AIML Chatbot

An OVOS question solver plugin that answers with an AIML chatbot. It uses the Alice chatbot brain to hold open-ended conversation. It answers utterances that no other skill or solver handles, so most utterances get some response.

Examples

• "Do you like ice cream"
• "Do you like dogs"
• "I have a jump rope"

Install

pip install ovos-solver-aiml-plugin

Usage

from ovos_solver_aiml_plugin import AIMLChatEngine
from ovos_plugin_manager.templates.agents import AgentMessage, MessageRole

engine = AIMLChatEngine({"lang": "en-us"})
reply = engine.continue_chat([AgentMessage(role=MessageRole.USER, content="hello")])
print(reply.content)
# Hi there!

Language support

The bundled AIML brain is English. Brain files live under ovos_solver_aiml_plugin/aiml_data/<lang>/ (bundled) or ~/.local/share/aiml/<lang>/ (user-supplied).

Opt-in query translation (enable_tx)

Set enable_tx: true in the plugin config to answer non-English queries. The plugin translates the user utterance into the brain language (English), answers it, then translates the answer back.

This is off by default, so a plain install needs no translate plugin and English deployments pay no extra cost.

{
  "ovos-solver-aiml-plugin": {
    "lang": "pt-pt",
    "enable_tx": true,
    "translate_plugin": "ovos-translate-plugin-server"
  }
}

The plugin loads the translator lazily, on the first non-English turn. If the translator fails to load, or a translation fails, the plugin answers in the original text instead of raising an error.

Translator plugin note: translation agents tend to be instantiated repeatedly, so a local model-based translate plugin pays its full model-load cost each time. Use a remote translate service, such as ovos-translate-plugin-server.

Converter scripts

The bundled aiml_data/<lang>/*.aiml files are the brain. The helper scripts in scripts/ are an optional convenience for authoring or inspecting content in OVOS notation. They are not part of the runtime and not wired into CI:

• brain_to_locale.py: exports the cleanly-mappable subset of an AIML or RiveScript brain to paired .intent / .dialog files ({query} slot for wildcards).
• locale_to_brain.py: the reverse. It regenerates a brain from such files.

The conversion is deliberately partial. Entries that rely on constructs with no direct OVOS equivalent (<srai>, <condition>, <star>, topic state, and so on) are skipped, and exports never emit residual <…> markup. See docs/converters.md.

Related projects

• OpenVoiceOS/ovos-plugin-manager: loads this plugin through the opm.agents.chat entry point.
• OpenVoiceOS/ovos-translate-plugin-server: recommended remote translate plugin for enable_tx.

Docker

This repo publishes ghcr.io/openvoiceos/ovos-solver-plugin-aiml, a standalone ovos-persona-server that serves one persona, AimlBot, backed by the AIMLChatEngine in this plugin. Matching runs entirely offline against the bundled aiml_data, so the image needs no key and no network access to answer.

docker run -p 8391:8337 ghcr.io/openvoiceos/ovos-solver-plugin-aiml:dev

curl http://localhost:8391/v1/models
curl http://localhost:8391/v1/chat/completions \
    -H "Content-Type: application/json" \
    -d '{"model": "AimlBot", "messages": [{"role": "user", "content": "hello"}]}'

A compose snippet:

services:
  ovos-aiml-persona:
    image: ghcr.io/openvoiceos/ovos-solver-plugin-aiml:dev
    ports:
      - "8391:8337"
    restart: unless-stopped

The image builds on every pull request touching Dockerfile, .dockerignore, pyproject.toml, or the docker workflow itself (build only, no push), and publishes on pushes to master (latest), dev (dev), and version tags. See the docker workflow.

License

Apache License 2.0