Vector stores (roadmap) |
Mental Model: Topics & Tools
Think of a Topic as a focused intent boundary. When a user asks: “Give me a quiz on black holes,” the agent:
- Classifies intent → Teaching Topic.
- Selects tool →
quiz_me
. - Executes with structured payload.
- Returns normalized JSON → formatted for display.
This structure keeps responses predictable and auditable.
Call to Action
Try the Teacher template now. Then imagine: What one domain task in your org could be templated tomorrow? Build that next.
Welcome to Agentspace. Build boldly. Teach intelligently. Ship confidently.