Interface: OpenAIResponsesAdapterOptions
Defined in: src/batteries/llm/openai_responses/types.ts:825
Configuration options for the OpenAI Responses adapter.
Properties
| Property | Type | Default value | Description | Defined in |
|---|---|---|---|---|
apiKey? | string | undefined | API key for authenticating requests. | src/batteries/llm/openai_responses/types.ts:828 |
autoAck? | boolean | undefined | Whether the executor should call ctx.ack() itself when a generation completes with no tool calls. | src/batteries/llm/openai_responses/types.ts:866 |
background? | boolean | undefined | Request the generation run in the background (async). NOT SUPPORTED — this adapter has no polling/resumption logic for a queued/in_progress background response, so true is rejected at validation time (E_INVALID_OPENAI_RESPONSES_OPTIONS) rather than silently treated as a completed, empty answer. false/undefined (the default, synchronous request/response) are unaffected. | src/batteries/llm/openai_responses/types.ts:973 |
baseURL? | string | undefined | Base URL for the OpenAI-compatible Responses endpoint. | src/batteries/llm/openai_responses/types.ts:830 |
bucketOrder? | ChatCompletionsBucketOrder | undefined | Determines order of memory and retrievable buckets in history assembly. | src/batteries/llm/openai_responses/types.ts:848 |
contextWindow? | number | undefined | Size of the model's token context window. | src/batteries/llm/openai_responses/types.ts:850 |
fetch? | { (input: URL | RequestInfo, init?: RequestInit): Promise<Response>; (input: string | URL | Request, init?: RequestInit): Promise<Response>; } | undefined | Custom fetch implementation to use for HTTP requests. | src/batteries/llm/openai_responses/types.ts:846 |
forgeToolsFilter? | (forged: ToolRegistry, ctx: DispatchContext) => ToolRegistry | undefined | Optional hook to shape forged artifact-query tools before they merge into the visible tool set. | src/batteries/llm/openai_responses/types.ts:878 |
headers? | Record<string, string> | undefined | Extra HTTP headers to include with each request. | src/batteries/llm/openai_responses/types.ts:836 |
helpers? | Partial<OpenAIResponsesHelpers> | undefined | Optional overrides for OpenAI Responses helpers. | src/batteries/llm/openai_responses/types.ts:860 |
include? | OpenAIResponsesIncludable[] | undefined | Additional output data requested from the model. Remarks 'reasoning.encrypted_content' is auto-added when reasoningReplay === 'encrypted'; a caller value does not need to repeat it (de-duplicated). | src/batteries/llm/openai_responses/types.ts:927 |
localToolCallParser? | | ToolCallParserName | ToolCallParserFn | undefined | Optional fallback parser for tool calls the provider did not return structurally. | src/batteries/llm/openai_responses/types.ts:876 |
max_output_tokens? | number | undefined | Hard limit on generated output tokens. Remarks The API enforces an UNDOCUMENTED minimum of 16 (400 Invalid 'max_output_tokens'... Expected a value >= 16) — enforced in this battery's validation schema so a too-small value fails at config time, not dispatch time. | src/batteries/llm/openai_responses/types.ts:941 |
metadata? | Record<string, string> | undefined | Metadata key-value pairs forwarded to the provider. | src/batteries/llm/openai_responses/types.ts:961 |
model | string | undefined | Name of the model to use for generation. REQUIRED. | src/batteries/llm/openai_responses/types.ts:919 |
onPromptAssembled? | PromptAssembledObserverFn | undefined | Observe the fully-assembled request before dispatch. | src/batteries/llm/openai_responses/types.ts:872 |
onRawGeneration? | RawGenerationObserverFn | undefined | Observe the model's raw response for each completed generation. | src/batteries/llm/openai_responses/types.ts:870 |
organization? | string | undefined | Optional OpenAI organization id, sent as OpenAI-Organization. | src/batteries/llm/openai_responses/types.ts:832 |
parallel_tool_calls? | boolean | undefined | Allow the model to emit multiple tool calls in one turn. | src/batteries/llm/openai_responses/types.ts:943 |
project? | string | undefined | Optional OpenAI project id, sent as OpenAI-Project. | src/batteries/llm/openai_responses/types.ts:834 |
prompt_cache_key? | string | undefined | Vendor cache key for caching system/prefix content. | src/batteries/llm/openai_responses/types.ts:955 |
prompt_cache_retention? | "in_memory" | "24h" | undefined | Cache retention strategy for cached prefix content. | src/batteries/llm/openai_responses/types.ts:957 |
reasoning? | { effort?: "none" | "minimal" | "low" | "medium" | "high" | "xhigh"; summary?: "auto" | "concise" | "detailed"; } | undefined | Reasoning-effort / summary configuration for reasoning-capable models. | src/batteries/llm/openai_responses/types.ts:929 |
reasoning.effort? | "none" | "minimal" | "low" | "medium" | "high" | "xhigh" | undefined | - | src/batteries/llm/openai_responses/types.ts:930 |
reasoning.summary? | "auto" | "concise" | "detailed" | undefined | - | src/batteries/llm/openai_responses/types.ts:931 |
