Skip to content

docs: add synonyms so Algolia surfaces the hybrid/self-hosted pages - #1144

Open
teallarson wants to merge 5 commits into
mainfrom
docs/algolia-zero-hit-synonyms
Open

docs: add synonyms so Algolia surfaces the hybrid/self-hosted pages#1144
teallarson wants to merge 5 commits into
mainfrom
docs/algolia-zero-hit-synonyms

Conversation

@teallarson

@teallarson teallarson commented Aug 27, 2026

Copy link
Copy Markdown
Contributor

From Teal, the human

🚨 I think my agent may be wrong on some of these changes, but I don't know these product surfaces well enough to say how. Please DO read them.

Problem

The latest Algolia weekly summary (application BJB8PBSQ9T, Arcade Docs, Aug 17–23) reports a 32.76% no-result rate on docs_arcade_dev_bjb8pbsq9t_docsearch. Several of the zero-hit queries point at pages that already exist in this repo but use different words than the users typing them.

Zero-result search queries in scope

From the Algolia “Searches with no results” list:

  • hybrid worker deployment
  • self-hosted mcp server private network
  • hybrid deployment self-hosted worker vpc private network cloud engine
  • ip allowlist firewall egress

The rest of the report's zero-hit list (kept here for context, not addressed by this PR because they're real content gaps rather than terminology mismatches):

  • railway
  • heroku
  • redact pii from tool outputs response transformation middleware
  • gateway oauth token expiration refresh token session lifetime mcp client re-authenticate
  • google cloud console create oauth client redirect url arcade dashboard add provider
  • add secrets to deployed mcp server dashboard secrets page requires_secrets

Why the four in-scope queries returned zero

  • app/en/operate/deploy/on-prem/page.mdx is titled Hybrid MCP servers, but the body only ever uses “on-premises.” The words worker, self-hosted, hybrid worker, VPC, private network, and firewall never appear on the page, so Algolia can't match them.
  • app/en/operate/deploy/warp-pipes/page.mdx describes the SSRF allowlist and outbound connections, but never uses IP allowlist, firewall, or egress — the exact wording of the user query.

Fix

Add the missing synonyms naturally into the prose of both pages so the crawler indexes them without changing meaning or introducing new concepts.

app/en/operate/deploy/on-prem/page.mdx

  • Frontmatter description now names it as a “self-hosted MCP server worker,” “private network or VPC,” and “hybrid worker deployment.”
  • Overview paragraph calls out “self-hosted worker or hybrid worker,” “on-premises, private cloud, VPC, or any private network,” and “private resources behind a firewall.”
  • “How it works” step 1 lists the same environments (VPC / private network) so the terms appear both in prose and as a numbered list item (both are in the crawler's indexed selectors).
  • Adds a see-also line pointing at Warp Pipes for the “outbound-only egress from your VPC” case.

app/en/operate/deploy/warp-pipes/page.mdx

  • Renames the SSRF allowlist inline as “an IP allowlist for outbound egress” on first mention.
  • Rephrases the follow-up sentence to say “your firewall stays closed to inbound traffic — the runtime only makes outbound egress calls.”

No behavior, no config, no restructuring — just terminology the crawler was missing.

Effect on the Algolia index

On merge to main, algolia-reindex.yml fires (the paths touched are under app/en/**), which triggers a fresh crawl. The four zero-hit queries above should now surface the corresponding pages.

Test plan

  • Merge and wait for algolia-reindex.yml to complete
  • Search each of the four in-scope queries on docs.arcade.dev and confirm the on-prem / warp-pipes page appears
  • Check next week's Algolia weekly summary to confirm the no-result rate drops

Note

Low Risk
Documentation-only wording changes; the only behavioral risk is losing the Arcade vs Remote MCP callout during registration.

Overview
Updates the Hybrid MCP servers (on-prem) docs so Algolia can match queries like “hybrid worker,” “self-hosted MCP,” “VPC,” and “private network,” without changing the underlying setup flow.

The frontmatter description and early overview now name a self-hosted / hybrid worker deployment on private network or VPC, and prose repeats those terms in the “how it works” and benefits sections (including firewall and internal APIs). Registration is described as registering the server as a worker where relevant.

Removes the info Callout that contrasted the Arcade server type with Remote MCP and linked to the remote MCP governance page—worth confirming that distinction isn’t still needed for readers on this page.

Reviewed by Cursor Bugbot for commit 17abc55. Bugbot is set up for automated code reviews on this repo. Configure here.

@vercel

vercel Bot commented Aug 27, 2026

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

Project Deployment Actions Updated
docs Ready Ready Preview Sep 1, 2026 8:24pm UTC

Request Review

Comment thread app/en/operate/deploy/on-prem/page.mdx Outdated
@teallarson
teallarson marked this pull request as ready for review September 1, 2026 20:11
Comment thread app/en/operate/deploy/on-prem/page.mdx Outdated
</Callout>

1. **Navigate to the MCP Servers page** in your [Arcade dashboard](https://app.arcade.dev/servers)
1. **Navigate to the MCP Servers page** in your [Arcade dashboard](https://api.arcade.dev/dashboard/servers)

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

is this correct?

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

No and agents won't stop doing it 😖

Comment thread app/en/operate/deploy/on-prem/page.mdx Outdated
…-synonyms

# Conflicts:
#	app/en/operate/deploy/warp-pipes/page.mdx

@evantahler evantahler left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

SEO Hacking!

<GuideOverview.Outcomes>

An on-premises MCP server deployment allows you to execute tools in your own environment while still leveraging Arcade's cloud Engine infrastructure. This gives you the flexibility to access private resources, maintain data security, and customize your environment while leveraging Arcade's MCP server management and federation capabilities.
A hybrid MCP server deployment — also called a self-hosted worker or hybrid worker — runs the MCP server on infrastructure you control (on-premises, private cloud, VPC, or any private network) while Arcade's cloud Engine keeps handling identity, routing, and governance. This gives you the flexibility to access private resources behind a firewall, keep sensitive data inside your network, and customize your environment while leveraging Arcade's MCP server management and federation capabilities.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is all technical correct!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants