generated: '2026-07-20'
method: derived
source: openapi/osmosis-agent-improvement-openapi.json
entities:
- name: AgentInput
  role: request
  operation: enhance_task
  fields: [tenant_id, input_text, context, agent_type]
  required: [tenant_id, input_text]
- name: EnhancedResponse
  role: response
  operation: enhance_task
  fields: [response, metadata]
  required: [response]
- name: AgentKnowledgeUpload
  role: request
  operation: store_knowledge
  fields: [tenant_id, query, turns, success, agent_type, timestamp, metadata]
  required: [tenant_id, query, turns]
- name: AgentTurn
  role: nested
  fields: [turn, inputs, decision, memory, result]
  required: [turn, inputs, decision]
relationships:
- from: AgentKnowledgeUpload
  to: AgentTurn
  kind: has_many
  via: turns
grouping_keys:
- field: tenant_id
  notes: Partitions all agent knowledge/enhancement by tenant.
- field: agent_type
  notes: Optional grouping of interactions by agent type.
- field: query
  notes: store_knowledge groups turns under an intent/query; delete_by_intent removes by it.
notes: |
  The Agent Improvement API models agent interaction memory: an agent's turns
  (inputs/decision/memory/result) are uploaded under a query/intent for a tenant,
  then enhance_task returns an improved response drawing on stored knowledge.