← Extractions

Optimizing Rails apps with a GraphQL layer (Caio Almeida, wroclove.rb 2024)

Caio Almeida from Meedan shares eight years of hard-won lessons running a Rails + GraphQL monolith (the Check fact-checking platform). He walks through performance testing, monitoring at HTTP/query/field levels, N+1 solutions (graphql-batch, graphql-preload, look-ahead), safeguards (max depth, timeout), caching (graphql-ruby-fragment-cache, event-based cached fields), query batching via multiplex, and Q&A on authorization and whether to go back to REST.

Model

claude-opus-4-7

Ingestion

3f33be13

Input tokens

110,123

fresh: 41,671
cached: 55,428
cache write: 13,024

Output tokens

11,701

Duration

167.5s

Roundtrips

Tool calls

Cost

$0.00

Nodes/edges extracted

34 / 54

Read set (nodes/edges)

68 / 2

Nodes (34)

create Sentry tool

kind (empty) → tool

name (empty) → Sentry

slug (empty) → sentry

attrs (empty) → {"category" => "service"}

description (empty) → Error-tracking service Meedan integrates with their Rails app. A graphql-ruby timeout handler calls Sentry.notify whe...

short_description (empty) → Error tracking and alerting service.

create PagerDuty tool

kind (empty) → tool

name (empty) → PagerDuty

slug (empty) → pagerduty

attrs (empty) → {"category" => "service"}

description (empty) → On-call alerting service triggered by uptime incidents (e.g. when an HTTP request exceeds a latency threshold) so tha...

short_description (empty) → On-call incident alerting service.

create Relay tool

kind (empty) → tool

name (empty) → Relay

slug (empty) → relay

attrs (empty) → {"category" => "library"}

description (empty) → Facebook's GraphQL client for React. Meedan's Check front-end uses React + Relay; the Relay network layer can be twea...

short_description (empty) → React GraphQL client with declarative data-fetching.

update Caio Almeida person

attrs (empty) → {"country" => "Brazil", "rails_since" => "2018"}

description Conference speaker. → Brazilian software engineer based in Brazil, working with Rails since 2018. Joined Meedan in 2016 as a software engin...

short_description Conference speaker. → Brazilian software engineer at Meedan, Rails/GraphQL practitioner.

create Meedan company

kind (empty) → company

name (empty) → Meedan

slug (empty) → meedan

attrs (empty) → {"industry" => "nonprofit / fact-checking tools", "headquarters" => "San Francisco"}

description (empty) → Nonprofit based in San Francisco that develops tools for collaborative fact-checking. Its main product, Check, was fu...

short_description (empty) → San Francisco-based nonprofit building collaborative fact-checking tools.

create Check project

kind (empty) → project

name (empty) → Check

slug (empty) → check

attrs (empty) → {"status" => "active", "license" => "open-source", "architecture" => "microservices, headless Rails GraphQL API"}

description (empty) → Meedan's main product: an open-source platform for fact-checking organizations around the world to track and verify m...

short_description (empty) → Meedan's open-source collaborative fact-checking platform.

create AWS CloudWatch tool

kind (empty) → tool

name (empty) → AWS CloudWatch

slug (empty) → aws-cloudwatch

attrs (empty) → {"category" => "service"}

description (empty) → AWS logging service where Meedan collects Rails HTTP request logs. Useful for total request duration but not for pinp...

short_description (empty) → AWS logging and metrics service.

update Optimizing performance in Rails apps with GraphQL layer talk

description Talk at wroclove.rb 2024. → Conference talk at wroclove.rb 2024 by Caio Almeida of Meedan, sharing eight years of lessons building and operating ...

short_description Talk at wroclove.rb 2024. → Caio Almeida's wroclove.rb 2024 talk on performance patterns for Rails + GraphQL.

update wroclove.rb 2024 event

No changes recorded.

update GraphQL tool

description API query language cited as having use cases where JSON API 'basically sucks'. Panel advice: be open-minded and pick ... → Query language describing data using a type system, with bindings for many programming languages. Open-sourced by Fac...

