Sally Enterprise
The AI office suite that runs on your own servers
Bring AI into your corporate network end to end: data never leaves your domain, models stay under your control, and every interaction is auditable. Employees get AI right inside the Word, Excel, PowerPoint, Outlook and WPS apps they already use, while IT and security teams keep full oversight.
Built for enterprise-scale procurement
Data stays inside your network — security starts with the architecture
Sally Enterprise runs as a self-hosted web service inside your corporate network. Business data, documents and chat history live in your own database and storage, and model requests are routed through your own gateway — no third-party SaaS involved at any point.
- All data is stored in your own PostgreSQL database, managed under your existing backup policies
- Employee devices only reach internal service addresses — business requests never cross your network boundary
- Local LLMs such as Ollama are supported, so AI inference can run fully offline
- A unified model gateway holds every model API key — invisible and non-exportable to end users
- Independent persistent data volumes make backup, migration and archiving simple
Nine capabilities your security and compliance teams will love
Every capability maps to a real product mechanism you can verify item by item after deployment
Data sovereignty
Documents, conversations and knowledge-base content are all stored in your own PostgreSQL database and data directories — fully owned and controlled by your organization.
End-to-end auditing
Every conversation and every model call leaves a trace. Admins can replay the full conversation context tree and LLM call history from the admin console.
Accountable usage
Token consumption, call counts and response times are tracked per user and per model, so departmental cost allocation is backed by hard numbers.
Least-privilege access
JWT authentication, tiered admin roles and one-click account activation/deactivation keep sensitive operations restricted to authorized administrators.
Hosted model keys
API keys for every model provider are configured and stored centrally by administrators — completely invisible to end users.
Controlled onboarding
Admins can disable public sign-up and switch to invitation-based account creation, keeping unauthorized people out.
Encrypted transport
HTTPS/SSL deployment and internal network isolation fit right into your existing security policies.
Model whitelisting
Admins decide which models everyone can use — down to local-only models for zero data egress.
No telemetry callbacks
The self-hosted service sends no business data back to the vendor; you monitor its health on your own terms.
One admin console, full control
A ready-to-use web admin console — IT admins handle day-to-day operations without touching the command line
Overview & analytics
Users & permissions
AI capabilities
Branding & distribution
Zero learning curve — AI inside the tools employees already know
Enterprise governance on top, a first-class employee experience underneath — call AI directly inside Office without switching windows
Word AI writing assistant
Polish, rewrite, LaTeX formula conversion, track changes and one-click formatting — all inside your Word documents.
Excel AI data analysis
Natural-language formulas, Python data analysis and automatic charting — spreadsheet work at double speed.
PowerPoint AI presentations
Generate slide decks from documents in one click with professionally matched templates — hours of deck-building reduced to minutes.
Outlook AI email assistant
Smart replies, email summaries and automatic calendar entries make business communication faster and more professional.
Web chat workspace
A browser-based AI workspace with agents, knowledge-base Q&A and document analysis — no installation required.
Cross-platform coverage
Works with Microsoft Office, WPS and OnlyOffice alike, fitting your existing office environment with no forced migration.
Consumer cloud service vs Sally Enterprise
Three steps to deployment — try it the same day
Standardized Docker delivery with no complex environment customization
Launch the services
Run docker compose up -d on a prepared Linux server — the application and PostgreSQL database come up automatically.
Initialize the admin
An admin account is created on first launch; sign in to the console to finish site configuration and model onboarding.
Distribute the add-ins
Push the Office / WPS / OnlyOffice add-ins to everyone from the distribution center — employees can start immediately.
What's included
- Enterprise server (with unified model gateway)
- Web admin console (17+ admin modules)
- Web chat workspace (11 interface languages)
- Office / WPS / OnlyOffice add-ins
- PostgreSQL database and deployment orchestration files
- Deployment documentation and technical support
Environment requirements
- A Linux server (bare metal, VM or private cloud)
- Docker and Docker Compose
- PostgreSQL 16 (bundled with the deployment package)
- Offline capable: pairs with on-premises Ollama models
Built for data-sensitive industries
For any organization with strict data-boundary requirements, self-hosting is the dependable choice
Finance & law firms
Contract review, legal drafting and research reports. Client and transaction data stays inside the firm — exactly what heavily regulated industries require.
