Stand up a self-hostable, typo-tolerant search environment with Typesense โ the open-source Algolia / ElasticSearch alternative (single C++ binary, <50ms instant search, no runtime deps). One routing-first skill: pick a server mode (binary download, official Docker image, or managed Typesense Cloud), install an API client (Python/JS/PHP/Ruby official; Go/Dart/C# community), design a collection schema, index documents, and run searches with typo tolerance, faceting/filtering, geo-search, sorting, grouping, synonyms, curation, scoped API keys, and federated multi-search โ then wire an InstantSearch.js UI and a Raft-based HA cluster for production. Use when the user wants to build or operate an installable search backend, add site/app/product search, or migrate off Algolia/Elasticsearch. Triggers on: typesense, search engine, typo-tolerant search, algolia alternative, elasticsearch alternative, instantsearch, faceted search, geo search, vector search, self-hosted search, site search, product search.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
Stand up a self-hostable, typo-tolerant search environment with Typesense โ the open-source Algolia / ElasticSearch alternative (single C++ binary, <50ms instant search, no runtime deps). One routing-first skill: pick a server mode (binary download, official Docker image, or managed Typesense Cloud), install an API client (Python/JS/PHP/Ruby official; Go/Dart/C# community), design a collection schema, index documents, and run searches with typo tolerance, faceting/filtering, geo-search, sorting, grouping, synonyms, curation, scoped API keys, and federated multi-search โ then wire an InstantSearch.js UI and a Raft-based HA cluster for production. Use when the user wants to build or operate an installable search backend, add site/app/product search, or migrate off Algolia/Elasticsearch. Triggers on: typesense, search engine, typo-tolerant search, algolia alternative, elasticsearch alternative, instantsearch, faceted search, geo search, vector search, self-hosted search, site search, product search.
allowed-tools
Bash Read Write Edit Glob Grep WebFetch
compatibility
Cross-platform search-backend wrapper usable from Claude Code, Codex, Gemini CLI, and OpenCode. Wraps the upstream Typesense server (binary / Docker / Cloud) and official API clients. Routes LLM-trace/eval observability to `opik`/`langsmith`, agent-facing code search to `semble`, and broad service telemetry to `monitoring-observability`.
Typesense is a fast, typo-tolerant
open-source search engine โ an Algolia alternative and an easier-to-use
ElasticSearch alternative. It is a single C++ binary with no runtime
dependencies, architected for low-latency (<50ms) instant search. This skill
is the routing-first wrapper: choose how to run the server, wire a client,
model the data, and drive search + UI + production hardening.
When to use this skill
The user wants to stand up a search backend for a site, app, catalog,
docs, or product browsing experience
The user asks to install/run Typesense (binary, Docker, or Typesense Cloud)
The user wants typo tolerance, faceting/filtering, geo-search, sorting,
grouping, synonyms, curation, scoped API keys, or federated multi-search
The user wants to migrate off Algolia or Elasticsearch to a self-hosted
or managed open-source engine
The user wants an InstantSearch.js UI or a Raft HA cluster in front
of / around Typesense
When not to use this skill
The user wants LLM trace/eval observability (hallucination, prompt
scoring) โ use opik / langsmith
The user wants token-efficient code search for agents over a repo โ
use semble
The user wants generic service dashboards / uptime alerts (non-search
telemetry) โ use monitoring-observability
The user wants a vector database purpose-built for embeddings only โ
Typesense does vector + hybrid search, but a dedicated store may fit better
for pure ANN at extreme scale; confirm the workload first
Prerequisites
Requirement
Notes
Docker (recommended)
Simplest local + prod path via the official image
or a binary host
Linux (x86-64) / macOS binary packages from typesense.org/downloads
or Typesense Cloud
Zero-ops managed cluster (fixed hourly + bandwidth, not per-record)
An API client
Python / JS / PHP / Ruby official; Go / Dart / C# community
An API key
Set at server start (--api-key); generate scoped keys per tenant
Prefer an official client over raw CURL โ they ship a smart retry strategy
for HA setups. See references/commands.md for the
full client + integration matrix.
Step 3 โ Design the collection schema
A collection is an index with a typed schema. Mark fields facet: true to
filter/drill-down, and set default_sorting_field for ranking:
Unlike Algolia, most settings (searchable fields, facets, ranking) are set at
query time, so one collection serves many sort orders โ less memory, more
flexibility.
client.collections["companies"].documents.search({
"q": "stork", # typo of "stark" โ handled out of the box"query_by": "company_name",
"filter_by": "num_employees:>100",
"sort_by": "num_employees:desc",
"facet_by": "country",
})
Capabilities to reach for: faceting/filtering, geo-search (sort by distance),
grouping & distinct, synonyms, curation/merchandizing (pin records),
federated multi-search across collections in one request, and vector /
hybrid search. Details in references/commands.md.
This skill folder is plugin-installable through the standard jeo-skills
flow so the wrapper, references, and installer land on disk for any supported
agent runtime:
# Project install (writes into .agents/skills/typesense/)
npx skills add https://github.com/akillness/jeo-skills --skill typesense
# Global install for every detected agent
npx skills add -g https://github.com/akillness/jeo-skills --skill typesense
# Target specific agents
npx skills add -g https://github.com/akillness/jeo-skills --skill typesense -a claude-code -a codex -y
Output format
When the user asks typesense for help, return a compact brief:
# typesense Routing Brief## Scope- Server mode: docker | binary | cloud | undecided
- Client: python | js | php | ruby | community
- Stage: install-server | install-client | schema-design | index | search | ui | production-ha
## Recommended next move- start-docker-server | install-client | create-collection | import-docs | run-search | wire-instantsearch | scoped-keys | cluster
## Why- 2-3 bullets grounded in the user's packet
## Route-outs-`opik` / `langsmith` for LLM trace/eval observability
-`semble` for agent-facing code search over a repo
-`monitoring-observability` for non-search service telemetry
Best practices
Pin a version tag, never latest โ typesense/typesense:27.1, and
keep the data dir on a real volume so restarts don't lose the index.
Set settings at query time โ searchable fields, facets, sort, and
ranking are per-query; you rarely need multiple collections for sort orders.
Mark facets in the schema โ facet: true is required for filtering /
drill-down on a field.
Use scoped API keys for clients โ the admin key stays server-side;
scoped keys enforce per-tenant record access.
Bulk import as JSONL with upsert โ far faster than per-document
creates for large datasets; size RAM to the index (memory-resident).
License awareness โ the server is GPL, the client libraries are
Apache-2.0; run the server as a separate daemon (the intended use).