create graphql-ruby tool

kind (empty) → tool

name (empty) → graphql-ruby

slug (empty) → graphql-ruby

attrs (empty) → {"category" => "library", "language" => "Ruby"}

description (empty) → The canonical Ruby implementation of GraphQL. Early versions used DSL-style definers that were hard to write and reus...

short_description (empty) → Ruby gem implementing the GraphQL specification for Ruby/Rails apps.

create graphql-batch tool

kind (empty) → tool

name (empty) → graphql-batch

slug (empty) → graphql-batch

attrs (empty) → {"category" => "library", "language" => "Ruby"}

description (empty) → Ruby gem originally implemented by Shopify that batches data loading in graphql-ruby resolvers. Using a RecordLoader ...

short_description (empty) → Shopify gem for batch-loading associations in GraphQL resolvers.

create graphql-preload tool

kind (empty) → tool

name (empty) → graphql-preload

slug (empty) → graphql-preload

attrs (empty) → {"category" => "library", "language" => "Ruby"}

description (empty) → Ruby gem used to preload has_many associations (e.g. a post's tags) at the beginning of a GraphQL query scope, avoidi...

short_description (empty) → Ruby gem that eager-loads has_many associations in GraphQL resolvers.

create graphql-ruby-fragment-cache tool

kind (empty) → tool

name (empty) → graphql-ruby-fragment-cache

slug (empty) → graphql-ruby-fragment-cache

attrs (empty) → {"category" => "library", "language" => "Ruby"}

description (empty) → Ruby gem that caches fragments (individual fields) at the GraphQL API layer in graphql-ruby. Fields declare cache_fra...

short_description (empty) → Gem for caching individual fields at the GraphQL API layer.

create Honeycomb tool

kind (empty) → tool

name (empty) → Honeycomb

slug (empty) → honeycomb

attrs (empty) → {"category" => "service"}

description (empty) → Observability tool used at Meedan to trace GraphQL requests via OpenTelemetry. Gives per-request spans for parsing, v...

short_description (empty) → Observability SaaS built around OpenTelemetry tracing.

create OpenTelemetry tool

kind (empty) → tool

name (empty) → OpenTelemetry

slug (empty) → opentelemetry

attrs (empty) → {"category" => "platform"}

description (empty) → Open observability standard that instruments applications with spans and metrics. Used under the hood by Honeycomb to...

short_description (empty) → Vendor-neutral standard for distributed tracing and metrics.

create Apollo GraphQL tool

kind (empty) → tool

name (empty) → Apollo GraphQL

slug (empty) → apollo-graphql

attrs (empty) → {"category" => "service"}

description (empty) → GraphQL platform used by Meedan for schema visualization (acting as living documentation and schema-change tracking) ...

short_description (empty) → GraphQL platform offering schema visualization, monitoring and client tooling.

update Pundit tool

description Popular Ruby authorization gem; used in Adrian's talk as the example runtime dependency declared in an engine's gemsp... → Popular Ruby authorization gem. Meedan adopted Pundit when they first built Check with GraphQL, because graphql-ruby ...

short_description Minimal authorization Ruby gem using policy classes. → Ruby gem for object-level authorization policies.

update N+1 Queries concept

description Common ORM query-per-row performance problem. Sergyenko lists it as a core thing Ruby engineers working with data mus... → Common ORM performance problem where the number of database queries is proportional to the number of rows loaded. In ...

create Nested GraphQL query bottleneck concept

kind (empty) → concept

name (empty) → Nested GraphQL query bottleneck

slug (empty) → nested-graphql-query-bottleneck

attrs (empty) → {"category" => "pattern"}

description (empty) → Concept discussed in the talk: because GraphQL is highly expressive, clients can issue deeply nested queries (and eve...

short_description (empty) → Deeply nested GraphQL queries hiding expensive or circular data fetches.

