@anvia/studio API reference

All supported exports come from @anvia/studio. Runtime route handlers and UI modules are internal.

Studio

class Studio implements AnviaStudio {
  constructor(targets?: StudioTarget[], options?: StudioOptions)
  get app(): import('hono').Hono
  fetch(request: Request): Response | Promise<Response>
  config(): StudioConfig
  traceObserver(): StudioTraceObserver
  start(options?: StudioServeOptions): this
  serve(options?: StudioServeLifecycleOptions): Promise<void>
  close(): void
}

start() starts a server and returns immediately. serve() follows a caller-owned lifecycle and resolves after shutdown. fetch() makes the same application usable in runtimes or servers that accept a Fetch handler.

type StudioServeOptions = {
  port?: number
  hostname?: string
  log?: boolean
  handleSignals?: boolean
}

type StudioServeLifecycleOptions = Omit<StudioServeOptions, 'handleSignals'> & {
  signal?: AbortSignal
  onShutdown?: () => void | Promise<void>
}

Targets and configuration

type StudioTarget = Agent | Pipeline<any, any>

type StudioOptions = {
  evals?: StudioEvalSuite<any, any, any>[]
  quickPrompts?: Record<string, string[]>
  stores?: StudioStores
  ui?: boolean | StudioUiOptions
  models?: StudioModelConfig
}

StudioUiOptions configures path, rootRoutes, title, redirectRoot, clientScript, and protectShell. StudioStores independently selects session, trace, pipeline-log, and pipeline-run stores; session and pipeline stores may be disabled with false where the type permits it.

The target configuration types are StudioAgent, StudioAgentConfig, StudioAgentRuntimeSummary, StudioPipeline, StudioPipelineConfig, StudioPipelineDetail, and StudioConfig.

Models

type StudioModelRef = string | { provider: string; model: string }

type StudioModelProvider = {
  id: string
  name?: string
  defaultModel?: string
  models?: StudioModelDefinition[]
  createCompletionModel(model: string): CompletionModel | StreamingCompletionModel
  listModels?: () => Promise<ModelList>
  metadata?: JsonObject
}

type StudioModelConfig = {
  providers: StudioModelProvider[]
  default?: StudioModelRef
  agents?: Record<string, StudioAgentModelPolicy>
}

Supporting public types include StudioModelDefinition, StudioModelModality, StudioModelModalities, StudioAgentModelPolicy, StudioModelSummary, StudioModelProviderConfig, StudioAgentModelPolicyConfig, StudioModelsConfig, and StudioAgentModelsSummary.

Stores

function createInMemoryStudioStore():
  & StudioSessionStore
  & StudioTraceStore
  & StudioPipelineLogStore
  & StudioPipelineRunStore

type SqliteSessionStoreOptions = { path?: string }

function createSqliteSessionStore(options?: SqliteSessionStoreOptions):
  & StudioSessionStore
  & StudioTraceStore
  & StudioPipelineLogStore
  & StudioPipelineRunStore

Store contracts are intentionally public so applications can provide another backend:

Contract	Responsibilities
StudioMemoryStore	Append messages and record memory errors.
StudioSessionStore	Create, list, load, update, and delete Studio sessions.
StudioTraceStore	Save, get, and list traces.
StudioPipelineLogStore	Append and list ordered pipeline log entries.
StudioPipelineRunStore	Save, get, and list replayable pipeline runs.

Their related input, list-options, record, status, summary, paging, and event types are public and listed by family below.

Trace observer

class StudioTraceObserver implements AgentObserver {
  constructor(options: StudioTraceObserverOptions)
  startRun(args: AgentRunStartArgs): AgentRunObserver
}

type StudioTraceObserverOptions = {
  store: StudioTraceStore | (() => StudioTraceStore | undefined) | undefined
}

function traceSummary(trace: StudioTrace): StudioTraceSummary

Trace records use StudioTrace, StudioTraceSummary, StudioTraceObservation, StudioTraceStatus, StudioTraceObservationKind, StudioTraceListOptions, and StudioSessionTraceListOptions.

Requests, events, and errors

AgentRunRequest accepts one message plus optional history; AgentRunUIRequest accepts a UI message array. Both support session, streaming, max-turn, tool-concurrency, model, metadata, and trace controls. AgentRunResponse aliases PromptResponse.

