Migration

Change your search engine without a blind migration

Importing documents is not enough. A real migration covers schema, ranking, filters, UI, analytics and the cutover. AACSearch helps you go through it step by step and keep a way to roll back.Start the migration Choose your current platform

Don't start by switching the old search off

The right order:

01
Inventory
Collect:
- indexes
- schema
- ranking
- synonyms
- rules
- facets
- filters
- frontend integrations
- events
02
Map
Match your current capabilities to AACSearch.
03
Import
Move the data and the settings across.
04
Benchmark
Run your real queries against both systems.
05
Shadow
Compare the answers before you switch any traffic.
06
Cutover
Move production over gradually.
07
Rollback
Keep a clear way back until the new system is confirmed.

A migration should show more than parity

If a feature works differently, that has to be visible in advance. A migration report should separate:

Supported
Partial
Alternative
Not supported
Planned

No marketing "100% compatibility" when it can't be proven.

In the console wizard, the review step lists every skipped or partially mapped item before you apply anything — the per-source breakdown is below.

What moves, per source

Algolia

Indices: Collections — searchable/facet attributes and custom ranking are mapped from your settings export; field types are inferred from sample records.
Records: Bulk-upserted into the new collection; objectID becomes id and the rest of each record maps field by field.
Synonyms: Synonym sets per collection — regular, one-way and altCorrection types; placeholder synonyms are skipped and listed as warnings.
Rules: Curation sets (pins, hides, replace-query, filters) — only the first condition per rule is mapped; redirects land in curation metadata for manual follow-up.
API keys with ACLs: Not created by the wizard — set up scoped keys for the new collection separately, after import.
Stopwords: Not imported automatically — add stopword lists for the collection separately in the console.
Analytics / Insights: Not something to import — built-in search analytics and the behavioural events API start collecting from day one on the new collection.

Typesense

Collections schema: Recreate the collection in the console or via the API — the wizard parses Algolia exports; Typesense dumps carry documents, not settings.
Documents (JSONL): Typesense JSONL exports import directly.
Override rules: Recreate as curation rules — overrides aren't parsed from Typesense dumps.
Scoped API keys: Not migrated — create scoped keys with embedded filters for the new collection in the console.
—: Plus what Typesense Cloud lacks: experiments, quality lab, assistant.

Meilisearch / self-hosted

Indexes: Export documents to JSONL and import them in bulk.
Synonyms & stop-words: Recreate in the console — Meilisearch settings aren't parsed; the wizard maps Algolia synonym exports.
Manual ops burden: Gone — managed engine, quotas and backups included.
Webhooks: Signed outbound webhooks with retries and a delivery log.
—: Plus recommendations, personalization and the AI assistant.

Migration paths

Prove the quality before the cutover

Don't compare engines on a few hand-picked queries. Import a real query set and compare relevance in Quality Lab.

Run a benchmark

Run the wizard

The migration wizard is a real tool in the console, not a form — it lives at Dashboard → Migrate from Algolia. Sign up (or log in) and it opens directly: paste your export, review the mapped schema, synonyms and rules, then apply.

In practice

POST /v1/collections/products/documents/import

# JSONL documents import as-is
{"id":"123","title":"Running shoes","price":89.9,"in_stock":true}

# or import a dump in bulk:
aacsearch import products \
  --file algolia-projects.jsonl \
  --format jsonl

Questions

The catalogue import itself is hours. Mapping keys, synonyms and rules usually takes a day or two; teams that go live fastest run both engines side by side for a week and compare the analytics.

Explore more

Integrations Compare with alternatives SDKs & API

Test AACSearch on your own problem

The best way to evaluate a search platform isn't reading one more comparison table.

Upload your own data.
Run real queries.
Look at the Query Trace.
Compare relevance.
Work out the cost.

Only then make the decision.

Start free Open Playground See the documentation

Don't start by switching the old search off

The right order:

01
Inventory
Collect:
- indexes
- schema
- ranking
- synonyms
- rules
- facets
- filters
- frontend integrations
- events
02
Map
Match your current capabilities to AACSearch.
03
Import
Move the data and the settings across.
04
Benchmark
Run your real queries against both systems.
05
Shadow
Compare the answers before you switch any traffic.
06
Cutover
Move production over gradually.
07
Rollback
Keep a clear way back until the new system is confirmed.

A migration should show more than parity

If a feature works differently, that has to be visible in advance. A migration report should separate:

Supported

Partial

Alternative

Not supported

Planned

No marketing "100% compatibility" when it can't be proven.

In the console wizard, the review step lists every skipped or partially mapped item before you apply anything — the per-source breakdown is below.

What moves, per source

Algolia

Indices: Collections — searchable/facet attributes and custom ranking are mapped from your settings export; field types are inferred from sample records.
Records: Bulk-upserted into the new collection; objectID becomes id and the rest of each record maps field by field.
Synonyms: Synonym sets per collection — regular, one-way and altCorrection types; placeholder synonyms are skipped and listed as warnings.
Rules: Curation sets (pins, hides, replace-query, filters) — only the first condition per rule is mapped; redirects land in curation metadata for manual follow-up.
API keys with ACLs: Not created by the wizard — set up scoped keys for the new collection separately, after import.
Stopwords: Not imported automatically — add stopword lists for the collection separately in the console.
Analytics / Insights: Not something to import — built-in search analytics and the behavioural events API start collecting from day one on the new collection.

Typesense

Collections schema: Recreate the collection in the console or via the API — the wizard parses Algolia exports; Typesense dumps carry documents, not settings.
Documents (JSONL): Typesense JSONL exports import directly.
Override rules: Recreate as curation rules — overrides aren't parsed from Typesense dumps.
Scoped API keys: Not migrated — create scoped keys with embedded filters for the new collection in the console.
—: Plus what Typesense Cloud lacks: experiments, quality lab, assistant.

Meilisearch / self-hosted

Indexes: Export documents to JSONL and import them in bulk.
Synonyms & stop-words: Recreate in the console — Meilisearch settings aren't parsed; the wizard maps Algolia synonym exports.
Manual ops burden: Gone — managed engine, quotas and backups included.
Webhooks: Signed outbound webhooks with retries and a delivery log.
—: Plus recommendations, personalization and the AI assistant.

Run the wizard

# JSONL documents import as-is {"id":"123","title":"Running shoes","price":89.9,"in_stock":true} # or import a dump in bulk: aacsearch import products \ --file algolia-projects.jsonl \ --format jsonl

Questions

The catalogue import itself is hours. Mapping keys, synonyms and rules usually takes a day or two; teams that go live fastest run both engines side by side for a week and compare the analytics.

Don't start by switching the old search off

Inventory

Map

Import

Benchmark

Shadow

Cutover

Rollback

A migration should show more than parity

What moves, per source

Algolia

Typesense

Meilisearch / self-hosted

Migration paths

Prove the quality before the cutover

Run the wizard

In practice

Questions

How long does a migration take?

Will there be downtime?

Do we have to change our documents?

What do we gain beyond parity?

What doesn't the wizard migrate automatically?

Test AACSearch on your own problem

Don't start by switching the old search off

Inventory

Map

Import

Benchmark

Shadow

Cutover

Rollback

A migration should show more than parity

What moves, per source

Algolia

Typesense

Meilisearch / self-hosted

Migration paths

Prove the quality before the cutover

Run the wizard

In practice

Questions

How long does a migration take?

Will there be downtime?

Do we have to change our documents?

What do we gain beyond parity?

What doesn't the wizard migrate automatically?

Test AACSearch on your own problem