Now that the **Interactions API** is the default Gemini API surface, maintaining separate `gemini-interactions-api` and `gemini-api-dev` skills creates redundant maintenance and confusion.
Promotes Gemini Omni Flash to 1.1 and adds prompt guide, CLI, and SDK updates:
- Promotes model endpoint to gemini-omni-1.1-flash
- Adds video extension capabilities (multi-turn and uploaded video)
- Adds first and last frame transition support (<FIRST_FRAME>, <LAST_FRAME>)
- Adds multi-video reference support (<VIDEO_REF_N>) and source declarations
- Adds CLI options for resolution controls, video extension, and interpolation
- Adds official documentation link to omni.md.txt across skills
- Add gemini-omni-flash-api skill and helper scripts
- Add gemini-omni-flash-preview to interactions API skill
- Simplify output of video checks and handle video-only inputs with no audio streams
- Add back-off and retry logic on media uploads
- Check for ffmpeg and system dependencies and update HTTP header key usage
- Correct description of aspect ratio scaling
The README links the phrase "thought circulation" to
`https://ai.google.dev/gemini-api/docs/thinking#signatures`. The page
section at that anchor is titled "Thought signatures", and the
underlying SDK field is exposed as `signature` (referenced in
`gemini-interactions-api/SKILL.md` line 246). The displayed link text
should match the destination's actual section name so a reader who
follows the link finds what the surrounding prose promises.
Madi