CLI: fetch AI summaries from share links

[ffu3xu4tw]

Could the Cap CLI retrieve the AI-generated summary from a shared recording URL and return it as JSON? Please also support custom/self-hosted Cap server URLs, such as those configured through CAP_SERVER_URL.

[linear]

CAP-758

[tembo]

The CLI already supports custom server URLs via CAP_SERVER_URL (or from Cap Desktop's stored serverUrl setting). This is documented in apps/cli/src/main.rs lines 120-121 and implemented in apps/cli/src/credentials.rs. So the self-hosted aspect is already handled for authenticated endpoints.

For fetching AI summaries, the relevant data lives in the video's metadata JSON field in the database (see packages/database/types/metadata.ts):

interface VideoMetadata {
  aiTitle?: string;
  summary?: string;
  chapters?: { title: string; start: number }[];
  aiGenerationStatus?: "QUEUED" | "PROCESSING" | "COMPLETE" | "ERROR" | "SKIPPED";
}

The share page at apps/web/app/s/[videoId]/page.tsx already fetches and displays this data. A new CLI command (e.g., cap video info <url-or-id>) could:

1. Parse the share URL to extract the video ID (the <id> from https://server/s/<id>)
2. Hit an API endpoint to retrieve the metadata

Currently there's no public unauthenticated endpoint that returns just the AI metadata as JSON. The closest is /api/video/ai which requires authentication. A new endpoint like /api/video/public-info?videoId=<id> could return the summary, chapters, and title for public videos.

For implementation, the CLI side would go in apps/cli/src/ as a new subcommand, similar to how upload.rs handles video operations. The endpoint would need to check videos.public and respect any password protection.

Related: issue #1996 tracks adding custom base URLs for the AI providers themselves (self-hosted Whisper/vLLM), which affects how summaries are generated on self-hosted instances.