querying-positron

name	querying-positron
description	Transforms natural language input into valid MongoDB queries for interrogating Positron's database, issues them to our Mongo cluster, retrieves and optionally analyzes or transforms the results. Use when the user wants to know about Positron's actual data about articles, authors, channels, verticals etc.
allowed-tools	Bash(.claude/skills/querying-positron/scripts/exec.sh:*)

Ask Positron

Construct valid MongoDB queries, issue them and present or analyze or transform the results.

• Use the wrapper script at .claude/skills/querying-positron/scripts/exec.sh to execute queries.

• The script handles connection setup and validates that mongosh is installed.

Examples

User question: "How many published articles are there?"

.claude/skills/querying-positron/scripts/exec.sh "db.articles.countDocuments({ published: true })"

User question: "Give me the most recent published article where layout is video"

.claude/skills/querying-positron/scripts/exec.sh 'db.articles.find( { published: true, layout: "video" } ).sort({ published_at: 1 }).limit(1)'

User question: "Count up the number of published articles for each layout value"

.claude/skills/querying-positron/scripts/exec.sh 'db.articles.aggregate([ { $match: { published: true } }, { $group: { _id: "$layout", count: { $sum: 1 } } }, { $sort: { count: -1 } }, { $project: { _id: 0, layout: "$_id", count: 1 } } ])'

General Rules

• Prefer simple queries when they will suffice, but use aggregation pipelines when necessary

• Format your queries as one-liners, to help avoid shell issues

• But if asked by the user to show your query, do so pretty-printed with concise comments ESPECIALLY for aggregation pipeline stages

• Transform Mongo's results into VALID json

  • Ensure keys are quoted
  • No ObjectId, ISODate, etc. Transform those values to plain strings.
  • Do not use the --json flag as this often fails to serialize correctly

• When querying by id, remember to wrap the id in Mongo's ObjectId function

  • Wrong ❌ db.articles.findOne({ _id: "696921ff5c8bc6fecf779dad" })
  • Right ✅ db.articles.findOne({ _id: ObjectId("696921ff5c8bc6fecf779dad") })

• When asked for urls, note the following patterns:

  • Public article urls https://www.artsy.net/article/{ARTICLE_SLUG}/edit
  • Admin editing urls https://writer.artsy.net/articles/{ARTICLE_ID}/edit
  • Artwork urls https://www.artsy.net/artwork/{ARTWORK_ID_OR_SLUG}/edit

Articles

See the example article to understand the shape of a typical document from the articles collection

This will be the subject of most queries.

Article-specific rules:

• Assume the user is only interested in articles where published: true, unless they explicitly tell you otherwise.
• The singular slug field is not reliable, instead we want to treat the final entry in the slugs array as the canonical slug.
• IMPORTANT: When asked about featured artworks (featured_artwork_ids), DO NOT conflate that with featured artists (primary_featured_artist_ids)

Article Authorship Fields

Articles contain multiple author-related fields that serve different purposes:

Display Attribution

• author: Object containing the organizational/editorial account (typically "Artsy Editorial")

  • author.name: Display name (e.g., "Artsy Editorial")
  • author.id: Gravity user ID for this account

• contributing_authors: Array of individual writer(s) who should be credited on the article

  • Each entry has name and id (Gravity user ID)
  • This field determines the byline shown to users on the website
  • WARNING: Often empty even when articles have actual authors

Author ID Fields

Articles store author references using multiple ID systems:

• author_ids (array): Positron's internal author record IDs from the authors collection

  • This is the authoritative source for author attribution
  • Use for lookups/joins with the authors collection to get actual author names

• author.id and contributing_authors[].id: Gravity user IDs

  • Use for Gravity API calls or authentication-related queries

• author_id (singular): Legacy field, purpose varies

Best Practice: Getting Author Names

To get correct author attribution, use an aggregation pipeline with $lookup to join with the authors collection:

db.articles.aggregate([
  { $match: { published: true } },
  {
    $lookup: {
      from: "authors",
      localField: "author_ids",
      foreignField: "_id",
      as: "author_records",
    },
  },
  // Then use this priority for display:
  // 1. contributing_authors[0].name (if array is non-empty)
  // 2. author_records[0].name (from the lookup)
  // 3. author.name (fallback)
])

Do NOT rely solely on contributing_authors or author.name - they may not reflect the actual article author. Always join with the authors collection using author_ids.

Known Issue: Partner-authored Articles

It is a known problem that a handful of partners continue to publish articles themselves, even though we officially stopped supporting this capability years ago.

Before querying for articles authored by partners you MUST refer to Partner-authored Article Problem

IMPORTANT: Do not conflate Partner-authored articles (where partner_channel_id is present) with Artsy-authored articles (where partner_channel_id is null).

Additional Article Relationships

When needed refer to the following documents:

• Article Artists, including "featured artists"
• Article Artworks, including "featured artworks"

Additional Mongo collections

Here is a list of all collections in Positron’s MongoDB.

Most of the time you will be interested in articles, but these others are available as well:

articles
authors
channels
curations
organizations
sections
sessions
tags
users
verticals

You can inspect any of these collections’ document shapes by requesting the most recent documents, e.g:

.claude/skills/querying-positron/scripts/exec.sh 'db.authors.find().sort({ _id: -1 }).limit(3)'