Inference
Send chat completion requests to managed inference endpoints. Supports text prompts and optional video URL inputs.
Send chat completion requests to managed inference endpoints. Supports text prompts and optional video URL inputs.
Command
cosmicac inference <subcommand> [options]Subcommands
| Subcommand | Description |
|---|---|
| chat | Chat with AI models |
| transcribe | Transcribe audio using AI models |
inference chat
Send chat completion requests to a managed inference endpoint. Supports text prompts and optional video URL inputs. Run without --message to start an interactive chat session. Reasoning models emit their thinking (shown dimmed above the answer) by default; use --disable-think to turn it off.
Usage
cosmicac inference chat [options]Options
| Option | Description |
|---|---|
--api-key | Use API key auth (reads COSMIC_API_KEY or prompts; never pass the key as an argument) |
--endpoint-id | Inference endpoint ID |
--message | Message to send to the model (omit for interactive mode) |
--video-url | URL of video to analyze |
--stream | Enable streaming response for real-time output |
--disable-think | Disable model reasoning/thinking output (enabled by default) |
--reasoning-effort | Reasoning effort level (e.g. low, medium, high) |
inference transcribe
Send audio transcription requests to a managed inference endpoint. Supports audio URL inputs for transcription.
Usage
cosmicac inference transcribe [options]Options
| Option | Description |
|---|---|
--api-key | Use API key auth (reads COSMIC_API_KEY or prompts; never pass the key as an argument) |
--endpoint-id | Inference endpoint ID |
--audio-url | URL of audio file to transcribe |
--prompt | Optional prompt to guide transcription |