🪽Hermes Agent Integration

Hermes Agent can use Sogni Intelligence through an OpenAI-compatible custom endpoint. Use this when you want Hermes Agent to route prompts to Sogni's Qwen models while keeping Hermes' local agent workflow.

#Prerequisites

• Hermes Agent installed.
• A Sogni API key from dashboard.sogni.ai/api-key.
• A Sogni account with Spark or SOGNI token balance.

#Interactive Setup

Run Hermes' model configuration command:

hermes model

Choose the custom endpoint/manual URL option, then enter:

If Hermes asks for a provider type, choose the OpenAI-compatible option.

#Config-File Setup

Hermes also supports direct endpoint overrides in configuration. Use these values if you manage Hermes through a config file:

model:
  provider: custom
  base_url: "https://api.sogni.ai/v1"
  api_key: "YOUR_SOGNI_API_KEY"
  default: "qwen3.6-35b-a3b-gguf-iq4xs"

Use GET https://api.sogni.ai/v1/models for the current model list. Current public model IDs include:

• qwen3.6-35b-a3b-gguf-iq4xs
• qwen3.5-35b-a3b-abliterated-gguf-q4km

#Recommended Defaults

For coding and agent work, start with:

{
  "model": "qwen3.6-35b-a3b-gguf-iq4xs",
  "task_profile": "coding",
  "chat_template_kwargs": {
    "enable_thinking": true
  }
}

Hermes may not expose every provider-specific request field in its UI. If it does expose extra body fields and you want a coding-only setup, set sogni_tools: false so Sogni does not inject media tools. Leave Sogni tools enabled if you want Hermes to be able to request image, video, or music generation through Sogni.

#Add Creative Media Tools

The endpoint setup above routes Hermes' LLM calls to Sogni. If you want Hermes, Manus, Claude Code, or another skill-aware agent to call Sogni media tools directly, install the public Sogni Creative Agent Skill and point the agent at the repo's SKILL.md:

npm install -g @sogni-ai/sogni-creative-agent-skill@latest
sogni-agent --version

The skill wraps direct image/video/music generation, hosted /v1/chat/completions, durable /v1/chat/runs, /v1/creative-agent/workflows, replay utilities, persona/memory stores, and the same per-skill creative-agent manifests used by the rest of the Sogni platform. Use the OpenAI-compatible endpoint when Hermes should only choose a model; use the skill when the agent should execute Sogni creative workflows through sogni-agent.

#Cost Control

Hermes uses Sogni for LLM coding/reasoning work today; media tools spend tokens if you leave them enabled.

• Pick the billing token. Pass token_type: "spark" (purchased) or "sogni" (native) in Hermes' provider extra-body if your version exposes one. Defaults to auto.
• Vendor models require explicit naming. OpenAI GPT Image 2 and ByteDance Seedance 2.0 require Premium Spark and are never selected by the router on your behalf.
• For coding-only setups that should never spend on media, leave sogni_tools: false (above).

For approvals, hard caps, and the durable cost-approval flow, see Billing & Cost Control.

#Verify

Ask Hermes:

Which model and API base URL are you using?

Then try a short coding prompt:

Write a TypeScript function that validates an Ethereum address and explain edge cases.

#Troubleshooting

See the Hermes Agent configuration docs for the current custom endpoint settings.