Agent & Solver Plugins

March 18, 2026 · View on GitHub

OVOS provides two generations of NLP backend plugins:

Agent engines (opm.agents.*) — current, actively developed
Solver plugins (opm.solver.*) — deprecated, will be removed in the next major release

Both live in ovos_plugin_manager.templates.agents and ovos_plugin_manager.templates.solvers respectively.

Agent Engines (current)

All agent engines ultimately derive from AbstractAgentEngine (in ovos_plugin_manager.thirdparty.solvers / AbstractSolver).

Common `AbstractSolver` constructor

AbstractSolver(
    config: Optional[dict] = None,
    translator: Optional[LanguageTranslator] = None,
    detector: Optional[LanguageDetector] = None,
    priority: int = 50,
    enable_tx: bool = False,
    enable_cache: bool = False,
    internal_lang: Optional[str] = None,
)

Attribute	Description
`default_lang`	Internal language (BCP-47). Queries not in `supported_langs` are auto-translated.
`supported_langs`	Languages the plugin natively supports (from config).
`enable_tx`	If `True`, inputs are translated to `default_lang` and outputs translated back.
`enable_cache`	Cache results on disk via `JsonStorageXDG`.
`priority`	Solver priority (lower = higher priority). Default 50.

Auto-translation decorators

@auto_translate(translate_keys=[...]) and @auto_detect_lang(text_keys=[...]) can be applied to solver methods to transparently handle translation.

`ChatEngine`

Entry point: opm.agents.chat

Streaming/non-streaming conversational LLM backend.

Abstract method

def continue_chat(self, messages: List[AgentMessage],
                  session_id: str = "default",
                  lang: Optional[str] = None,
                  units: Optional[str] = None) -> AgentMessage

Generate a response given a list of AgentMessage objects.

`RetrievalEngine`

Entry point: opm.agents.retrieval

Base for retrieval-augmented generation backends.

`DocumentIndexerEngine`

Entry point: opm.agents.retrieval.documents

Index and search a corpus of documents.

`QAIndexerEngine`

Entry point: opm.agents.retrieval.qa

QA-specific retrieval over a question→answer corpus.

`ReRankerEngine`

Entry point: opm.agents.reranker

Rank a list of candidates by relevance to a query.

`SummarizerEngine`

Entry point: opm.agents.summarizer

Summarize a document.

`ExtractiveQAEngine`

Entry point: opm.agents.extractive_qa

Extract the best passage from a document that answers a question.

`NaturalLanguageInferenceEngine`

Entry point: opm.agents.nli

Determine whether a premise entails a hypothesis.

`AgentContextManager`

Entry point: opm.agents.memory

Manage conversational context / session memory for agent plugins.

from ovos_plugin_manager.templates.agents import AgentContextManager, AgentMessage

Abstract methods

def get_history(self, session_id: str) -> List[AgentMessage]
def update_history(self, new_messages: List[AgentMessage], session_id: str)

Property

@property
def system_prompt(self) -> str

Returns the system prompt string from config (config['system_prompt']).

`AgentMessage` dataclass

from ovos_plugin_manager.templates.agents import AgentMessage, MessageRole

Field	Type	Description
`role`	`MessageRole`	`SYSTEM`, `DEVELOPER`, `USER`, or `ASSISTANT`.
`content`	`str`	Message text.

Deprecated Solver Plugins

These classes are in ovos_plugin_manager.templates.solvers and will be removed in the next major release. Migrate to the corresponding agent engines.

Deprecated class	Replacement
`QuestionSolver`	`ChatEngine` / `RetrievalEngine`
`ChatMessageSolver`	`ChatEngine`
`CorpusSolver`	`DocumentIndexerEngine` / `QAIndexerEngine`
`TldrSolver`	`SummarizerEngine`
`EvidenceSolver`	`ExtractiveQAEngine`
`MultipleChoiceSolver`	`ReRankerEngine`
`EntailmentSolver`	`NaturalLanguageInferenceEngine`

`QuestionSolver` (deprecated)

Entry point: opm.solver.question

Abstract method

def get_spoken_answer(self, query: str, lang: Optional[str] = None, units: Optional[str] = None) -> Optional[str]

Public methods (with auto-translation)

spoken_answer(query, lang, units) — return spoken answer string
search(query, lang, units) — return data dict (cached)
visual_answer(query, lang, units) — return image path/URL
long_answer(query, lang, units) — return [{"title", "summary", "img"}, ...]
stream_utterances(query, lang, units) — yield answer sentences

`TldrSolver` (deprecated)

Entry point: opm.solver.summarization

Abstract method

def get_tldr(self, document: str, lang: Optional[str] = None) -> str

Public method

tldr(document, lang) — with auto-translation

`EvidenceSolver` (deprecated)

Entry point: opm.solver.reading_comprehension

Abstract method

def get_best_passage(self, evidence: str, question: str, lang: Optional[str] = None) -> str

Public method

extract_answer(evidence, question, lang) — with auto-translation

`MultipleChoiceSolver` (deprecated)

Entry point: opm.solver.multiple_choice

Abstract method

def rerank(self, query: str, options: List[str], lang: Optional[str] = None, return_index: bool = False) -> List[Tuple[float, Union[str, int]]]

Public method

select_answer(query, options, lang, return_index) — returns best option (or its index)

`EntailmentSolver` (deprecated)

Entry point: opm.solver.entailment

Abstract method

def check_entailment(self, premise: str, hypothesis: str, lang: Optional[str] = None) -> bool

Public method

entails(premise, hypothesis, lang) — with auto-translation