ai_gateway_audio_speech_create
ChatGPTGenerate speech from text. First select a voice from the model's supported_voices in ai_gateway_models_get. Returns stored audio for playback or download, never a server file path.
ai_gateway_audio_transcriptions_create
ChatGPTTranscribe an explicitly user-selected audio file with automatic language detection. This produces text without creating a Robin note. Use Robin's import tool when the user wants a note.
ai_gateway_chat_completions_create
ChatGPTAsk a selected Vecbase language model to generate text. This is a paid model call and does not replace the host assistant. Reuse idempotency_key after interruption.
ai_gateway_images_create
ChatGPTGenerate images with a discovered model. Returns stored results and an Operation ID. Reuse idempotency_key to recover; do not regenerate to check progress.
ai_gateway_models_get
ChatGPTRead one discovered model's supported inputs, parameters and voices. Does not return a price quote.
ai_gateway_models_list
ChatGPTFind available models before generation. For reading text aloud, use output_modalities=["speech"]; for speech to text use ["transcription"]. Omit q unless filtering by model name or ID. Returns capabilities, not price quotes; do not invent models or prices.
ai_gateway_open
ChatGPTOpen one compact result panel for an existing authorized model Operation when a preview or player helps. Use once, then use ai_gateway_results_get for progress. Never generate from this tool.
ai_gateway_results_get
ChatGPTRecover a model call's status and stored result using its Operation ID. Does not open a panel or generate anything.
ai_gateway_videos_content
ChatGPTGet stored output references for a completed video using its Operation ID. Pending tasks return their current status.
ai_gateway_videos_create
ChatGPTGenerate a video with a discovered model. Returns one Operation UUID as both operation_id and the video response id. Query progress; never start a second generation to poll.
ai_gateway_videos_get
ChatGPTRead an existing video using its Operation ID without generating again.
robin_files_upload
ChatGPTUpload one bounded chunk of an explicitly selected audio file or recording. Reserves encrypted, caller-bound bytes for Robin transcription.
robin_notes_create
ChatGPTCreate a meeting note. Returns a canonical Operation; its response contains note_id.
robin_notes_get
ChatGPTRead a meeting note, its summary and current revision.
robin_notes_list
ChatGPTList meeting note previews, newest first.
robin_notes_update
ChatGPTUpdate note fields at an exact revision. Returns a canonical Operation.
robin_open
ChatGPTOpen the notes panel for interactive review or recording. Provide note_id for one note; omit it to browse. The compact panel has Start recording and Import audio; the expanded panel has these actions under New note (+). An empty note also has these actions. Recording requires the user's click and microphone permission; finishing submits for automatic transcription. Each note holds one recording. Opening this panel does not start recording or create a note. If recording is unavailable, offer a selected audio file. Open once; use data tools for follow-up reads, edits and progress without another panel. Give only a brief explanation using these actual controls.
robin_operations_get
ChatGPTRead an authorized Robin operation and its result. A waiting client does not cancel the work.
robin_segments_update
ChatGPTCorrect transcript text at an exact revision. Returns a canonical Operation.
robin_speakers_list
ChatGPTRead a meeting's speakers and their current revisions.
robin_speakers_update
ChatGPTRename a speaker at an exact revision. Returns a canonical Operation.
robin_summary_create
ChatGPTRegenerate a stale or failed summary, replacing its existing text, including manual edits, when generation succeeds. Does not change the recording, transcript or handwritten notes. Use only when the user requests regeneration. Requires the current note revision and returns a canonical Operation.
robin_transcript_get
ChatGPTRead timestamped transcript segments and speaker references.
robin_transcriptions_create
ChatGPTImport an audio or video file into an unrecorded note. Format validation and bounded audio decoding run after canonical Operation acceptance.
robin_transcriptions_import
ChatGPTTranscribe a completed caller-bound Robin file reservation. Returns the same canonical Operation as an attached file import.