Olivier Dupuis

How AI Agents Assemble the Climate Narratives Report This post explores how the assembler agent transforms raw data into the Climate Narratives Report, showcasing AI’s role in streamlining workflows and generating actionable insights.

I'm continuing my explorations of how to use AI agents within a data platform. In this post and the accompanying video, I break down an agent that assembles the final weekly reports used for the Climate Narratives substack. Of course, all this is made possible with the help of Hex and LangChain.

The ideal setup would be a Daily Journal template that automatically generates a summary at the top of each entry.

I thought using a Text Generator template to parse a dynamic list of entries via Dataview might work.

I thought wrong :(

I recently started using #Obsidian. It took some time to settle on an approach that works for me, but I'm enjoying it - especially for journaling.

A challenge has been keeping track of ongoing threads of thought. I don't want to read through past entries; I’d prefer an updated summary each day.

Classics, looking forward to it 👍

Cool project Alex! Watched the first 2 videos. Wasn't aware of Evidence, seems like a nice data viz framework. Looking forward to the next update - hopefully sooner than 5 though ;)

Investor Anxiety Amidst Political Shifts in Climate Policy Weekly Climate Narratives Review - Week ending December 29, 2024

The Weekly Climate Narratives Review is back! After months of updates to the platform, this edition is the 1st in this “new era” and has been primarily assembled by AI agents.

This edition covers investor concerns over clean energy policies and more.

climatenarratives.substack.com/p/investor-a...

Finally, I can run this function within a Python script, and the output will be returned as a JSON object.

```
result = show_asset('media/nytimes_articles', orchestrator='dagster', data_warehouse='bigquery', configs_dir="../.damn/")
data = json.loads(result)
```

Once the BigQuery connector configured, I can run the following:

`damn metrics media/nytimes_articles --data-warehouse bigquery --configs-dir .damn/`

And that outputs metadata from both the orchestrator and the DW.

I had not played with the DAMN tool in a while, but lately, I've been working on a project where I needed to fetch metadata for a bunch of assets I'm producing. However, I was missing support for metadata from BigQuery. DAMN 0.1.5 now supports BQ.

github.com/republicofda...

I like the promise of Quarto. I do enjoy working on notebooks directly from my IDE. The number of configurations ahead of me is a bit intimidating, though. But at least I can get a preview going, and that's all that matters :)

It’s amazing how many open-source or inexpensive tools we have for analyzing large, unstructured data. Using @langchain.bsky.social, I added another agent to my Hex report and summarized climate events discussed in the past 28 days in under an hour.