Application Settings

Application Settings allow you to configure and customize your branchly application's behavior, search retrieval mode, AI model preferences, and user interaction tracking. These settings ensure a cohesive, optimized experience across all user-facing interfaces.

General

• Application Name: Assign a name to your Application for easy identification.

• Website Location: Enter the exact URLs where the User Interfaces will be embedded. Multiple URLs can be specified if the interfaces are to appear on various pages. Important: For security reasons, all domains where your user interfaces appear must be configured here. If a user interface is loaded on an unlisted domain, it will fail to initialize or return an error to prevent unauthorized embedding.

• Knowledge Base Language: Select the language and country code for your Knowledge Base content. For multi-language setups, add combinations of two-digit ISO language and country codes (e.g., en_US for English/United States, de_DE for German/Germany).

• Use Browser Language: When enabled, all branchly interfaces — chat widget, search widget, and any embedded UI — automatically adapt to the visitor's browser locale. This affects two things simultaneously:

  1. UI text localisation — all static interface texts (input placeholders, button labels, disclaimer messages, suggested questions, and error messages) are served in the visitor's language. Translations for configured locales are stored and managed server-side; for visitors using a language you haven't explicitly configured, branchly auto-generates translations on first visit and caches them for subsequent requests.
  2. Content retrieval and AI responses — branchly separates the content locale (which language of your indexed content to search) from the user language (which language to respond in). When a visitor's browser locale is not one of your configured valid locales, branchly switches to multilingual dense semantic search — bypassing the language-specific keyword (BM25) model — and instructs the AI to reply in the visitor's language while retrieving the nearest available content.

  How language is determined

  Every request carries two signals: the page locale (the language of the page the widget is embedded on) and the browser locale (the visitor's preferred language as reported by their browser). branchly resolves these together:

  Visitor's browser locale	Content retrieved from	AI replies in
  Matches one of your valid locales (e.g. en, de)	The page locale — visitor is on the right language page	Page locale
  Does not match any valid locale (e.g. ja on a de/en site)	Page locale as fallback, using multilingual semantic search	Browser locale — best-effort for unsupported visitors
  Not provided	Page locale	Page locale

  tip

  When a visitor's language is not in your valid locales, branchly uses dense (semantic) search only rather than hybrid search. This is because keyword-based matching (BM25) is language-specific and produces no useful results across different scripts (e.g. Japanese queries against German content). Multilingual semantic models still return meaningful results in these cases.

  When to use this feature

  Disabled (default) — The widget always responds in the language of the page it's on. Safe and predictable. Use this when your audience speaks the same language as your content.

  Enabled — Allows cross-lingual responses for visitors whose browser language isn't one of your configured valid locales. Use this when your site may receive visitors in languages you haven't indexed content for.

  Scenario	Recommendation
  Single-language site, homogeneous audience — e.g. a German SaaS for German-speaking customers	Disable — simplest setup, no surprises. The widget always replies in the page language.
  Single-language site, occasional international visitors — e.g. a German site visited by Japanese or French users	Enable — your German content stays German, but international visitors get a response in their own language.
  Multi-language site (e.g. separate /de/ and /en/ pages) with a consistent audience	Disable — each page already serves the right language. The widget picks up the page locale automatically; no extra configuration needed.
  Multi-language site + occasional visitors in unsupported languages — e.g. an EN/DE site with French visitors	Enable — EN and DE visitors get responses matching the page they're on; a French visitor on either page still gets a French reply, backed by the nearest available content.

  note

  Use Browser Language is a global setting that applies to all interfaces. The chat-specific Reply in User Language setting (in the Chat section below) handles a separate concern: detecting the language of a user's typed message within a session and instructing the AI to reply in kind. Both settings can be active at the same time.

• Status: Choose between Development, Production and Disabled modes. Switch to Production when your user interfaces are ready to go live.

• Dynamic Navigation: Configure dynamic navigation based on usage data within a specific timeframe.

  • Primary Metric: The primary method for edge weighting. We recommend starting with weight_manual and weight.

    tip

    Tip: Start with weight_manual and weight to establish a baseline before transitioning to dynamic edge weighting based on user interaction data.