create GraphQL look_ahead concept

kind (empty) → concept

name (empty) → GraphQL look_ahead

slug (empty) → graphql-look_ahead

attrs (empty) → {"category" => "pattern"}

description (empty) → graphql-ruby feature where a field receives a look_ahead object describing the tree of fields the caller requested. R...

short_description (empty) → graphql-ruby feature exposing the requested query tree to resolvers.

create GraphQL max_depth concept

kind (empty) → concept

name (empty) → GraphQL max_depth

slug (empty) → graphql-max_depth

attrs (empty) → {"category" => "practice"}

description (empty) → A graphql-ruby schema option that rejects queries nested beyond a configured depth (e.g. 4). Simple safeguard against...

short_description (empty) → Schema-level limit on nesting depth of incoming GraphQL queries.

create GraphQL query timeout concept

kind (empty) → concept

name (empty) → GraphQL query timeout

slug (empty) → graphql-query-timeout

attrs (empty) → {"category" => "practice"}

description (empty) → Control added to a graphql-ruby schema that caps the total execution time of a GraphQL query. When exceeded, the hand...

short_description (empty) → Per-query maximum execution time enforced at the GraphQL schema level.

create GraphQL query batching concept

kind (empty) → concept

name (empty) → GraphQL query batching

slug (empty) → graphql-query-batching

attrs (empty) → {"category" => "pattern"}

description (empty) → Pattern where a GraphQL client (Apollo's client-side batching, or React Relay's network layer) combines queries trigg...

short_description (empty) → Combining multiple GraphQL queries into one HTTP request and executing them concurrently.

create Event-based GraphQL field caching concept

kind (empty) → concept

name (empty) → Event-based GraphQL field caching

slug (empty) → event-based-graphql-field-caching

attrs (empty) → {"category" => "pattern"}

description (empty) → Caching approach Meedan built in-house: each expensive cached field declares a starting value and a list of model eve...

short_description (empty) → Explicit list of model events that invalidate a cached GraphQL field.

create GraphQL performance regression tests concept

kind (empty) → concept

name (empty) → GraphQL performance regression tests

slug (empty) → graphql-performance-regression-tests

attrs (empty) → {"category" => "practice"}

description (empty) → Practice of writing tests (e.g. Minitest assert_queries helper) that assert how many database queries a given GraphQL...

short_description (empty) → Minitest helpers asserting maximum database queries per GraphQL operation.

create Premature optimization concept

kind (empty) → concept

name (empty) → Premature optimization

slug (empty) → premature-optimization

attrs (empty) → {"category" => "practice"}

description (empty) → Well-known computer-science adage cited as the 'root of all evil'. In the talk used to temper optimization advice: an...

short_description (empty) → Classic warning that optimizing too early wastes effort on the wrong code.

create Monitoring is the most important optimization habit takeaway

kind (empty) → takeaway

name (empty) → Monitoring is the most important optimization habit

slug (empty) → monitoring-is-the-most-important-optimization-habit

attrs (empty) → {"type" => "recommendation"}

description (empty) → Speaker's single most-emphasized takeaway: you can't improve what you don't measure. Set up monitoring at multiple le...

short_description (empty) → Without layered monitoring you can't identify real GraphQL bottlenecks.

create Make performance a team mindset takeaway

kind (empty) → takeaway

name (empty) → Make performance a team mindset

slug (empty) → make-performance-a-team-mindset

attrs (empty) → {"type" => "recommendation"}

description (empty) → Performance is an ongoing process, not a one-off sprint. For sustained adoption of monitoring and optimization tools,...

short_description (empty) → Optimization must be continuous and owned by the whole engineering team.

create Always evaluate trade-offs of GraphQL optimizations takeaway

kind (empty) → takeaway

name (empty) → Always evaluate trade-offs of GraphQL optimizations

slug (empty) → always-evaluate-trade-offs-of-graphql-optimizations

attrs (empty) → {"type" => "insight"}

description (empty) → Every GraphQL optimization (extra gems, look_ahead resolvers, custom caching DSLs) adds dependencies, learning curve ...

