v1.7.0

Minor Changes

• e6fb51b: Align SDK with Bold AI API v1 specification

  Breaking Changes:

  • Video-scoped chat: pass videoId in options instead of as separate arg: bold.ai.chat({ videoId, prompt })
  • Source.timestamp_end and Source.playback_id are now required (were optional)
  • AIUsage now uses input_tokens/output_tokens (was prompt_tokens/completion_tokens/total_tokens)
  • Removed TopicInput type - RecommendationsOptions.topics now only accepts string[]
  • RecommendationsOptions.context is now AIContextMessage[] (was string)

  New:

  • bold.ai.chat(opts) - Single method for all chat (pass videoId to scope to a video)
  • bold.ai.recommendations(opts) - AI-powered video recommendations (replaces recommend)
  • ChatOptions.videoId - Scope chat to a specific video
  • ChatOptions.currentTime - Pass current playback position for context

  Deprecated (still work, will be removed in v2):

  • bold.ai.ask() → use bold.ai.chat()
  • bold.ai.coach() → use bold.ai.chat()
  • bold.ai.recommend() → use bold.ai.recommendations()
  • AskOptions type → use ChatOptions
  • RecommendOptions type → use RecommendationsOptions
  • RecommendResponse type → use RecommendationsResponse

  Type Changes:

  • Added Source.id field (chunk identifier)
  • Added conversation_id and video_id to message_start event
  • Added conversation_id, recommendations, guidance to message_complete event
  • Simplified clarification event to include content field
  • Removed legacy event types: token, answer, complete