reasoningReplay? | ReasoningReplayMode | 'off' | Reasoning-replay mode. Remarks 'off' (default) — thoughts never replay as native reasoning items; they render as plain text instead. 'encrypted' — replay signed reasoning items, auto-adding 'reasoning.encrypted_content' to include. 'summary-only' — replay reasoning items without requesting the encrypted payload. This is real, scoped, stateful machinery (persisted opaque payloads, prefix-fence fingerprinting, adjacency enforcement, item-id manipulation) — see the battery's design notes for the reasoning/output-item pairing hazard this default guards against. | src/batteries/llm/openai_responses/types.ts:908 |
replayCompatibility? | readonly string[] | undefined | List of replay labels supported by the assistant. | src/batteries/llm/openai_responses/types.ts:858 |
requestTimeoutMs? | number | undefined | Request timeout in milliseconds for API calls. | src/batteries/llm/openai_responses/types.ts:842 |
retry? | ChatCompletionsRetryConfig | undefined | Configures request retry behavior. | src/batteries/llm/openai_responses/types.ts:844 |
safety_identifier? | string | undefined | Unique safety system identifier or configuration id. | src/batteries/llm/openai_responses/types.ts:959 |
selfIdentity? | string | undefined | Unique identity label for the assistant instance. | src/batteries/llm/openai_responses/types.ts:852 |
service_tier? | "default" | "auto" | "flex" | "scale" | "priority" | undefined | Service reliability tier for processing the request. | src/batteries/llm/openai_responses/types.ts:953 |
spoolStore? | SpoolStore | undefined | Backing store for string / Uint8Array tool returns. | src/batteries/llm/openai_responses/types.ts:862 |
stream? | boolean | undefined | Whether to stream the response via SSE. | src/batteries/llm/openai_responses/types.ts:838 |
stream_options? | { include_obfuscation?: boolean; } | undefined | Configuration options for response streaming. | src/batteries/llm/openai_responses/types.ts:975 |
stream_options.include_obfuscation? | boolean | undefined | - | src/batteries/llm/openai_responses/types.ts:975 |
streamIdleTimeoutMs? | number | undefined | Idle timeout in milliseconds for the stream before aborting. | src/batteries/llm/openai_responses/types.ts:840 |
strict? | boolean | undefined | Global strict default forwarded to every emitted tool declaration's strict field, when the caller hasn't already got a reason not to (kept undefined by default since strict: true requires additionalProperties: false and full required coverage that ADK's Joi-derived schemas don't guarantee). | src/batteries/llm/openai_responses/types.ts:915 |
strictToolChoice? | boolean | undefined | Whether a tool_choice forcing a forged ephemeral artifact-query tool throws instead of warning. | src/batteries/llm/openai_responses/types.ts:864 |
systemPromptChannel? | SystemPromptChannel | 'instructions' | Where the ADK-rendered system prompt is placed on the wire. Remarks 'instructions' (default) — the top-level instructions string field. 'developer-item' / 'system-item' — a leading developer/system-role input message item instead, for gateways that only understand the item form. Either way the CONTENT is always ADK-rendered; there is no consumer-facing escape hatch to hand-author it directly (see the instructions note below). | src/batteries/llm/openai_responses/types.ts:892 |
temperature? | number | undefined | Sampling temperature control. | src/batteries/llm/openai_responses/types.ts:945 |
text? | { format?: Record<string, unknown>; verbosity?: "low" | "medium" | "high"; } | undefined | Structured-output format / verbosity configuration. | src/batteries/llm/openai_responses/types.ts:965 |
text.format? | Record<string, unknown> | undefined | - | src/batteries/llm/openai_responses/types.ts:965 |
text.verbosity? | "low" | "medium" | "high" | undefined | - | src/batteries/llm/openai_responses/types.ts:965 |
thoughtSurfacing? | "all" | "all-self" | "latest-self" | undefined | Determines which thoughts are surfaced back to the model. | src/batteries/llm/openai_responses/types.ts:854 |
tokenEncoding? | | TokenEncodingId | null | undefined | Tokenizer encoding configuration for token counting. | src/batteries/llm/openai_responses/types.ts:856 |
tool_choice? | OpenAIResponsesToolChoice | undefined | Tool-choice directive. | src/batteries/llm/openai_responses/types.ts:963 |
toolCallIdFilter? | ToolCallIdFilterFn | undefined | Optional ingress hook for adapting provider tool-call ids; absent preserves the vendor id. | src/batteries/llm/openai_responses/types.ts:874 |
top_logprobs? | number | undefined | Top log-probability tokens limit. | src/batteries/llm/openai_responses/types.ts:949 |
top_p? | number | undefined | Nucleus sampling probability threshold. | src/batteries/llm/openai_responses/types.ts:947 |
truncation? | "auto" | "disabled" | undefined | Context-truncation strategy. | src/batteries/llm/openai_responses/types.ts:951 |
unsupportedMediaPolicy? | UnsupportedMediaPolicy | undefined | Policy for a @nhtio/adk!Media instance whose modality this wire cannot represent (audio, video). | src/batteries/llm/openai_responses/types.ts:868 |