Venice API

Venice API

Privacy-first AI platform with OpenAI SDK compatibility. Zero data retention.

Base Config

Base URL: https://api.venice.ai/api/v1
Auth: Bearer token (API key from venice.ai/settings/api)

Venice-Specific Parameters

Use extra_body with venice_parameters:

Also in extra_body (but outside venice_parameters):

• prompt_cache_key (string) - Routing hint for cache hits in multi-turn conversations

Key Endpoints

POST /chat/completions     # Text + vision + audio + video input
POST /image/generate       # Image generation
POST /image/upscale       # Upscale images (<25MB)
POST /image/edit          # Inpaint/edit
POST /audio/speech        # TTS (60+ voices)
POST /audio/transcriptions # Transcription (Beta)
POST /video/quote         # Get price estimate
POST /video/queue         # Queue video generation
POST /video/retrieve      # Poll for completion
POST /video/complete      # Delete from storage
POST /embeddings          # Vector embeddings
POST /image/remove-background # Remove background
GET  /image/styles        # List style presets
GET  /characters/{slug}   # Get character info
GET  /models              # List models with capabilities
GET  /models/traits       # Auto-select models (default, fastest, etc.)
GET  /models/compatibility_mapping # OpenAI-to-Venice model map
GET  /api_keys/rate_limits     # View your rate limits
GET  /api_keys/rate_limit_logs # Rate limit history
GET  /billing/balance     # Check USD/DIEM balance

Rate Limits

Image: 20/min | Audio: 60/min | Video queue: 40/min

Video Workflow (Async)

1. Quote: POST /video/quote → get price
2. Queue: POST /video/queue → get queue_id
3. Poll: POST /video/retrieve with queue_id
   • Returns JSON while PROCESSING
   • Returns video/mp4 when done
4. Complete: POST /video/complete → delete storage

Pricing

• Text: Per 1M tokens (check model pricing)
• Web Search: $10/1K calls
• Web Scraping: $10/1K calls (up to 3 URLs)
• TTS: $3.50/1M characters
• Transcription: $0.0001/audio second

See: https://docs.venice.ai/overview/pricing

Model Selection

Models change frequently. Verify current models with GET /models or use GET /models/traits for auto-selection (returns default, fastest, default_code, etc. without hardcoding IDs).

Privacy Tiers

• Private: Zero data retention
• Anonymized: Not affiliated with user
• Uncensored: No content filtering

Gotchas

1. API key shown once only - copy immediately
2. Image/audio URLs must be publicly accessible
3. Video is async - must poll, not instant
4. Venice appends system prompts (disable with include_venice_system_prompt: false)
5. Rate limits are per-key, not per-account

Error Codes

• 401: Invalid API key
• 402: Insufficient balance (check /billing/balance)
• 429: Rate limit (check x-ratelimit-* headers)
• 422: Content policy violation (try uncensored model)

Response Headers

Key headers to check:

• CF-RAY - Request ID for support
• x-ratelimit-remaining-requests
• x-venice-balance-usd
• x-venice-model-deprecation-warning

Community SDKs

Beyond the OpenAI SDK approach, dedicated packages exist:

• Python: venice-ai | pyvenice
• TypeScript: venice-ai-sdk-provider (Vercel AI SDK)
• Ruby: venice_client
• CLI: veniceai/venice-cli

Resources

• Docs: https://docs.venice.ai
• Models: https://docs.venice.ai/overview/models
• Pricing: https://docs.venice.ai/overview/pricing
• Keys: https://venice.ai/settings/api
• Swagger: https://api.venice.ai/doc/api/swagger.yaml
• Changelog: https://featurebase.venice.ai/changelog