Berlin Transport AI Assistant

Introduction

The virtual_hackathon_VuHo workflow is designed to assist users with real-time public transportation information specifically for Berlin. It interacts with a transport information server to answer queries about routes, disruptions, and schedules through an iterative AI-driven process.

It does 7 things:

1. Listens for incoming webhook triggers that start the process.
2. Initializes a session with the public transport information server.
3. Extracts session details and server capabilities from the initialization response.
4. Fetches the list of available tools from the server using the session id.
5. Enters an iterative loop where an AI agent decides which transport tool to call based on the user's query and previous results.
6. Processes the AI decisions either to call a specific tool or finalize the answer.
7. Aggregates results and outputs a comprehensive response to the user.

What you need

1. Access to the Needle automation platform.
2. An active webhook to receive incoming user requests.
3. Network access to the Berlin public transport information server at http://89.167.56.59:3000/mcp.
4. The workflow template named 'virtual_hackathon_VuHo' imported into your Needle account.

How the flow works

Webhook trigger: Waits for incoming queries from users to start the flow.

Initialize MCP session (HTTP request): Sends a JSON-RPC 'initialize' request to the Berlin transport server to begin a session.

Parse initialization response (Code step): Extracts the session id from response headers and parses server capabilities from the body.

Fetch tools list (HTTP request): Requests a list of available transport tools using the session id to understand what queries can be made.

Parse tools list (Code step): Extracts tool information and prepares a default user query for iterative processing.

Iterative AI loop (Loop with AI node): Uses an AI assistant to analyze the user query and past tool results to decide the next step. The AI can:

1. Call a transport tool with specific arguments to gather data.
2. Provide a final travel answer including route planning and disruption info.

Decision processing (Code steps): Interprets the AI output to inform whether to continue iterating or finish.

Final response formatting (Code step): Compiles the final AI-generated answer or combined tool results into a user-friendly message.

Output

At the end of the workflow, users receive a well-structured and emoji-enhanced response about Berlin public transportation. This can include recommended routes, real-time schedules, alerts about disruptions, and other helpful travel details.

Notes

1. The iterative AI loop limits to five iterations to balance thoroughness and efficiency.
2. The workflow includes safeguards to avoid repeated tool calls with identical parameters.
3. Emoji usage helps highlight different types of transport and alert statuses for better readability.
4. Make sure your network environment allows POST requests to the specified server for seamless operation.
5. Custom user queries can be sent to the webhook to get personalized transport info dynamically.