AgentRunStreamEvent combines core agent events with Studio approval, question, session-log, pipeline-log, and pipeline-final events.

type StudioErrorCode =
  | 'bad_request'
  | 'conflict'
  | 'not_found'
  | 'payload_too_large'
  | 'unsupported_capability'
  | 'internal_error'

StudioErrorResponse wraps the code, message, and optional JSON details.

Public type inventory

The remaining public types are grouped by the Studio surface that produces or consumes them.

Surface	Public types
Capabilities and status	StudioCapability, StudioCapabilityConfig, StudioStatusSummary, StudioStores, StudioConfig
Evaluations	StudioEvalSuite, StudioEvalSuiteConfig, StudioEvalCasePreview, StudioEvalMetricSummary, StudioEvalRunRequest, StudioEvalRunResponse
Tools and MCP	StudioAgentToolSource, StudioAgentToolApprovalMetadata, StudioAgentToolMetadata, StudioAgentToolsSummary, StudioToolRunRequest, StudioToolRunResponse, StudioAgentMcpToolMetadata, StudioAgentMcpServerMetadata, StudioAgentMcpsSummary
Approvals	StudioToolApproval, StudioToolApprovalDecision, StudioToolApprovalStatus, StudioToolApprovalTranscript, StudioToolApprovalRequestEvent, StudioToolApprovalResultEvent
Questions	StudioToolQuestion, StudioToolQuestionChoice, StudioToolQuestionPrompt, StudioToolQuestionAnswer, StudioToolQuestionStatus, StudioToolQuestionTranscript, StudioToolQuestionRequestEvent, StudioToolQuestionResultEvent
Sessions and transcripts	StudioSession, StudioSessionSummary, StudioSessionCreateInput, StudioSessionListOptions, StudioSessionRunStatus, StudioSessionRunTranscriptInput, StudioTranscriptEntry, StudioTranscriptChatEntry, StudioTranscriptReasoningEntry, StudioTranscriptToolEntry, StudioTranscriptAttachment, StudioTranscriptChildAgentEvent
Session logs	StudioSessionLogEntry, StudioSessionLogAppendInput, StudioSessionLogListOptions, StudioSessionLogLevel, StudioSessionLogCategory, StudioSessionLogEvent
Pipeline runs	StudioPipelineRunRequest, StudioPipelineReplayRequest, StudioPipelineRunResponse, StudioPipelineRunRecord, StudioPipelineRunSaveInput, StudioPipelineRunListOptions, StudioPipelineRunGetOptions, StudioPipelineRunStatus, StudioPipelineFinalEvent
Pipeline logs	StudioPipelineLogEntry, StudioPipelineLogAppendInput, StudioPipelineLogListOptions, StudioPipelineLogLevel, StudioPipelineLogCategory, StudioPipelineLogEvent
Knowledge	StudioAgentKnowledgeConfig, StudioKnowledgeSourceKind, StudioKnowledgeSourceSummary, StudioStaticKnowledgeDocument, StudioKnowledgeEvidence, StudioKnowledgeEvidenceDocument, StudioKnowledgeItem, StudioKnowledgeItemKind, StudioKnowledgeItemsPage, StudioKnowledgeSummary
Memory inspection	StudioMemoryContext, StudioMemoryAppendInput, StudioMemoryErrorInput, StudioMemoryUserSummary, StudioMemoryConversationSummary, StudioMemoryConversationsPage, StudioMemoryUsersPage, StudioMemoryConversationMessages, StudioMemoryConversationSteps, StudioMemoryMessageRecord, StudioMemorySourceKind, StudioMemorySourceSummary, StudioMemorySourcesPage, StudioMemorySourceConversationSummary, StudioMemorySourceConversationsPage, StudioMemorySourceUsersPage, StudioMemorySourceConversationMessages, StudioMemorySourceConversationSteps
Sandboxes	StudioSandboxCapabilities, StudioSandboxSummary, StudioSandboxesSummary, StudioSandboxFileType, StudioSandboxFileEntry, StudioSandboxFilesResponse, StudioSandboxPort, StudioSandboxPortsResponse, StudioSandboxProcessStatus, StudioSandboxProcess, StudioSandboxProcessesResponse, StudioSandboxProcessLogsResponse
Observability	StudioObservabilityEventType, StudioObservabilityEvent, AgentTraceInfo, AgentTraceOptions

The exact fields for these transport and storage types are part of the published declarations. Prefer consuming the types directly instead of reproducing request shapes locally.