Source: https://docs.sogni.ai/sogni-intelligence/introduction/

![Introducing Sogni Intelligence — The AI layer making every Sogni app smarter.](https://docs.sogni.ai/assets/sogni-intelligence-hero.jpg)

# 👋Sogni Intelligence

## [#](https://docs.sogni.ai/sogni-intelligence/introduction/#what-it-is)What it is

Sogni Intelligence is the OpenAI-compatible LLM and agent API on the Sogni Supernet. Drop-in chat, vision, streaming, custom function tools, and **built-in creative media tools** for image, video, and music — all behind one bearer token and one base URL.

```
POST https://api.sogni.ai/v1/chat/completions
```

Use [`/v1/chat/completions`](https://docs.sogni.ai/sogni-intelligence/chat-completions/) when an LLM should interpret a user request, decide whether tools are needed, and return a final assistant response. Use [`/v1/creative-agent/tools/execute`](https://docs.sogni.ai/sogni-intelligence/chat-completions/#direct-synchronous-tool-execution) when your application already knows the exact synchronous composition/planning tool and arguments to run. Use [`/v1/chat/runs`](https://docs.sogni.ai/sogni-intelligence/durable-chat-runs/) for model-driven behavior with durable progress, replayable SSE events, recovery, and cancellation. Use [`/v1/creative-agent/workflows`](https://docs.sogni.ai/sogni-intelligence/creative-agent-workflows/) when your application already knows the exact media workflow to run.

## [#](https://docs.sogni.ai/sogni-intelligence/introduction/#whats-new)What's new

Introduced in March 2026, Sogni Intelligence is moving fast. The default `qwen3.6-35b-a3b-gguf-iq4xs` model brings stronger reasoning, sharper tool calling, and a roughly 264k-token context window — enough working memory for long creative briefs, multi-step plans, and rich media analysis.

In practice, every Sogni SuperApp gets smarter, easier to use, more personalized, and more interoperable. In [Sogni Photobooth](https://docs.sogni.ai/sogni-photobooth/web/), you can describe the booth you want instead of browsing dozens of preset categories — try ideas like "put me on famous magazine covers" or "turn me into famous paintings." In [Sogni 360](https://docs.sogni.ai/sogni-360/), the assistant can analyze your trip photos and choose smoother object-to-object transitions across a sequence.

Developers can plug in through OpenAI Chat Completions-compatible clients including [Open WebUI](https://docs.sogni.ai/sogni-intelligence/integration-open-webui/), [OpenClaw](https://docs.sogni.ai/sogni-intelligence/integration-openclaw/), [Hermes Agent](https://docs.sogni.ai/sogni-intelligence/integration-hermes-agent/), the [OpenAI SDK](https://docs.sogni.ai/sogni-intelligence/openai-sdk-compatibility/), and agentic coding clients that support custom OpenAI-compatible endpoints — all powered by Sogni's decentralized compute network.

## [#](https://docs.sogni.ai/sogni-intelligence/introduction/#current-model-ids)Current model IDs

`GET /v1/models` returns the live catalog. Today:

| Model ID | Use |
| --- | --- |
| `qwen3.6-35b-a3b-gguf-iq4xs` | Default chat, reasoning, coding, vision, and tool calling. |
| `qwen3.5-35b-a3b-abliterated-gguf-q4km` | Alternative Qwen 3.5 variant. |

## [#](https://docs.sogni.ai/sogni-intelligence/introduction/#what-it-supports)What it supports

-   OpenAI-compatible chat completions and SSE streaming
-   Multi-turn conversations with `developer`, `system`, `user`, `assistant`, and `tool` messages
-   Inline base64 PNG/JPEG vision input (up to 20 images per request)
-   Custom OpenAI-style function tools alongside Sogni's built-in tools
-   Sogni creative tools with media indexing across rounds — generate, edit, analyze, and deterministically upscale images through 16K with NVIDIA RTX VSR
-   Synchronous composition tools — `enhance_prompt`, `compose_script`, `compose_lyrics`, `compose_instrumental`
-   Synchronous workflow planners — `compose_workflow` for one-shot durable plans and `compose_workflow_template` for savable/editable recipes
-   Direct synchronous tool execution when your app already has exact tool arguments
-   Durable hosted chat runs with persisted state, replayable events, cost-approval pause, recovery, and cancellation
-   Durable creative workflows with cancel, resume, reseed, and full SSE replay
-   Saveable workflow templates invocable by ID with typed inputs

## [#](https://docs.sogni.ai/sogni-intelligence/introduction/#where-to-go-next)Where to go next

[Reference **API Reference →**  
Endpoint-by-endpoint lookup with curl, JS, and Python examples.](https://docs.sogni.ai/api-reference/) [Surface · Chat **Chat Completions →**  
OpenAI-compatible chat, streaming, vision input, and built-in tools.](https://docs.sogni.ai/sogni-intelligence/chat-completions/) [Compatibility **OpenAI SDK Compatibility →**  
Point the official OpenAI SDKs at `api.sogni.ai/v1` and go.](https://docs.sogni.ai/sogni-intelligence/openai-sdk-compatibility/) [Surface · Runs **Durable Chat Runs →**  
Long tool-calling turns with cost-approval pauses, SSE replay, recovery.](https://docs.sogni.ai/sogni-intelligence/durable-chat-runs/) [Surface · Workflows **Creative-Agent Workflows →**  
Pre-planned multi-step jobs — storyboards, image→video, batch generation.](https://docs.sogni.ai/sogni-intelligence/creative-agent-workflows/) [Surface · Templates **Workflow Templates →**  
Saveable, parameterized recipes you can invoke by ID with typed inputs.](https://docs.sogni.ai/sogni-intelligence/workflow-templates/) [Assets **Media Upload URLs →**  
Presigned POST URLs for reference media that durable workflows can use.](https://docs.sogni.ai/sogni-intelligence/media-upload-urls/) [Operations · Billing **Billing & Cost Control →**  
Sogni vs Premium Spark, vendor-model gating, cost-approval flow.](https://docs.sogni.ai/sogni-intelligence/billing-and-cost-control/) [Agent SDK **Sogni Creative Agent Skill →**  
Public agent skill + `sogni-agent` CLI for Claude Code, Manus, others.](https://docs.sogni.ai/sogni-intelligence/creative-agent-skill/)

## [#](https://docs.sogni.ai/sogni-intelligence/introduction/#tool-families)Tool families

By default Sogni Intelligence injects the `creative-tools` family — image, video, music generation; image / video analysis; metadata extraction; plus synchronous composition tools (`enhance_prompt`, `compose_script`, `compose_lyrics`, `compose_instrumental`).

Set `sogni_tools: "creative-agent"` to add workflow control, asset-manifest tools, and both workflow planners: `compose_workflow` for one-shot durable plans and `compose_workflow_template` for reusable templates. Set `sogni_tools: false` or `"none"` to disable tool injection entirely.

If your app already knows it needs one of the synchronous tools above, call it directly with `POST /v1/creative-agent/tools/execute` instead of routing through a chat-completions tool-choice round.

## [#](https://docs.sogni.ai/sogni-intelligence/introduction/#quick-test)Quick test

```
curl https://api.sogni.ai/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3.6-35b-a3b-gguf-iq4xs",
    "messages": [
      {"role": "user", "content": "Make me something beautiful, use your imagination!"}
    ],
    "sogni_tools": true
  }'
```

For the full REST reference — every endpoint, parameter, and response — see the [API Reference](https://docs.sogni.ai/api-reference/).