> For clean Markdown of any page, append .md to the page URL. > For a complete documentation index, see https://developers.alephant.io/llms.txt. > For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://developers.alephant.io/_mcp/server. # Chat Preferences > Configure organization-level chat behavior, default assistant capabilities, privacy protection, retention, and user-facing chat controls. Chat Preferences is the administrator control surface for the default chat experience. It governs organization-wide defaults that affect new chats, the default assistant, privacy protection, history visibility, file limits, retention, and selected user-facing controls. Use this page for deployment-level behavior. It does not replace model credentials, Agent permissions, tool authorization, connector access, or the Privacy Dictionary; those controls remain in their own configuration pages and are enforced independently. ## Capability Boundary | Area | What You Can Configure | Boundary | | --------------------- | ------------------------------------------------------------------------------------------ | -------------------------------------------------------------------------------------------------------------------------- | | Organization identity | Company name and company description used as deployment context. | Does not grant access to documents, tools, or models. | | Default assistant | Default system prompt and tools available to the built-in assistant. | Custom Agents keep their own prompt, knowledge, model, and tool configuration. | | Privacy protection | Whether chats and Agent sessions detect and mask sensitive information before model calls. | Dictionary actions and route readiness still decide whether content is sanitized, local-only, blocked, or review-required. | | History and retention | Query-history mode, admin-panel visibility, and maximum chat retention. | Hiding the admin page is visibility-only; data recording depends on the selected history mode. | | User experience | Anonymous access, auto-scroll, search mode, file limits, and selected chat UI defaults. | These settings do not override authentication, role, or group permissions. | ## Main Sections | Section | Use It To | | --------------------- | ------------------------------------------------------------------------------------------------------------------------------ | | Company information | Set the organization name and description that help the assistant understand the deployment context. | | Privacy Protection | Choose the organization policy, inspect readiness, and open the Privacy Dictionary summary. | | Default Assistant | Edit the default assistant system prompt and enable built-in tools, connector-backed tools, MCP tools, and OpenAPI tools. | | History and retention | Decide how query history is recorded, whether the admin sidebar shows Query History, and how long inactive chats are retained. | | Access and experience | Control anonymous access, auto-scroll behavior, search mode availability, and related chat defaults. | | File limits | Set maximum user-upload size and file token threshold for chat attachments. | ## Privacy Protection | Policy | Behavior | Notes | | -------------------------- | --------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------- | | Force on | Enhanced privacy is enabled for every new chat and Agent session. | Use this for deployments where sensitive-content masking must stay organization-controlled. | | Allow employees to control | New sessions start protected, and users may change only the current session. | Use this when teams need per-session flexibility but secure defaults. | | Globally off | Enhanced masking is disabled. Sensitive entities, including secrets, still follow default tokenization and explicit dictionary actions. | Reducing protection requires confirmation because approved model providers may receive more raw chat content. | The readiness panel shows whether the privacy data plane is usable: Privacy Gateway, required scanner, encryption keys, Redis, local secret route, and egress policy. If core readiness fails, privacy controls remain hidden or unavailable and the system fails closed instead of silently weakening protection. The Privacy Dictionary summary shows the current published version, active entries, and drafts when readiness is available. Use the linked [Privacy Dictionary](/aivis/features/privacy-dictionary) page to manage entity rules for tokenization, generalization, local-only routing, blocking, and review. ## Default Assistant Tools | Tool Group | What It Controls | Before Enabling | | ----------------------------- | ------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------- | | Built-in search and retrieval | Search, connector-backed retrieval, and document-aware answering for the default assistant. | Confirm the deployment has knowledge indexing enabled and that users can access the intended document sets. | | Web and URL tools | Web search and URL-reading capabilities when available. | Confirm public-network and content-source policy with administrators. | | Media and analysis tools | Image generation, video generation, code execution, and similar built-in capabilities. | Confirm model credentials, service readiness, cost controls, and policy boundaries. | | MCP tools | Tools exposed by authenticated MCP servers. | Authenticate the server first; unauthenticated tools cannot be enabled. | | OpenAPI tools | Tools created from approved API specifications. | Confirm schema, credentials, authorization, and high-impact action policy before enabling broadly. | Tool toggles affect the default assistant. They do not automatically enable the same tool on every custom