Agent Requirements

Agent Objective:

What is the agent designed to accomplish? Define the core task and success criteria...

Requirements

1. can search flight - keep eye on budget
2. can search hotels
3. Can check activities
4. Booking Reservations
5. Check realtime pricing
6. Can accommodate user preference changes

Autonomy & Constraints:

What level of autonomy does the agent have? What are the boundaries and constraints it must operate within?

Autonomy

1. can execute read query - like search flight, hotel, activites
2. Keep HITL prior booking
3. Keep HITL prior accommodation preference change

Constraints

1. prior suggesting the plan calculate the overall cost of the trip and only communicate to user if it is less than or equal to the budget entered by customer

Tool & Action Design

Define the tools and actions available to the agent. Specify tool interfaces, input/output schemas, and how the agent selects which tool to use...

Agent Architecture

Design the overall agent architecture. Choose an orchestration pattern (single agent, multi-agent, hierarchical). Identify key components and how they interact. Use the diagramming tool to illustrate the architecture.

Architecture - It will be multi-agent orchestration.

One coordinator agent orchestrates the call between subagents.

Agents -

1. UserQuery
2. Flight
3. Hotel
4. Activities
5. Cost calculator
6. Validator
7. User preference

Now these all agent has there subagent to achieve the activities

Agent and subagent have access to set of tools to achieve atomic task

Memory & Context Design

How does the agent manage memory and context?

Define short-term (conversation) and long-term (persistent) memory strategies. How is context window managed? What retrieval mechanisms are used (RAG, vector stores, summarization)?

For optimal Context and memory, predefined optimized system prompt will be coded.

To achieve the deterministic in the flow, we can design multiple hooks. For example - always call search agent proceeding with the booking. So here we can define a PretoolUseHook for booking.

To achieve the consistency in data transfer between agents through the coordinator agent, we can define the output format and schema

To achieve reliability, after fetching of data, it will be loaded in scratchpad. So that if the session ended , once session is resumed , agent wont lose the context.

For retrieval, we can built a RAG and store the session data in an in-memory vectordb and add the archival rule.

By using these two strategies (ScratchPad+RAG) context window will be maintained.

Core Reasoning

Deep dive into the agent's reasoning strategy. How does it plan, decide, and self-correct? Discuss prompt engineering, chain-of-thought, reflection loops, failure handling, and tradeoffs.

One central Orchestrator agent.

Responsibilites -

1. Take user queries
2. Decide and spawn agent call
3. maintain the context
4. Consoliate agents response
5. Respond back to user

Multiple Subagents - Each subagent should have appropriate name based on the domain they will act.

Each subagent should have set of tools. Tools have appropriate tool descriptions.

Subagents

1. Flight
2. Hotel
3. Activities
4. Resevation

The tools mentioned in the subagent have proper error handling to handle resilency. e.g trasient error will be retryable with exponential backoff . Nonretyable error will be returning proper error code and description to the coordinator.

Safety & Permissions

What guardrails and safety mechanisms does the agent have? Define permission boundaries, human-in-the-loop checkpoints, content filtering, rate limiting, and how the agent handles adversarial inputs or unexpected states.

Define each subagent boundaries based on their business goal.

Non of the subagent should be overlapping

maintain a state by a stateid and save the stateid on scratch pad to maintain idempotency.