Complete Feature List

Standard email/password registration and login with password strength validation and hashed credentials.

6-digit OTP code sent to users' email with 120-second countdown, auto-focus inputs, and resend support.

Secure session cookies with SHA-256 hashed tokens stored in PostgreSQL. Logout invalidates the token server-side.

Sign in with Google or link an existing account to Google. CSRF-protected with state cookies.

Create, rename, and delete organizations. Each organization is a fully isolated workspace.

Invite team members, assign roles (Owner, Admin, Editor, Viewer), and manage permissions. Plan limits enforced at creation.

Four roles with granular permission matrices. Owners have full control; Viewers have read-only access to search and analytics.

Invite teammates who haven't registered yet. Lunexa stores a pending invite, sends a branded email with a single-use link (7-day expiry), and auto-joins the invitee to the org when they finish registering.

Create, list, rename, and delete projects. Plan limits enforced at creation (e.g., 1 project on Free, 3 on Starter, 50 on Pro, 200 on Enterprise).

Every project gets a unique collection prefix, ensuring complete data isolation between projects.

Create collections with custom schemas supporting string, int32, int64, float, bool, string[], int32[], int64[], float[], bool[], auto, and geopoint field types.

Paste a collection schema JSON to auto-populate field definitions when creating a new collection.

View document count, index size, and basic health metrics for each collection.

Opt-in support for hierarchical documents with object and object[] field types. Query sub-fields with dot notation (e.g. author.name) and facet on nested attributes. Enabled per-collection at creation time; immutable afterwards. Useful for product variants, author bios, location blocks, and other structured sub-records.

Create, read, update, patch, and delete individual documents. Every create and delete increments/decrements the records usage counter for billing.

Import thousands of documents at once via JSONL upload. Supports create, upsert, update, and emplace actions. Line count tracked for billing.

Export an entire collection as JSONL for backup, migration, or downstream processing.

Generate realistic sample documents for your collection using AI. Useful for testing and demos.

Full-text search with typo tolerance, filters, sorting, and pagination. Every request is logged for analytics and billing.

Execute multiple queries across different collections in a single request. Ideal for federated search interfaces.

Ask questions in plain English. An LLM converts your query into structured filter_by and sort_by parameters automatically.

Generate time-limited, CORS-safe, read-only API keys for frontend search widgets. No admin credentials ever reach the browser.

A built-in search playground in the dashboard with an NL mode toggle and a panel showing how the AI interpreted your query.

Define bidirectional synonyms (car ↔ automobile) or one-way root mappings. Improves recall for user queries that use different terms than your data.

Pin specific documents at fixed positions for certain queries, exclude documents, or apply custom filters. Rule types: exact, contains, starts_with, ends_with.

Define words to ignore during search. Multi-language support: English, Spanish, French, German, Italian, Portuguese, Japanese, Chinese.

Save reusable search parameter sets (query_by, filter_by, sort_by, facet_by) under a named preset. Clients can reference presets by name.

Create virtual names that point to collections (e.g., products_latest → products_v2). Enables zero-downtime reindexing.

Define seed URLs, crawl depth, max pages, extraction mode, and cron schedule for recurring crawls.

Lunexa crawls a few sample pages and uses AI to propose the best collection schema — field names, types, facets, sample values — all automatically.

Before indexing, review detected page patterns, the proposed schema, sample extracted documents, and estimated storage/monthly cost.

Content hashes track which pages have changed since the last crawl. Unchanged pages are skipped, reducing cost and load on your site.

Use cron expressions to schedule recurring crawls (every hour, every 6 hours, daily, weekly, custom). A background worker checks every minute.

Every run shows live status messages, page counts, and a Stop button. Expandable run history with per-run details and error lists.

Preview the monthly cost before starting a crawl based on your plan's included pages and overage rate.

Create, list, update, and delete NL search models per project. Supported: Google Gemini 2.5 Flash, Gemini 2.5 Pro, Gemma 4 31B.

The LLM provider API key is stored centrally in Lunexa and injected into the model config server-side. Users never provide raw keys.

Add a custom system prompt to any NL model to steer the AI toward your domain's terminology and preferred response style.

Interactive calculator showing monthly NL search cost across all plan tiers based on expected query volume.

22 distinct action scopes across 6 categories: documents, collections, synonyms, curations, stopwords, aliases, presets, analytics, NL search, and more.

Pre-built templates for common use cases: Search-Only (frontend safe), Admin (full access), Write (indexing/sync), Analytics (read-only monitoring), Custom.

Restrict an API key to specific collections. Wildcard patterns are supported for easy scoping.

Limit a key's usage to specific HTTP Referer headers. Combined with CORS, this protects frontend keys from abuse.

Set an expiration time (minutes, hours, days) and revoke any key instantly. Full keys are returned only once at creation and stored as hashes.

Per-key rate limits (queries per hour) prevent runaway clients from exhausting your plan quota.

Time-series chart of search requests over custom date ranges. Total record count and month-to-date usage per project.

Paginated log of every search request: query text, query_by fields, filters, sort, result count, latency, HTTP status, user-agent.

Average, P90, and P99 response times plotted over time. Identify latency regressions before your users do.

Surface queries that returned zero hits. Use this to improve synonyms, expand your dataset, or refine curations.

Breakdown of search operations by type and endpoint over a selected date range.

Health check showing search engine connectivity and collection availability for the project.

Define aggregation rules that compute popular queries, no-hits queries, or custom counters across source collections and write them to a destination collection.

View current subscription, upgrade/downgrade plans, and see available packages with pricing and feature limits.

Tracks four metered counters: search requests, records stored, crawler pages, and NL searches. All counters are surfaced in the dashboard and reported to Stripe.

Upgrade plans in-app with Stripe's embedded checkout component. No redirect required.

Redirect users to Stripe's hosted portal to manage payment methods, download invoices, and update billing details.

List all past invoices with amount, date, status, and PDF download links.

Add and remove payment methods directly from the dashboard. Uses Stripe SetupIntent for secure card collection.

A background worker reports search, crawler, and NL search usage to Stripe every 5 minutes as meter events. Overages billed automatically.

Sign in with Google and link/unlink Google accounts. Optional — requires GOOGLE_CLIENT_ID and GOOGLE_CLIENT_SECRET.

Full subscription and metered-billing integration. Optional — requires STRIPE_SECRET_KEY. Gracefully disabled when not configured.

Powers Smart Crawler and Natural Language Search. Optional — requires GEMINI_API_KEY. Features gracefully disabled when not configured.