short_description (empty) → Performance wins like look_ahead can cost readability and onboarding time.

create Delay mutations to background jobs takeaway

kind (empty) → takeaway

name (empty) → Delay mutations to background jobs

slug (empty) → delay-mutations-to-background-jobs

attrs (empty) → {"type" => "recommendation"}

description (empty) → For GraphQL mutations with heavy side-effects — calls to external services, bulk operations like 'select all + tag' o...

short_description (empty) → Offload slow mutation work (external calls, bulk ops) to background jobs.

create Denormalization vs graphql-batch / graphql-preload question

kind (empty) → question

name (empty) → Denormalization vs graphql-batch / graphql-preload

slug (empty) → denormalization-vs-graphql-batch-graphql-preload

attrs (empty) → {"answer_summary" => "Didn't evaluate denormalization specifically; graphql-batch was the first fix, and because Chec...

description (empty) → Audience question asking whether the team evaluated data denormalization (e.g. adding a tags field directly on the Po...

short_description (empty) → Did you compare denormalizing fields versus using batch/preload gems?

create Authorization strategies in GraphQL question

kind (empty) → question

name (empty) → Authorization strategies in GraphQL

slug (empty) → authorization-strategies-in-graphql

attrs (empty) → {"answer_summary" => "Authenticate at the controller; authorize at type/field level using graphql-ruby + Pundit. Meed...

description (empty) → Audience question about authorization patterns in large GraphQL apps: field level vs layer level, how layers coordina...

short_description (empty) → How to handle authorization for flexible GraphQL queries?

create Have you considered going back to REST? question

kind (empty) → question

name (empty) → Have you considered going back to REST?

slug (empty) → have-you-considered-going-back-to-rest

attrs (empty) → {"answer_summary" => "No. Check is open-source with three in-house clients and many external fact-checking agencies —...

description (empty) → Audience question at the end of the talk. Speaker jokes he 'scratches himself every day' but no: Check is open-source...

short_description (empty) → Would you drop GraphQL and return to REST APIs?

Edges (54)

update Caio Almeida → authored → Optimizing performance in Rails apps with GraphQL layer

context (empty) → Delivered the talk at wroclove.rb 2024.

update Optimizing performance in Rails apps with GraphQL layer → presented_at → wroclove.rb 2024

context (empty) → Talk given at wroclove.rb 2024 (originally scheduled for 2020 before pandemic cancellation).

create Caio Almeida → attended → wroclove.rb 2024

context (empty) → Speaker at the conference.

relation (empty) → attended

source_node_id (empty) → 8e94e35e-1e33-48c7-ac3e-b72e4e1c8c9e

target_node_id (empty) → 748e0524-a102-41d4-97dc-575881583d91

create Caio Almeida → works_at → Meedan

attrs (empty) → {"role" => "software engineer", "since" => "2016"}

context (empty) → Software engineer at Meedan since 2016.

relation (empty) → works_at

source_node_id (empty) → 8e94e35e-1e33-48c7-ac3e-b72e4e1c8c9e

target_node_id (empty) → cc6d7a67-eef0-4196-9a7e-0874b4868ece

create Caio Almeida → works_on → Check

attrs (empty) → {"role" => "lead engineer of PHP→Rails rewrite"}

context (empty) → Led migration of Meedan's main product Check from PHP to Rails.

relation (empty) → works_on

source_node_id (empty) → 8e94e35e-1e33-48c7-ac3e-b72e4e1c8c9e

target_node_id (empty) → 8b4e5d9c-9987-43bb-b7bd-3c9f2fc5af4a

create Check → uses → GraphQL

context (empty) → Check's core Rails API is a headless GraphQL API.

relation (empty) → uses

source_node_id (empty) → 8b4e5d9c-9987-43bb-b7bd-3c9f2fc5af4a

target_node_id (empty) → c6a1041c-3116-4194-b7f0-eba7cd4a7e64

create Check → uses → graphql-ruby

context (empty) → Ruby implementation of GraphQL powering Check's API.

