Workbench — Sovereign LLM

New chat

Your question appears as Ask; the reply appears as Response after you send.

Type a prompt — chat, research, RAG, and workbench tools are selected automatically

Presentation mode

Export audience-themed slide decks (PPTX, HTML, or ZIP bundle) grounded in your library when RAG is on.

Audience Infographics

Answer from

Product questions use platform context automatically. Enable RAG for uploaded documents; add Exact match only to answer strictly from matching text (no general AI guesses when nothing is found).

Use knowledge base (RAG) Exact match only Stream response Tool calling (KB search)

Learning profile

Platform applications

Extended applications

Governance assurance, audit archives, enterprise administration, and the advanced workbench — sovereign AI operations in one place.

Advanced workbench

KB upload, diagrams, reports, batch tools

Governance hub

Frameworks, risks, OWASP LLM, and GRC repository

Assurance dashboard

EU AI Act, OWASP LLM, and unified trust index

Audit archive

WORM seals and tamper-evident logs

Enterprise account

Global billing, usage, SSO/SAML, branding, ML hub

Implementation tracker

Live control gaps for production sign-off

Production maturity

Operational readiness and assurance scores

Platform capabilities

Live API and feature inventory

Enterprise SSO

OIDC, SAML, and OAuth federation guide

Deep research runs multi-step RAG + LLM analysis with structured findings. Uses your knowledge base; enable ENABLE_WEB_SEARCH on the server for external sources.

Your topic appears as Ask; findings appear as Response after you start.

Account & platform

Dashboard

Use the sidebar for workspace chat, profile, and billing. Platform destinations (Enterprise ML, Governance, Assurance, Capabilities) are in the Platform section of the sidebar, or open Enterprise account for billing and SSO.

— —

Account panels (Upgrade, Profile, Billing) are in the sidebar under Account. Platform apps (Enterprise ML, Governance, Assurance, Capabilities) are under Platform.

Conversations

Session history

Chats and auto-routed research sessions. Select any row to reopen the conversation in New chat.

Sessions—

Metering

Usage analytics

Per-request token metering and Enterprise MLOps chargeback. Click any summary card for a breakdown — LLM calls and corpus operations appear in the table below.

Requests—

Tokens—

Est. cost—

Date	Type	Repository	Model / Op	Tokens / Units	Cost
Loading…

FinOps

Spending & limits

Plan entitlements, included usage, on-demand controls, and MLOps chargeback by repository.

Loading…

Finance

Billing & invoices

Subscription plan, payment method, invoice history, and on-demand settings — aligned with sovereign deployment billing.

Loading…

Programmatic access

API keys

Cursor, VS Code, CI, and automation. Send the secret once per request as X-API-Key. Requires an authorized enrolled account.

Active keys—

Quota—Per account

Auth headerX-API-Key

Listed values are identifiers only (swb_… + ellipsis), not the full secret. Use Generate API key, Browser connect, or New secret on IDE Integration keys to get a copyable secret once.

Name	Identifier	Last used	Created	Actions

Account

Profile

Your signed-in identity, plan entitlements, and shortcuts to API keys and enterprise administration.

Signed in

Plan:

Developer experience

IDE & integrations

Connect Cursor or VS Code — sidebar chat, MCP Agent tools, and auto-provisioned API keys. No cloud model configuration required.

Editor setup

Sovereign LLM Workbench extension

v1.1.9 · Cursor, VS Code, VSCodium · Activity bar chat + MCP sovereign_chat

Click Install in Cursor — provisions your IDE API key
Run the downloaded sovereign-cursor-setup.bat (installs extension via CLI, not Visual Studio)
Reload Cursor · chat in Sovereign LLM sidebar · Agent uses MCP automatically

Install in Cursor (one-click) VS Code connect Setup script (.ps1) VSIX (CLI)

Windows: do not double-click the .vsix file — it opens Visual Studio's installer, not Cursor. Use Install in Cursor or cursor --install-extension.

Advanced — MCP bundle (Continue, Cline, Aider)

ZIP with .continue/config.yaml, .vscode/settings.json, and MCP — use after Cursor one-click or for non-Cursor editors.

Usage & metering

Per-user IDE connections, API requests, token estimates, spend per request, and sessions.

Loading usage…

Active sessions

Web (browser portal) and device (Cursor / VS Code / API) sessions — created, status, actions. Revoke to sign out a device or browser.

Show revoked

Loading sessions…

Integration API keys

Keys inherit your account permissions (chat, RAG, uploads, reports).

Copy now — shown once only:

Prefix column is not the full key — use Generate key, New secret on IDE Integration, or Browser connect.

Name	Identifier	Created

Loading plans…

Preferences

Personalization

Theme, language, and default chat behavior — including RAG chunking and embedding model overrides.

Appearance

Theme

Language

Display language

Used for chat attachment analysis and multilingual replies when supported.

Chat defaults

Default to RAG in new chats Default to exact match only (KB terms must appear in documents)

RAG tuning

Minimum relevance score (0–1)

Higher values require stronger semantic similarity. Exact match mode also requires query terms in the text.

Chunk size (tokens) Top-K results Embedding model

Security & compliance

Settings

Privacy controls, multi-tenant scoping, GDPR consent, and two-factor authentication for your workbench account.

Privacy

Allow usage analytics Marketing emails Retain chat history

Organization (multi-tenant)

Scope KB and Chroma to your tenant. Send header X-Tenant-ID.

Tenant ID

Consent

Analytics consent (GDPR)

Two-factor (TOTP)

Add an authenticator app for sign-in verification.