Skip to content

Run Several Personas on One Server

Alpha channel only

The prerelease channel carries this content. A stable or testing install does not have it, and a plain pip install does not get it: the install needs --pre. Read Release Channels before you mix channels.

This page serves several personas from one ovos-persona-server process, with a separate memory for each user. You want to reach your personas over HTTP, from apps that speak the OpenAI or Ollama API.

What the setup includes

  • The model field of each request selects the persona.
  • With CHAT_MEMORY=transparent, the user field selects the memory of one person.
  • Each persona keeps its own memory, so a user has a separate history with each persona.

Steps

Community servers are best-effort demos

The public OVOS servers exist for easy onboarding and demos only. They are best-effort, not optimized, carry no uptime guarantees, and may vanish at any time. OVOS will be slow and unreliable if you rely on them. The official recommendation is to self-host — or skip servers entirely: fully offline plugins exist for everything.

  1. Install the packages:

    pip install --pre ovos-persona-server ovos-openai-plugin

  2. Save two personas, ~/personas/ada.json and ~/personas/bo.json. This is ada.json:

    {
  "name": "ada",
  "handlers": ["ovos-chat-openai-plugin"],
  "ovos-chat-openai-plugin": {
    "api_url": "https://llama.smartgic.io/v1",
    "key": "sk-none",
    "model": "qwen2.5:7b",
    "system_prompt": "Your name is Ada. You answer in at most fifteen words. When asked your name, say Ada."
  }
}

    For bo.json, change the name to bo and change the system prompt to the name Bo.

  3. Start the server. It loads every *.json file in the directory:

    CHAT_MEMORY=transparent ovos-persona-server --personas-dir ~/personas

  4. List the personas:

    curl -s localhost:8337/openai/v1/models | python3 -c 'import sys,json; print([m["id"] for m in json.load(sys.stdin)["data"]])'

  5. Send requests to both personas for one user:

    ask() {
  curl -s localhost:8337/openai/v1/chat/completions \
    -H 'content-type: application/json' \
    -d "{\"model\":\"$1\",\"user\":\"$2\",\"messages\":[{\"role\":\"user\",\"content\":\"$3\"}]}" |
    python3 -c 'import sys,json; print(json.load(sys.stdin)["choices"][0]["message"]["content"])'
}
ask ada alice "What is your name? Answer with your name only."
ask bo alice "What is your name? Answer with your name only."
ask ada alice "Remember this: my cat is called Pixel. Reply only with OK."
ask ada alice "What is my cat called? Answer with the name only."
ask bo alice "What is my cat called? If you do not know, say UNKNOWN."

Output of a run:

['ada', 'bo']
Ada.
Bo.
OK.
Pixel.
UNKNOWN.

bo does not see what Alice told ada. An unknown model value gets HTTP 404 with the list of available personas.

The server trusts the user value and does not check it, so this setup fits trusted clients only. On a server that untrusted clients share, keep CHAT_MEMORY off. See Memory, RAG & embeddings.

Troubleshooting

Problem Cause Fix
The server does not start Two persona files have the same name Give each persona a unique name
All users share one history The client sends no user field Send a different user value for each person
The server forgets between requests CHAT_MEMORY is not transparent Start the server with CHAT_MEMORY=transparent