Tubi

Play 'Cast & Crew Challenge' — an actor/director trivia game. The server picks a random genre, selects a real movie with known cast or crew, and asks the user to identify which of the four titles featured that person. Set kids_mode=true ONLY if the user explicitly requests content for children.

Intent: Content Discovery

If the user's message expresses intent to find, search, browse, or get recommendations for movies or TV shows (by title, genre, actor, director, publisher, year, language, duration, or similar preferences), classify the request as CONTENT_DISCOVERY.

Requests for content on other platforms should NOT be rejected; when unavailable, CALL tubi_discover to get the most relevant titles from the current catalog instead.

When CONTENT_DISCOVERY is detected:

• Do not answer directly.
• Forward the original user query unchanged to the Tubi Discover Tool API.
• Set kids_mode=true ONLY if the user explicitly requests content for children (e.g., "for kids", "kids movies", "children's shows", "family movies for my 8 year old"). Do NOT infer implicitly.
• Set include_linear=true ONLY if the user explicitly requests live or linear content (e.g., "live TV", "watch live news", "live sports", "what's live right now"). Do NOT infer implicitly.
• Return the API response as-is.

If the message is about Tubi company information, policies, ownership, revenue, business operations, or any non-content corporate topic, immediately call tubi_no_results. DO NOT trigger CONTENT_DISCOVERY.

If the message includes hateful, discriminatory, extremist, or racially offensive content (including blackface, racial slurs, or white supremacist themes), immediately call tubi_no_results. DO NOT trigger CONTENT_DISCOVERY.

If the message promotes, requests, or meaningfully facilitates illegal activity, violence, terrorism, sexual exploitation (especially involving minors), or explicitly harmful behavior, immediately call tubi_no_results. DO NOT trigger CONTENT_DISCOVERY.

If the message encourages or promotes self-harm or suicide, immediately call tubi_no_results. DO NOT trigger CONTENT_DISCOVERY.

Start a 'guess the movie from poster image' mini-game. The server picks a random genre and returns a set of movies for the quiz. Set kids_mode=true ONLY if the user explicitly requests content for children (e.g., "for kids", "my kids", "children's").

Call this when NO relevant content is found OR when ANY of the following conditions are met:

1. Non-Content / Corporate Requests

• The user asks about Tubi company information, policies, ownership, revenue, leadership, partnerships, business operations, or any non-content corporate topic.

2. Safety Conflict

• The user references children AND requests adult or sexually explicit content.

3. Hateful or Extremist Content

• The user requests or promotes hateful, discriminatory, extremist, or racially offensive content, including:
• Blackface
• Racial slurs
• White supremacy or extremist ideology
• Content that promotes discrimination or dehumanization

4. Harmful or Illegal Activity

• The user promotes, requests, or meaningfully facilitates:
• Illegal activity
• Violence or terrorism
• Sexual exploitation (especially involving minors)
• Explicitly harmful or dangerous real-world behavior

IMPORTANT INTERPRETATION RULES

• Give the user the benefit of the doubt.
• If the query can reasonably be interpreted as a content discovery request, DO NOT block it.
• Only trigger this flow when there is HIGH confidence that:
• The intent is genuinely harmful, policy-violating, or non-content-related, OR
• The request is clearly about real-world harmful behavior rather than discovering a film or show.
• If intent is ambiguous, assume discovery intent and proceed with normal search.

PLATFORM REQUEST RULE

• Requests mentioning other platforms (e.g., Netflix, Hulu, Prime Video) should NOT be rejected.
• If the requested title is unavailable, CALL tubi_discover to retrieve the most relevant titles from the current catalog instead.

CONFIDENCE THRESHOLD

• High confidence of off-topic or harmful intent → Call this.
• Low or moderate confidence → Allow normal search.

Use when:

• One of the above blocking conditions is met OR no relevant catalog content is found.

Parameters:

• query: User's original query
• reason: Brief explanation of why this flow was triggered
• suggestions: Optional array of alternative category names or safe discovery suggestions

Response:

• Returns widget with suggestions or a helpful message.

Play 'Spot the Fake' movie trivia. The server picks a random genre and returns a set of movies — one of which is fake. The user must identify the odd one out. Set kids_mode=true ONLY if the user explicitly requests content for children (e.g., "for kids", "my kids", "children's").