QuestionSolversService

Module: ovos_persona.solvers.QuestionSolversService

Manages the ordered pipeline of utterance handler plugins for a Persona. Tries each handler in order and returns the first successful response.

---

Supported Plugin Types

QuestionSolversService supports six plugin categories, all discovered via ovos-plugin-manager:

Entry point group	Base class	Chat history support	Streaming
opm.solver	QuestionSolver	No, last message only	Yes (stream_utterances)
opm.solver.chat	ChatMessageSolver	Yes	Yes (stream_chat_utterances)
opm.agents.chat	ChatEngine	Yes	Yes (stream_sentences)
opm.agents.chat.multimodal	MultimodalChatEngine	Yes	Yes (stream_sentences)
opm.agents.retrieval	RetrievalEngine	No, last message only	No
opm.agents.indexer.document	DocumentIndexerEngine	No, last message only	No
opm.agents.indexer.qa	QAIndexerEngine	No, last message only	No

QuestionSolversService treats all plugin types the same way. It dispatches to the method that matches the plugin's class.

---

Plugin Ordering

The sort_order parameter (set from the persona's handlers list) controls the order in which plugins are tried:

service = QuestionSolversService(
    config={"ovos-solver-openai-plugin": {"enabled": True}},
    sort_order=["ovos-solver-wolfram-alpha-plugin", "ovos-solver-openai-plugin"]
)

If sort_order is empty, plugins are sorted by their priority attribute (lower = tried first).

stream_completion stops after the first handler that yields at least one sentence. It does not try the remaining handlers.

---

get_utterance_handler_plugins()

from ovos_persona.solvers import get_utterance_handler_plugins

plugins = get_utterance_handler_plugins()
# → {entry_point_name: PluginClass, ...}

Merges all six plugin entry point groups into a single dict. Used by Persona.__init__ to build the handler config.

---

chat_completion(messages, lang, units) → Optional[str]

Single-shot: tries each handler in order and returns the first non-empty string response.

Dispatch logic:

Plugin type	Method called
ChatEngine / MultimodalChatEngine	continue_chat(messages, ...) → response.content
RetrievalEngine / DocumentIndexerEngine / QAIndexerEngine	query(last_message, k=1) → first document
ChatMessageSolver	get_chat_completion(messages, ...)
QuestionSolver	spoken_answer(last_message, ...) (no history)

---

stream_completion(messages, lang, units) → Iterable[str]

Streaming: yields response sentences from the first handler that produces output. Stops trying further handlers once any handler yields at least one sentence.

Dispatch logic:

Plugin type	Method called
ChatEngine / MultimodalChatEngine	stream_sentences(messages, ...)
RetrievalEngine / DocumentIndexerEngine / QAIndexerEngine	query(last_message, k=1) → yields document
ChatMessageSolver	stream_chat_utterances(messages, ...)
QuestionSolver	stream_utterances(last_message, ...) (no history)

---

Configuration

Each handler plugin is configured by its entry point name inside the persona config:

{
  "handlers": ["ovos-solver-openai-plugin", "ovos-solver-wolfram-alpha-plugin"],
  "ovos-solver-openai-plugin": {
    "enabled": true,
    "api_url": "https://api.openai.com/v1",
    "key": "sk-..."
  },
  "ovos-solver-wolfram-alpha-plugin": {
    "enabled": true,
    "key": "..."
  }
}

Handlers not in the persona's handlers list are disabled regardless of their enabled key. If a handler listed in handlers is not installed, ImportError is raised during load_plugins().

---

← Persona · Home · Memory →