relation (empty) → uses

source_node_id (empty) → 8b4e5d9c-9987-43bb-b7bd-3c9f2fc5af4a

target_node_id (empty) → be81c4e9-2645-4f1d-90cc-0d6cfe93cb6c

create Check → uses → graphql-batch

context (empty) → Used to batch-load belongs_to associations and avoid N+1 queries.

relation (empty) → uses

source_node_id (empty) → 8b4e5d9c-9987-43bb-b7bd-3c9f2fc5af4a

target_node_id (empty) → f3af2e65-d82f-4f24-81e2-ac1f84dd0b2b

create Check → uses → Relay

context (empty) → Check's React front-end uses Relay with query batching in the network layer.

relation (empty) → uses

source_node_id (empty) → 8b4e5d9c-9987-43bb-b7bd-3c9f2fc5af4a

target_node_id (empty) → 204d8f3f-4978-4072-ac57-92411341cca4

create Check → uses → Honeycomb

context (empty) → Monitoring GraphQL query execution via OpenTelemetry spans.

relation (empty) → uses

source_node_id (empty) → 8b4e5d9c-9987-43bb-b7bd-3c9f2fc5af4a

target_node_id (empty) → 9fe05a75-6726-4be8-b578-0058603267a6

create Check → uses → OpenTelemetry

context (empty) → Instrumentation underlying Honeycomb tracing.

relation (empty) → uses

source_node_id (empty) → 8b4e5d9c-9987-43bb-b7bd-3c9f2fc5af4a

target_node_id (empty) → 39ce1c2f-63e7-4a8d-859f-40e7f1bd6d4b

create Check → uses → Apollo GraphQL

context (empty) → Schema visualization, per-field monitoring, and Slack daily summaries of slow queries.

relation (empty) → uses

source_node_id (empty) → 8b4e5d9c-9987-43bb-b7bd-3c9f2fc5af4a

target_node_id (empty) → cebd5aab-3634-4afe-b0a8-ba8fb96ddffc

create Check → uses → AWS CloudWatch

context (empty) → Collects Rails HTTP request logs.

relation (empty) → uses

source_node_id (empty) → 8b4e5d9c-9987-43bb-b7bd-3c9f2fc5af4a

target_node_id (empty) → a2a3a19e-767f-42a8-8776-c2fcf62c6397

create Check → uses → Sentry

context (empty) → Receives notifications when GraphQL queries exceed timeout.

relation (empty) → uses

source_node_id (empty) → 8b4e5d9c-9987-43bb-b7bd-3c9f2fc5af4a

target_node_id (empty) → 9b81df6c-7817-4cd8-91a1-bfeefb7fbce8

create Check → uses → PagerDuty

context (empty) → Triggers on-call incidents from uptime alerts.

relation (empty) → uses

source_node_id (empty) → 8b4e5d9c-9987-43bb-b7bd-3c9f2fc5af4a

target_node_id (empty) → fefd3a9e-5cec-4b86-8ba8-cb49dc16f4cd

create Check → uses → Pundit

context (empty) → Authorization enforced below GraphQL because graphql-ruby didn't offer it at the API layer when Check adopted GraphQL.

relation (empty) → uses

source_node_id (empty) → 8b4e5d9c-9987-43bb-b7bd-3c9f2fc5af4a

target_node_id (empty) → 1ebb50fc-98ba-48a0-8720-cbf7138c2eed

create Meedan → works_on → Check

context (empty) → Check is Meedan's flagship product.

relation (empty) → works_on

source_node_id (empty) → cc6d7a67-eef0-4196-9a7e-0874b4868ece

target_node_id (empty) → 8b4e5d9c-9987-43bb-b7bd-3c9f2fc5af4a

create Optimizing performance in Rails apps with GraphQL layer → about → GraphQL

context (empty) → Talk focuses on Rails applications with a GraphQL layer.

relation (empty) → about