Agent catalogue¶

API version

Chatbot API: 2.41.7

This catalogue describes every agent node compiled into the Oria graph. It is a functional view: an agent is described by the user need it owns, the business tools it can choose and the result it produces.

The supervisor selects one of seven specialists, also available as dedicated subgraphs. Career Chat can then hand off to its own Knowledge Base Deep Agent.

Routing node¶

Supervisor¶

The supervisor identifies the intent of an allowed request and transfers the turn to one enabled specialist. It receives conversation context, the current tenant agent configuration and a dynamic routing prompt.

It uses OpenAI Decisions when the configured mini profile supports it, otherwise a structured choice from the same configured provider, to select one enabled agent. The application executes the transition and persists active_agent. The supervisor has no business tools and cannot write a user-facing answer. See Decisions routing for provider requirements, context handling and failure behavior.

Guidance and development¶

Agent node	Functional responsibility	Business tools	Main user-facing result
`career_chat_agent`	Discuss career and internal-knowledge questions using relevant profile, document, job, training and skills context	Internal knowledge retrieval; selective profile retrieval; document retrieval; job/skill mapping; job/training retrieval	Conversational career guidance grounded in available business data
`knowledge_base_agent`	Research internal documents and configured public pages, compare sources and synthesize a grounded answer	Bounded RAG/web research	Exact final answer, or a return to Career Chat for clarification
`profile_analysis_agent`	Analyze professional positioning, strengths and priority development areas	Profile analysis	Structured profile-analysis card and explanatory text
`task_helper_agent`	Explain how to complete a professional task such as document, office-tool or technical work	Task helper	Direct step-by-step assistance

Profile Analysis can be reached from the supervisor and, for profile action-plan elaboration, through the supervisor's Training Finder destination adjustment.

Discovery and recommendation¶

Agent node	Functional responsibility	Business tools	Main user-facing result
`opportunities_suggestion_agent`	Search opportunities from natural language or recommend them from an objective	Opportunity search; objective-based suggestion; objective retrieval	Text plus structured opportunity suggestions
`training_finder_agent`	Search learning content, create learning-plan suggestions and propose a role-play when relevant	Selective profile retrieval; training search; simulation suggestion	Text plus structured training or simulation data
`talent_finder_agent`	Find internal people matching explicit criteria	Talent search	Ranked talent suggestions and explanation
`person_disambiguation_agent`	Resolve which employee a name refers to when several records match	Person disambiguation	Candidate identities/IDs or a resolved person

The opportunity, training and talent agents retain identifiers already returned in graph state so a “show more” request can avoid duplicates.

Relationships at a glance¶

                         supervisor
                              |
          +--------------+----+--------+-------------+
          |              |             |             |
       guidance       discovery     training       people
          |              |             |             |
   career / profile  opportunity    finder +      talent /
   / task                           simulation   disambiguate

The diagram groups responsibilities; it does not add extra graph nodes. The exact runtime transitions are documented in agent orchestration.

Runtime restrictions¶

Agent availability can differ by tenant. The functional catalogue represents everything compiled into AI Hub, while chatbot.agents determines which destinations are offered for an individual request. Some agents also support tenant-level tool exclusions. See tenant-dependent availability.