Agent, and they do not bypass tool-level authentication or policy checks. ## Default System Prompt The default assistant system prompt shapes the built-in assistant across new conversations. Use it for general behavior, tone, response rules, and organization-wide guidance that should apply when users do not select a custom Agent. Before changing it: * Test the prompt with normal chat, search, attached files, and enabled tools. * Avoid placing secrets, credentials, or hidden policy assumptions in the prompt. * Keep operational rules concrete so they can be verified by admins. * Remember that custom Agents may add their own instructions or replace parts of the base prompt depending on their configuration. ## Query History and Retention | Setting | Behavior | Governance Note | | ----------------------------------- | ------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------- | | Query history type | Controls whether query history is disabled, anonymized, or recorded normally. | Choose the least detailed mode that still supports debugging, audit, and product operations. | | Hide Query History from admin panel | Removes the Query History page from the admin sidebar. | Visibility-only; it does not by itself change whether query events are recorded. | | Maximum chat retention | Deletes chats with no activity beyond the selected retention window, or keeps them forever. | Shortening retention requires confirmation because older chats can become permanently deleted. | | Custom retention | Allows a whole number of days when presets are insufficient. | Keep retention aligned with legal, security, and support obligations. | ## Access, Search, and File Controls | Setting | Behavior | | ---------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Anonymous access | Allows users to start chats without signing in when the deployment supports anonymous sessions. Anonymous users cannot manage Agents or administrator settings. | | Auto-scroll | Sets the workspace default for whether chats scroll to new responses automatically. | | Search mode | Enables the dedicated search experience for eligible tiers and deployments. | | File size limit | Caps the size of user-uploaded files accepted in chat. The deployment may also enforce a hard maximum. | | File token limit | Controls the token threshold for file processing. A value of zero means no configured limit in the UI. | ## Change Management * Change one behavior at a time when validating a production deployment. * Treat privacy-policy reductions and retention reductions as high-impact changes. * Record the reason for broad tool enablement, especially web, code, external API, or MCP tools. * Test with a regular user, an administrator, and a user who lacks access to the relevant Agent or tool. * Review query history, tracing, or administrator logs only according to the deployment's privacy policy. ## Verification After updating Chat Preferences: * Start a new chat and confirm company context and default assistant behavior match the saved settings. * Edit the default assistant prompt and confirm it affects the default assistant but not unrelated custom Agents. * Toggle a built-in tool and confirm the default assistant can or cannot use it as expected. * Enable an MCP or OpenAPI tool only after authentication and confirm unauthenticated tools remain disabled. * Change the privacy policy and confirm risky reductions require confirmation. * Open Privacy Dictionary from the summary card and confirm published version, active entries, and drafts are visible when readiness passes. * Change query-history mode and confirm the selected visibility and recording policy with an approved test request. * Shorten chat retention and confirm the reduction warning appears before saving. * Upload files around the configured size and token thresholds. ## Troubleshooting | Symptom | Likely Cause | Resolution | | ----------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------ | | Settings do not save | The settings request failed, the admin session expired, or another admin changed the same field. | Refresh the page, reapply one setting, and check the toast or server logs for the failed request. | | Privacy controls are missing | Privacy readiness failed or the deployment has not enabled the privacy data plane. | Check the readiness card and verify gateway, scanner, encryption keys, Redis, route, and egress policy. | | Privacy dictionary summary is unavailable | Readiness has not passed or the summary request failed. | Open [Privacy Dictionary](/aivis/features/privacy-dictionary) and verify the current configuration directly. | | A tool toggle is disabled | The tool's MCP server is unauthenticated, the deployment lacks the capability, or the tool is not available for Agent creation. | Authenticate the server, configure the provider, or verify capability availability. | | Query History is not visible | It may be hidden from the admin sidebar or unavailable for the deployment. | Check the hide-query-history setting and the selected query-history mode. | | Retention change asks for confirmation | The new value is shorter than the current retention window. | Confirm only after verifying the deletion impact. | | File upload is rejected | The file exceeds the workspace limit or the deployment hard maximum. | Lower file size, adjust the workspace limit, or ask an administrator to review deployment limits. | ## Related Pages * [Privacy Dictionary](/aivis/features/privacy-dictionary) * [Language Models](/aivis/models/language-models) * [Model Credentials](/aivis/models/model-credentials) * [Agents](/aivis/agents/agents) * [Data Flows](/aivis/governance/data-flows) * [Security](/aivis/governance/security)