juspay/neurolink

Summary Create comprehensive multimodal.md documentation for video support covering SDK usage, CLI usage, provider differences, configuration options, and best practices. Technical Details File(s): docs/features/multimodal.md (update existing) Section: Video Support (new section) Effort: 2h Acceptance Criteria Video Support section added to multimodal.md SDK usage examples (basic, custom frames, native video) CLI usage examples (all flags) Provider comparison table (Gemini native vs others) Configuration options documented Frame extraction explained with diagrams/examples Audio transcription documented Best practices section Troubleshooting section Performance considerations Token cost estimation guide No typos or formatting errors Implementation Notes Document structure: Overview: Video support capabilities Supported Formats: MP4, WebM, MOV, AVI, MKV SDK Usage: Basic video analysis Custom frame extraction Native video (Gemini) Audio transcription CLI Usage: All vid

Summary Create a new CLI command 'neurolink voices' that lists all available TTS voices for a provider. Technical Details File(s): src/cli/commands/voices.ts (create new), src/cli/index.ts Effort: 2h Acceptance Criteria Created src/cli/commands/voices.ts Implemented voices command with yargs Accept --provider flag (required) Accept --language flag (optional filter) Call TTSProcessor.getVoices(provider, { languageCode }) Display voices in formatted table Sort voices alphabetically by name Show voice count at end Handle errors gracefully Register command in src/cli/index.ts Examples in help text Dependencies Depends on: TTS-005, TTS-010, TTS-014 Blocks: none Priority: low | Effort: 2h | Complexity: simple

Summary Update the system prompt generation in MessageBuilder to detect audio files and inform the LLM that audio content has been automatically transcribed and included in the message. Technical Details File(s): src/lib/utils/messageBuilder.ts Effort: 0.5 hours Acceptance Criteria Detect if audioFiles are present in input or files array Add "audio files (transcribed)" to file types list Update system prompt to mention transcription Clear guidance that audio is already processed Dependencies Depends on: AUDIO-021 Blocks: none Priority: medium Effort: 0.5h Complexity: simple

Summary Create a SentenceBuffer utility class for intelligent text chunking during streaming synthesis. Technical Details File(s): src/lib/utils/sentenceBuffer.ts (create new) Effort: 2h Acceptance Criteria Created src/lib/utils/sentenceBuffer.ts Implemented SentenceBuffer class Method: add(text: string) - accumulate text chunks Method: hasCompleteSentence() - detect sentence boundaries Method: extractSentence() - extract and remove complete sentence Method: hasRemainingText() - check for buffered text Method: flush() - return all remaining text Sentence detection regex Handle edge cases: abbreviations Unit tests pass Dependencies Depends on: TTS-001 Blocks: TTS-009, TTS-013, TTS-017 Priority: medium | Effort: 2h | Complexity: simple

Summary Add provider configuration getter methods to VideoProcessor: getProviderConfig(), supportsNativeVideo(), and estimateTokens(). These methods provide provider-specific video capabilities and cost estimation. Technical Details File(s): src/lib/utils/videoProcessor.ts Methods: getProviderConfig(), supportsNativeVideo(), estimateTokens() Effort: 1h Acceptance Criteria getProviderConfig() returns config for given provider Returns null for unknown providers supportsNativeVideo() returns boolean based on config estimateTokens() calculates token count based on frames and transcription All methods have JSDoc comments Tests pass for all providers No regressions Implementation Notes static getProviderConfig(provider: string): VideoProviderConfig | null { return VIDEO_PROVIDER_CONFIGS[provider] || null; } static supportsNativeVideo(provider: string): boolean { return this.getProviderConfig(provider)?.supportsNativeVideo || false; } static estimateTokens(duration: number, f