Skip to main content

Settings

The Settings panel lets you configure authentication, tune response behavior, and manage per-nexset filter schemas. Open it by clicking the gear icon in the top bar or pressing Ctrl/Cmd+,.

Introduction​

Settings is organized into three tabs covering distinct configuration scopes: Authentication manages your service key and connection state, Response Tuning controls model output behavior such as temperature, and Filters lets you register and manage per-nexset filter schemas that govern access control and pre-retrieval filtering. The Sidebar — outside the Settings panel itself — handles per-session credential, model, and nexset selection.

Filter management in this UI mirrors the programmatic Filter Registration API. Changes made in either surface take effect for subsequent queries.

Authentication Tab​

The Authentication tab manages your connection to the GenAI RAG service.

  • Service Key field — paste your Nexla service key (password-type input, paste-friendly)
  • Connect button — validates and stores the key. On success, a green Authenticated badge appears
  • Sign out — clears the stored key and resets the session
  • "Where do I find my key?" link — opens dataops.nexla.io/settings/authentication where you can create a service key
Session Storage

The service key is stored in your browser's session storage — it survives page refreshes but is cleared when you close the tab.

Response Tuning Tab​

The Response Tuning tab controls how the LLM generates answers.

Temperature​

A slider from 0 to 1 that controls randomness in the model's output:

ValueBehavior
0–0.3 (Precise)More factual, deterministic answers. Best for data lookup and factual queries
0.4–0.7 (Balanced)Mix of accuracy and expressiveness. Good for most queries
0.8–1.0 (Creative)More varied and creative responses. Best for exploratory or open-ended questions

Lower values produce consistent, factual answers. Higher values allow more creative but potentially less precise responses.

Filters Tab​

The Filters tab lets you register, view, and manage filter schemas for your nexsets. Filter schemas define how user context (access rules, ownership scope, and session filters) maps to metadata filters at query time. For the full API reference, see Filter Registration API.

Select a Nexset​

Use the nexset filter picker dropdown to select which nexset's filter schema to manage.

Filter Type Sections​

Three sections display the registered filter keys for the selected nexset:

TypePurpose
Access RulesPolicy-level gates (e.g., "can this user access confidential docs?")
Access ScopeData ownership scope (e.g., "only docs from this tenant"). Values use IN semantics
FiltersPage or session context (e.g., "only docs from this project")

Each section lists filter keys with their operator badge, nexset field mapping, description, and edit/delete buttons.

Add a Filter​

Click Add filter within any section. The inline editor prompts for:

FieldDescription
KeyMetadata field name to filter on
OperatorComparison operator (EQ, NEQ, IN, NOT_IN, GT, GTE, LT, LTE, CONTAINS, NOT_CONTAINS, EXISTS, NOT_EXISTS, BETWEEN)
Nexset fieldBackend field name if it differs from the key (e.g., map tenant_id to org_tenant)
TypeWhich filter layer to register under (Access Rules, Access Scope, or Filters)
DescriptionHuman-readable description (optional)

Edit a Filter​

Click the pencil icon on any filter key to open the inline editor. You can change the operator, nexset field, type, and description. Key names cannot be renamed — delete and re-add to change a key name.

Delete a Filter​

Click the trash icon on any filter key. A two-click confirmation prevents accidental deletion: click once to arm, then click again within 2 seconds to confirm.

Refresh Cache​

Click Refresh cache to invalidate the Redis-backed filter cache for the selected nexset. The next query reloads the filter schema from the database. Use this after making out-of-band changes to filter configuration.

The sidebar provides the primary controls for configuring each query session.

LLM Credential Picker​

Select which LLM credential to use.

LLM Model Selector​

After selecting a credential, the model dropdown populates with available models for that provider, ranked by capability. The top-ranked model is auto-selected.

Embedding Configuration (Optional)​

Toggle the Embeddings switch to enable custom embedding configuration:

  • Embedding Credential — select a credential with embedding capability (typically OpenAI or Azure)
  • Embedding Model — choose from available models (e.g., text-embedding-3-small at 1536 dimensions)

When embedding is disabled, the system infers the embedding model per-nexset from the vector store configuration.

Nexset Selector​

Search and select the nexsets to query:

  • Search — type-ahead search by name or ID
  • Multi-select — click to toggle nexsets on/off
  • Chips — selected nexsets appear as removable chips with a "Clear all" button
  • Collapse/expand — toggle button in the top bar; state persists across sessions
  • Resize — drag the right edge on desktop
  • Mobile — swipe from left edge to open, swipe left to close