---
id: ch5-guide-raven-magpie
type: spine
title: "How to find a dataset with Magpie and publish yours with Raven"
readingTime: 2
standalone: true
core: false
teaser: "If your region's data is not in the catalogue, JackDaw cannot use it. Here is how data gets found, and how it gets in."
parent: null
diagram: fig-guide-raven-magpie
recallQ: "What must a data provider fill in when publishing a dataset in Raven, and why does it decide what JackDaw can do with the data?"
recallA: "After a resumable upload, the provider gives a title, a description (ideally in several languages), tags, temporal coverage or an explicit 'unknown', and a description of every column, then publishes from draft. JackDaw finds datasets by semantic matching on this metadata, so weak descriptions mean the data is found and used poorly."
highlights:
  - "Magpie demonstrator: jackdaw.avinet.no/magpie; in the Beta its search runs inside Raven and shows up through JackDaw"
  - "Raven uploads large vector files in resumable chunks"
  - "Describe every column; multilingual descriptions improve search recall"
  - "Datasets stay drafts until you publish them"
status: accepted
concept: tool-ecosystem
lang: en
lens: finland
depth: standard
formalism: none
visuality: balanced
lengthBand: standard
genre: worked-example
state: edited
carriers: prose|diagram
---

**When to use it:** you are looking for geodata for a question, or you hold a local dataset that other tools should be able to use.

## What you need

- To search: the Magpie demonstrator at **[jackdaw.avinet.no/magpie](https://jackdaw.avinet.no/magpie)** (D4.1 §6.5), or simply a dataset question in [JackDaw](#c/jackdaw).
- To publish: access to Raven. The pack gives no public address; Raven sits behind the platform's shared identity provider, so access comes through your pilot's technical partner.
- A geospatial **vector** dataset and what you know about it: source, period, meaning of each column.

## Steps

1. **Search in plain language.** Magpie turns your query into an embedding and compares it with embedded ISO 19115/19119 metadata. Results are records with layer names, descriptions, keywords and WMS/WFS endpoints, and it can build a map from them (D4.1 §6.5). In the Beta this function runs inside Raven and usually appears through JackDaw (D4.2 §3.4).
2. **Upload what is missing.** Raven sends large files in chunks, up to three in parallel; an interrupted upload resumes where it stopped (D4.2 §3.3.3).
3. **Describe it.** Give a title, a description, classification tags, temporal coverage (a date range, or marked explicitly as unknown) and a **semantic description of every column**. The interface encourages descriptions in more than one language to improve search recall.
4. **Publish.** The dataset stays a draft until you publish it to the catalogue.
5. **Maintain it.** Edit metadata later, attach methodology documents or data dictionaries, grant or revoke access for named users, or replace the geometry layer without a full re-ingest where the state allows it.
6. **Test it in JackDaw.** Ask about your area; JackDaw's GeoRAG step finds datasets in the Raven catalogue by semantic matching and then queries them spatially.

## What you get

A dataset in PostGIS that can be queried by SQL and by natural-language semantic search, and that platform tools can use **without code changes** (D4.2 §3.3.3, §6.3).

## Pilot use

In Mallusjoki, Raven metadata services ran alongside JackDaw, the Advisor and the MAA tool. D4.2 reports municipal-level spatial analysis that combined open geospatial datasets with local planning needs for the rural events ecosystem (Table 7).

## Watch out

- **Search cannot fix metadata.** Incomplete records, weak keywords and inconsistent classification limit what Magpie or Raven can find (D4.2 §3.4).
- Uneven source quality still causes uneven behaviour across datasets (§3.2).
- Magpie will not grow as a stand-alone product; development continues through Raven.

**Sources:**
- D4.2 PoliRuralPlus Platform Beta, §3.2 Data and metadata services; §3.3 Jackdaw (3.3.3 Raven); §3.4 Magpie; §6.2 Table 7; §6.3 Usability observations. [PDF](https://www.poliruralplus.eu/docs/421/D4.2_PoliRuralPlus_Platform_Beta.pdf)
- D4.1 PoliRuralPlus platform design, §6.5 Magpie: AI-driven OGC Metadata Catalog Search. [PDF](https://www.poliruralplus.eu/docs/162/D4.1_PoliRuralPlus_platform_design_v1.pdf)