Operators

Every AI job, routed to the right model.

Friday separates concerns. Each operator owns one kind of work - voice, images, video, music, or scheduling - and resolves its own provider and model from your saved settings.

AI Assistant

The main Friday agent. Runs a provider-neutral tool loop with system prompts, context compaction, pre-run safety hooks, and memory across sessions.

Any configured chat provider and model
Pre-run policy hooks block unsafe requests
Automatic context compaction on overflow
Per-agent startup files and persistent memory
Per-run provider and model overrides

Supported providers

OpenAI · Anthropic · Google · xAI · Mistral · DeepSeek · Qwen · Kimi · Perplexity · NVIDIA · and more

Speech to Text

Live dictation and audio transcription. Streams 24 kHz PCM audio from the microphone and returns partial and final transcript events in real time.

Real-time streaming transcription
Echo cancellation and noise suppression
Configurable language and sample rate
Provider-agnostic session contract
Graceful close on renderer teardown

Supported providers

OpenAI Realtime Whisper · Deepgram · Any STT-capable provider

Text to Speech

Synthesize spoken audio from assistant responses. Configurable voice, format, and streaming style through any TTS-capable provider.

Any configured TTS provider and model
Streaming or batch audio output
Voice, format, and speed options
Background task with progress tracking
Cron-triggered audio delivery

Supported providers

ElevenLabs · Deepgram · Cartesia · OpenAI · Any TTS-capable provider

Image Creator

Generate and edit images from prompts or reference assets. Supports any image-capable provider with normalized result records.

Text-to-image and image editing
Aspect ratio and count controls
Background image.create task
Cron-scheduled image generation
Provider-specific adapters behind a shared API

Supported providers

Black Forest Labs · Stability AI · Ideogram · OpenAI · Google · Any image-capable provider

Video Creator

Create videos via async generation jobs. Supports prompt-to-video, reference assets, duration, and aspect ratio through any video-capable provider.

Async video generation with polling
Prompt, duration, and aspect ratio inputs
Background video.create task
Cron-triggered video production
Webhook and download support in adapters

Supported providers

Runway · Kling · Pika · Luma AI · OpenAI · Any video-capable provider

Music Creator

Generate music, sound effects, and audio from text prompts. Covers any provider with music or audio generation capability.

Text-to-music and text-to-audio
Duration, format, and stem controls
Background sound.create task
Cron-triggered audio generation
Provider-agnostic normalized audio results

Supported providers

MiniMax · Stability AI · ElevenLabs · Google · Suno · Any sound-capable provider

Document OCR

Extract text from images and documents via a configured OCR endpoint. Validates image input, posts to the endpoint, and returns extracted text.

Base64 image input with MIME type
Configurable OCR endpoint
Background ocr.run task
Language hint support
Migration path to provider/model picker

Supported providers

Any OCR endpoint · Provider/model picker planned

Cron Scheduler

Schedule recurring tasks, one-time runs, and interval jobs with timezone support, missed-run handling, concurrency policies, and full audit trails.

Cron expression, interval, and one-time schedules
Missed-run policies: skip, run-once, catch-up, fail
Concurrency policies: skip, queue, cancel-previous
Timezone-aware with 30-second poll interval
Agent turn, image, video, sound, and reminder tasks

Supported providers

Built-in — no external provider required

Background Tasks

Run agent, OCR, image, video, and sound work concurrently. Each task has a lifecycle, progress tracking, cancellation via AbortSignal, and in-memory result storage.

Concurrent task execution
Cooperative cancellation via AbortSignal
Lifecycle events: queued, running, succeeded, failed, cancelled
No default execution timeout
Renderer-safe IPC preload API

Supported providers

Built-in — delegates to operator modules

Bring your own provider

Every operator is provider-agnostic.

Configure your preferred provider and model in Settings. Credentials stay local; the operator resolves them at runtime.

Download desktop app View integrations