Itinerary planner

Qdrant vector database and agent skill for Big Data LDN conference itinerary planning and ranking.

code View on GitHub

⚡ The Prototype // Idea #2

Itinerary planner

Built in 1.5h with Antigravity, Qdrant, FastEmbed, Python, uv

code View on GitHub

tips_and_updates The 5 Morning Sparks

Conference is coming up tomorrow

IDEA #1 lightbulb

Voice note organiser

Organise the notes against the content

Voice, Events, Productivity

IDEA #2 lightbulb

Itinerary planner

A skill to build a personalized itinerary

https://www.bigdataldn.com/en-gb/conference.html#/sessions is where the sessions are. 2 days. There is too many talks to sift through. It should rank which topics I am most interested in, scrape the data, info about the speakers and let me rank/re-rank it using human language

AI, Events, Productivity SHIPPED

IDEA #3 lightbulb

Active AI dictation

Bespoke voice-to-text setup

There are a lot of things being talked about during conferences. Some of the words are very unusual, domain specific, jargon. A good AI setup will condense knowledge into a template and collect/label different topics. But the most important and hardest thing is accurate voice attribution/transcription. The correct setup would need a custom, frictionless transcription based on the topics discussed

Voice, AI, Events

IDEA #4 lightbulb

Big Data LDN Tech Radar

A ranker for the tech discussed across 4 quadrants

This thing for tech discussed based on the talks https://www.thoughtworks.com/radar

Data Viz, Events

IDEA #5 lightbulb

Company info lookup

Find out who you are about to talk to

When walking across the booths, you see companies and the social thing is to ask them about what they are doing. But can we just learn and say "not for me" or let me prep what I want to waste their time on talking about

AI, Events

construction

How It Was Built With AI in 1.5 Hours

Agent Request

Asked to create an agent.

Repointing Scope

Had to repoint the idea to not be a web app, and to use Qdrant instead of SQLite, luckily I had another project with a similar setup.

Ingestion & Repo Setup

Agent scraped the data, deployed qdrant and created a repo, going straight into skill demonstration.

Itinerary & PDF Generation

I asked it for an itinerary, it generated a decent one, refined it a little bit, made a pdf.

thumb_up

What Rocked

✓The itinerary is actually decent

bug_report

What Broke & Fixed

✗The initial idea wanted to make too much custom vector based stuff, and was too swayed by presence of web, even though the idea was explicitly about creating a skill

smart_toy AI Interaction Summary

Genesis Prompt: "/spec-implementation"
Interaction Turns: 11 human turns across 614 execution steps

Key Guidance Turns:
  • Turn 2: lets just use something ready made like qdrant and run it via Docker. The knowledge building should be separate to the skill th...
  • Turn 3: Good except call it "big-data-ldn".
  • Turn 4: yes
  • Turn 5: mostly mlops, using AI in data processes, technologies like dbt and dlt, power and financial sector
  • Turn 6: can you make this into a pdf file
  • Turn 7: there are few I don't like, mostly around observability and compliance. I would rather keep focused on engineering and architec...
  • Turn 8: /record-implementation
  • Turn 9: 1. Asked to create an agent. Had to repoint the idea to not be a web app, and to use Qdrant instead of SQLite, luckily I had an...
  • Turn 10: Remove extra stuff that was added. Use my words from before for the 3 sections I completed. Then push the changes and create th...
  • Turn 11: just one more adjustment: break the process up into multiple steps

Tools Executed: run_command (175), manage_task (41), write_to_file (20), schedule (16), view_file (15), find_by_name (3)
Files Modified: .gitignore, AGENTS.md, Big_Data_LDN_2026_Itinerary.html, README.md, SKILL.md, __init__.py, app.py, cli.py