Government & public sector
Official correspondence, drafting and reporting. Deployed on isolated internal networks with local models so nothing ever leaves the premises.
Research & academia
Paper writing, literature search, Zotero reference management and data analysis. Research data and findings stay under your control.
Healthcare & hospitals
Clinical documentation and report preparation. Patient data never leaves the hospital network — privacy protected by architecture.
Manufacturing & R&D
Knowledge-base Q&A and technical documentation. Process and R&D know-how becomes a private corporate asset, not third-party platform data.
Consulting & professional services
Data analysis, reporting and proposal writing. Deliverables stay controllable and auditable, with client data kept isolated.
From POC to go-live, someone owns every step
A clear delivery process and ongoing technical support — procurement without second-guessing
Requirements assessment
We review your office environment and data-security requirements and confirm the deployment plan.
POC validation
A quick trial deployment in your own environment — validate the core scenarios before committing.
Deployment & implementation
Hands-on help with production deployment, model onboarding and security policy configuration.
Training & handover
Admin console training and end-user training, plus complete operations documentation.
Go-live acceptance
Item-by-item acceptance against functionality and security baselines for a smooth launch.
Ongoing support
A dedicated support channel for issues, plus continuous version upgrades and feature updates.
Frequently asked questions
Does the deployment need internet access?+
No. Sally Enterprise runs entirely inside your network. Paired with local LLMs such as Ollama, AI inference runs fully offline — the whole system works in a physically air-gapped network.
Where is our data stored, and who owns it?+
All data — users, conversations, documents, knowledge base and audit logs — is stored in your own PostgreSQL database and data directories. Your organization owns it completely and manages it under your own backup and retention policies.
Which models are supported? Can we use local models?+
Any OpenAI-compatible model works, including OpenAI, Anthropic, Gemini, DeepSeek, Qwen, GLM and Mistral — plus self-hosted models via Ollama or vLLM. Admins configure models centrally and can add or remove them anytime.
What server specifications do we need?+
A standard deployment needs one Linux server with Docker; PostgreSQL 16 is bundled. For local LLMs, size your GPU server to the models you choose — cloud model APIs need no GPU at all.
How does licensing and pricing work?+
Per-enterprise licensing — one license, long-term use, no per-seat subscription. Contact us by email or WhatsApp and we'll put together a proposal based on your organization's size and deployment shape.
How is this different from the desktop personal edition?+
The personal edition targets individuals, storing data on personal devices and using cloud services. The enterprise edition runs on your own servers and adds full user management, audit logs, usage analytics, model gateway governance and brand customization — plus Office/WPS add-ins for the entire workforce.
Is it hard for employees to learn?+
Employees use AI inside the Word, Excel, PowerPoint, Outlook and WPS interfaces they already know — almost no learning curve. We provide admin and end-user training along with complete operations documentation.
How do upgrades work?+
The enterprise edition keeps evolving. Upgrades ship as standardized delivery packages, and the independent data-volume design means upgrades never touch your business data. Enterprise support customers get new versions and upgrade assistance first.
Why enterprises need a self-hosted AI office suite
Your data boundary is non-negotiable
Documents, contracts and research reports carry your core IP and personal information. Public cloud AI means that data leaves your boundary and passes through third parties. Self-hosting runs AI where the data already lives, eliminating leakage risk by architecture rather than remediation.
Auditable is what passes compliance
Internal audits, certifications and sector regulators all require traceability of system activity. Sally Enterprise records every conversation and model call, and admins can replay the full context — making AI usage as auditable and accountable as any other information system.
Models under your control
Models are a procurement decision, not vendor lock-in. The unified model gateway lets you switch between cloud and local models anytime: cost-effective models for everyday work, local-only for sensitive tasks, with keys held centrally and invisible to end users.
Predictable costs at scale
Per-seat SaaS subscriptions balloon as you grow. A per-enterprise license plus built-in usage analytics lets you account for AI costs by department and optimize your model mix — keeping long-term cost of ownership firmly under control.