TripSync-AI: Multi-Agent Orchestration System
AI-powered group trip planning where every traveler's preferences get a voice. A multi-agent system negotiates a fair itinerary for the whole group and no one gets steamrolled, no one is ignored.
Multi-Agent Architecture
TripSync uses a sequential multi-agent orchestration model. Agents are not run in parallel, each proposal goes through a defined pipeline before the next one starts. This ensures budget and schedule constraints are enforced incrementally, not retrofitted at the end.
Agents
Agent | Role |
|---|---|
Preference Agent | Conversational AI that guides each traveler through 5 categories of preferences (activities, food, logistics, constraints, budget). Produces a structured preference profile. |
Activity Agent | Proposes activities for each time slot (morning, afternoon, evening) of each day. Proposals are grounded in the group's preference summary and remaining budget. |
Food Agent | Proposes meals (lunch after morning activities, dinner after afternoon activities) near the location of the preceding activity. Respects dietary restrictions. |
Accommodation Agent | Proposes lodging for each night. Considers the day's activities, group accommodation style preferences, and budget remaining. |
Transportation Agent | Plans ground transport for each day — airport transfers on day 1, inter-venue logistics on subsequent days. Constrained to 25% of the daily budget cap. |
Budget Agent | Heuristic validator. Checks each proposal against remaining budget and daily cap. Approves or rejects before it goes to debate. |
Logistics Agent | Heuristic validator. Checks that the day's schedule isn't overloaded given the group's travel pace (slow = max 4 events, moderate = 6, fast = 8). |
Traveler Advocate Agents | One per traveler. During each debate, each advocate responds to the proposal with a stance (support / neutral / object) and a utility score based on that traveler's preferences. |
Consensus Engine (Mediator) | Receives all advocate stances and validator results, then issues a decision: |
Orchestrator | Drives the full loop. Tracks budget spend, triggers budget rebalancing if earlier days over-spent, handles deferred proposals, fills under-planned days, and assembles the final itinerary. |
https://github.com/pjsk02/TripSync-AI
