First Class Modules
Global secured search
Searches permitted records across modules and also uses permission-filtered vector retrieval to find meaningfully related information in indexed uploads, PDFs, images, screenshots, tables, diagrams, and text.
module.global-searchUse Global Search as the universal finder when users need to locate customers, files, work, conversations, events, custom records, or related evidence from uploaded content in one place.
The bundled renderer key is module.global-search. First-class means BuildWithHQ supplies a native, typed, secured runtime experience inside normal app provisioning and the signed-in user's existing permissions.
What it does
Searches permitted records across modules and also uses permission-filtered vector retrieval to find meaningfully related information in indexed uploads, PDFs, images, screenshots, tables, diagrams, and text.
Key capabilities
- Run bounded text search across authorized modules and records.
- Search eligible indexed uploads, PDFs, images, screenshots, tables, diagrams, and extracted text in a shared vector space.
- Find conceptually related information even when the source uses different words from the query.
- Retrieve the original text, page, image, and source identity so results can retain evidence and citations.
- Filter by module and page through stable result envelopes.
- Show safe matched-field snippets plus module, status, location, relation, activity, and update context.
- Add or remove favorites without leaving the result list.
How vector and AI search improve discovery
Global Search has two complementary secured paths. Operational record search is best for exact names, identifiers, fields, statuses, and module records. The AI search path vectorizes eligible indexed content so a query can match by meaning, not only by identical keywords.
| Search path | What it finds | Why it helps |
|---|---|---|
| Record and keyword | Exact or close text in permitted records and fields | Fast, predictable lookup for known business data. |
| Semantic vector | Related passages and documents that express the same idea with different wording | Surfaces useful context that keyword-only search can miss. |
| Multimodal vector | Relevant uploads, PDF pages, images, screenshots, charts, tables, and diagrams | Lets a text question find visual evidence, including information that weak OCR or text extraction may lose. |
| AI relevance and reranking | The strongest authorized candidates, with their original source and page context | Brings more closely related information toward the top while preserving evidence for review. |
For example, a search for declining renewal risk can surface a customer record, a differently worded conversation, a chart on an uploaded PDF page, and a screenshot containing relevant evidence—provided each source is indexed, AI-eligible, and visible to the current user.
AI relevance is a discovery aid, not proof that every result is correct. Users can open the original source, page, or visual evidence to verify the result in context.
Common uses
- App-wide command and search experiences.
- Finding a record when its owning module is unknown.
- Discovering related material across records and uploaded documents.
- Finding a PDF page, image, screenshot, chart, table, or diagram from a natural-language description.
- Cross-module navigation in relationship-heavy applications.
How it connects
Record results point to the canonical record in its owning module. AI search uses the platform's secured source, document-element, chunk, visual-index, and vector retrieval chain; it returns original source/page/asset references instead of copying the evidence into a new business record.
Where applicable, its records use the universal RecordId conventions so they can participate in secured relationships, activity history, favorites, dynamic fields, notifications, Inbox attention, and global search without copying the source record.
Security and data boundary
The server applies SaaS, tenant, CanAiReadRecords, DataRole, location, record, field, and AI-eligibility rules before vector candidates or visual evidence can be returned. Unindexed, deleted, AI-excluded, or otherwise unauthorized content is absent from results. Relevance never grants access, and counts, snippets, citations, and related context must not reveal inaccessible records or fields.
- The authenticated service derives the SaaS app, tenant account, user, DataRole, and location scope; browser identifiers are never authorization proof.
- The page editor composes React components with validated data bindings. Those bindings call typed runtime APIs, whose application services execute reviewed stored procedures.
- List, search, detail, relation, activity, favorite, and write operations reapply their required server-side permissions.
Add it to an app
- Open the SaaS app in the Developer Console and identify the user journey and page where this module belongs.
- Add the validated
module.global-searchmodule block through the supported page/template authoring flow. - Configure the module with the page editor's React components and validated data bindings; the bindings call authenticated platform APIs backed by reviewed stored procedures.
- Place the page in the correct user-type menus and assign existing DataRole, record, field, and location permissions.
- Test list, detail, search, empty, denied, stale-update, and cross-location behavior before publishing an exact version.
Keep module-specific filters and lists for focused workflows. Use Global Search as a fast cross-module entry point, and clearly distinguish exact record matches from AI-ranked related evidence so users understand why each result appeared.
A renderer being bundled in the tenant application does not make its data visible in every app. The server returns only components and records authorized for the current app and signed-in user; unavailable or unauthorized blocks fail closed.