# Venice API

> Build AI apps with Venice API - privacy-first, uncensored, OpenAI-compatible. Covers chat, images, audio, video, embeddings, web search.

- Skill: `wulfmeister/venice-api` (Agent Skill, multi-file: 5 files)
- Install (CLI): `npx skillmds@latest add wulfmeister/venice-api`
- Raw SKILL.md: https://api.skillmd.com/api/skills/wulfmeister/venice-api/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Integrations & APIs
- License: MIT
- Author: wulfmeister (https://skillmd.com/u/wulfmeister)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/wulfmeister/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`:

| Parameter | Type | Description |
|-----------|------|-------------|
| `enable_web_search` | string | "off", "on", "auto" |
| `enable_web_scraping` | boolean | Scrape URLs in message (max 3) |
| `enable_web_citations` | boolean | [REF]0[/REF] format citations |
| `strip_thinking_response` | boolean | Show/hide think blocks |
| `disable_thinking` | boolean | Disable reasoning |
| `character_slug` | string | Use AI character |
| `include_venice_system_prompt` | boolean | Disable Venice defaults |

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

| Tier | Requests/min | Tokens/min |
|------|-------------|------------|
| XS   | 500         | 1M         |
| S    | 75          | 750K       |
| M    | 50          | 750K       |
| L    | 20          | 500K       |

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

| Use Case | Model |
|----------|-------|
| Fast chat | `qwen3-4b` |
| General | `zai-org-glm-4.7` |
| Reasoning | `kimi-k2-thinking` |
| Code | `grok-code-fast-1` |
| Vision | `gemini-3-pro-preview` |
| Uncensored | `venice-uncensored` |
| Budget reasoning | `deepseek-v3.2` |
| Frontier | `openai-gpt-52` |

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`](https://pypi.org/project/venice-ai/) | [`pyvenice`](https://pypi.org/project/pyvenice/)
- **TypeScript**: [`venice-ai-sdk-provider`](https://www.npmjs.com/package/venice-ai-sdk-provider) (Vercel AI SDK)
- **Ruby**: [`venice_client`](https://rubygems.org/gems/venice_client)
- **CLI**: [`veniceai/venice-cli`](https://github.com/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