  • Secondary Metric: Used when the primary metric does not yield a distinct edge weight or when node weights are identical.

  • Edge Weighting Methods:

    • weight_manual: Manually highlight specific nodes (e.g., keeping a contact node prominent).
    • weight: A customizable value set during data upload.
    • weight_rolling_30: Calculates average weights from the past 30 days of usage.
    • weight_rolling_60: Averages weights from the past 60 days of usage.
    • weight_rolling_alltime: Averages weights from all usage data since deployment.

• Data Retention Policy: Defines the number of days after which user request data is deleted. Note that insights and detailed analytics will not show historical data older than this limit, with the exception of the active sessions count.

Search

• Dense Embedding Model: A proprietary semantic AI model that supports major languages and excels at conceptual search tasks.
• Sparse Embedding Model: An information retrieval model optimized for advanced keyword lookup (currently uses a custom BM25 implementation).
• Search Mode: Defines the search algorithm used across all interfaces:
  • Dense: Default semantic search.
  • Hybrid (recommended): Combines sparse and dense models using reciprocal rank fusion (RRF) for highly accurate keyword and conceptual results.
  • Sparse: Pure keyword-based search.
• Search Result Highlighting: Customize which text is highlighted in results. Choose Chunk to highlight the best-matching text snippet, or Page Metadata Description to show the page's HTML meta description.
• Rerank Settings: Advanced settings to fine-tune the relevance and ordering of search results. You can configure different rules for search and chat interfaces:
  • Datetime reranking: Pushes newer content higher in search results. You can evaluate by published_date or modified_date.
  • Record source reranking: Boosts scores of results based on whether the query matches the title or text fields.
  • Custom boosting: Uses the score_boost field on nodes to manually adjust search ranking priorities.

Chat

• Chat Strategy: Choose between Agent (which supports active tools and workflows) and Standard RAG. This can only be enabled if at least one knowledge_base tool is active. See AI Actions for details.
• Reply in User Language: Enabled by default. Automatically detects the language of user queries and instructs the AI to reply in the detected language. If the query language is ambiguous, it defaults to the knowledge base locale.
• Follow Up Actions (Beta): Enables suggested questions and navigation recommendations based on conversational context. If a suggested link points to your configured domain, the chat interface will navigate within the same frame; otherwise, it opens in a new tab.
• Routing Model (Deactivated): The model used for orchestrating multi-tool routing.
• Generation Model (Deactivated): The model used to generate conversational responses, utilizing gpt-5.4 class models via Azure OpenAI.

tip

AI Model Providers: Aside from Azure OpenAI models, branchly supports a variety of model providers including Mistral, Cohere, Google, DeepSeek, and Qwen.

Tracking

Configure page and element interaction tracking to analyze user pathways and behavior across your website.

• Track Page Navigations / Link Clicks: Toggle to record user navigations and clicks. When active, navigation and click events are sent to the API, allowing you to trace user journeys in Docusaurus and other layouts. This setting can be activated individually for each application and user.
• Custom Element Tracking: Explicitly target specific elements (such as custom cards or banners) for tracking by adding the data-branchly attribute to your HTML elements.
• Tracking Rules: Define rules to track all anchor (<a>) and button elements globally, or target specific CSS selectors.
• Dashboard Exclusion: Navigation events occurring inside the branchly user interfaces are automatically ignored.

Integrations

Connect third-party platforms to extend the capabilities of your branchly application. This tab centralizes secure OAuth connections and setup configurations.

• WordPress: Connect your WordPress site via the REST API v2 to import posts and pages directly into your knowledge base, and configure webhooks to sync updates automatically.
• Calendly: Connect your Calendly account to authorize the AI Agent to book meetings and schedule calls on your behalf directly within the chat widget.

User Preferences

Manage personal notification preferences and report subscriptions for the current application.

• Weekly Digest Email: Opt in to receive periodic automated email reports detailing session metrics, trending classifications, and out-of-scope queries. This preference can be configured individually for each application and user. See Scheduled Email Reports for more details.

Advanced

Developer-focused settings for API integrations and security.

• API Keys: Generate and manage secure API keys to authenticate custom API requests to the branchly endpoints.