feat(openai-official): add OpenAI Responses API module - #3078
Conversation
There was a problem hiding this comment.
🟡 Changes recommended
There are correctness/usability issues in provider configuration and connection-field validation (env base URL resolution and blank-value normalization) that should be fixed before merging.
Once you've addressed the issues Copilot identified, you can request another Copilot review.
Pull request overview
This PR introduces a new model extension module, agentscope-extensions-model-openai-official, integrating OpenAI models via the official OpenAI Java SDK against the Responses API, and wires it into docs, distributions, and the e2e harness.
Changes:
- Adds the
openai-officialprovider implementation (request mapping, response parsing, streaming assembly, error translation, credential type, SPI registration). - Updates distributions/BOM and extensions aggregator to include the new module, and adds e2e provider coverage.
- Updates v2 docs (EN/ZH) and TOC to document the new provider and module.
File summaries
| File | Description |
|---|---|
| docs/v2/zh/integration/overview.md | Adds openai-official provider row to the ZH integration overview table. |
| docs/v2/zh/integration/model/openai-official.md | New ZH provider doc page for the official SDK / Responses API module. |
| docs/v2/zh/integration/model/index.md | Links the new ZH provider doc in the model index. |
| docs/v2/zh/docs/building-blocks/model.md | Updates ZH “model building blocks” doc to include the new module/provider references. |
| docs/v2/zh/docs/building-blocks/agent.md | Mentions openai-official as a supported ModelRegistry provider in ZH agent docs. |
| docs/v2/en/integration/overview.md | Adds openai-official provider row to the EN integration overview table. |
| docs/v2/en/integration/model/openai-official.md | New EN provider doc page for the official SDK / Responses API module. |
| docs/v2/en/integration/model/index.md | Links the new EN provider doc in the model index. |
| docs/v2/en/docs/building-blocks/model.md | Updates EN “model building blocks” doc to include the new module/provider references. |
| docs/v2/en/docs/building-blocks/agent.md | Mentions openai-official as a supported ModelRegistry provider in EN agent docs. |
| docs/_toc.yml | Adds TOC entries for the new EN/ZH provider doc pages. |
| agentscope-extensions/agentscope-extensions-model/pom.xml | Enables the new agentscope-extensions-model-openai-official module in the extensions reactor build. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/test/java/io/agentscope/extensions/model/openaiofficial/TestSdkFixtures.java | Shared SDK object/exception fixtures for unit tests. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/test/java/io/agentscope/extensions/model/openaiofficial/ResponsesStreamingAssemblerTest.java | Unit tests for streaming event assembly to ChatResponse chunks. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/test/java/io/agentscope/extensions/model/openaiofficial/ResponsesResponseParserTest.java | Unit tests for non-streaming Response → ChatResponse parsing. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/test/java/io/agentscope/extensions/model/openaiofficial/ResponsesMultiAgentFormatterTest.java | Unit tests for multi-agent conversation merging/formatting. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/test/java/io/agentscope/extensions/model/openaiofficial/ResponsesHelperTest.java | Unit tests for metadata/usage extraction helpers. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/test/java/io/agentscope/extensions/model/openaiofficial/OpenAISdkClientFactoryTest.java | Unit tests for official SDK client construction and error wrapping. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/test/java/io/agentscope/extensions/model/openaiofficial/OpenAIResponsesChatModelTest.java | Unit tests for streaming/non-streaming flows, retry predicate behavior, and builder boundaries. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/test/java/io/agentscope/extensions/model/openaiofficial/OpenAIOfficialModelProviderTest.java | Unit tests for SPI provider supports/create behavior and advanced options. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/test/java/io/agentscope/extensions/model/openaiofficial/OpenAIOfficialModelExceptionTest.java | Unit tests for provider exception type and retryable-status classification. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/test/java/io/agentscope/extensions/model/openaiofficial/OpenAIErrorTranslatorTest.java | Unit tests for SDK exception translation coverage. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/test/java/io/agentscope/extensions/model/openaiofficial/credential/OpenAIOfficialCredentialTest.java | Unit tests for credential JSON round-tripping and validation. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/main/resources/META-INF/services/io.agentscope.core.model.spi.ModelProvider | Registers OpenAIOfficialModelProvider via ServiceLoader. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/main/java/io/agentscope/extensions/model/openaiofficial/ResponsesStreamingAssembler.java | Implements Responses API streaming event → incremental ChatResponse assembly. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/main/java/io/agentscope/extensions/model/openaiofficial/ResponsesResponseParser.java | Parses non-streaming SDK Response objects into ChatResponse. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/main/java/io/agentscope/extensions/model/openaiofficial/ResponsesRequestMapper.java | Maps AgentScope history/options/tools into SDK ResponseCreateParams. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/main/java/io/agentscope/extensions/model/openaiofficial/ResponsesMultiAgentFormatter.java | Implements multi-agent history grouping/merge into Responses API input items. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/main/java/io/agentscope/extensions/model/openaiofficial/ResponsesHelper.java | Extracts response metadata and usage into AgentScope-friendly structures. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/main/java/io/agentscope/extensions/model/openaiofficial/OpenAISdkClientFactory.java | Centralizes OpenAI SDK client creation with retries disabled and optional headers/timeout. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/main/java/io/agentscope/extensions/model/openaiofficial/OpenAIResponsesChatModel.java | Main ChatModelBase implementation for Responses API (streaming + non-streaming). |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/main/java/io/agentscope/extensions/model/openaiofficial/OpenAIOfficialModelProvider.java | Adds SPI provider supporting openai-official:<model> resolution. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/main/java/io/agentscope/extensions/model/openaiofficial/OpenAIOfficialModelException.java | Provider-specific exception type implementing ModelHttpException. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/main/java/io/agentscope/extensions/model/openaiofficial/OpenAIOfficialConstants.java | Centralizes provider id, metadata keys, and additionalBodyParams whitelist. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/main/java/io/agentscope/extensions/model/openaiofficial/OpenAIErrorTranslator.java | Translates SDK exceptions into OpenAIOfficialModelException. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/main/java/io/agentscope/extensions/model/openaiofficial/credential/OpenAIOfficialCredential.java | Adds credential type for the new provider. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/pom.xml | New module POM with openai-java dependency and test deps. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-e2e-tests/src/test/java/io/agentscope/core/e2e/providers/OpenAIOfficialResponsesProvider.java | Adds e2e provider(s) for official SDK Responses API path. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-e2e-tests/src/test/java/io/agentscope/core/e2e/ProviderFactory.java | Registers the new e2e providers in the factory list. |
| agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-e2e-tests/pom.xml | Adds test-scope dependency on the new openai-official module. |
| agentscope-distribution/agentscope-bom/pom.xml | Adds the new module to the published BOM. |
| agentscope-distribution/agentscope-all/pom.xml | Adds the new module as an optional dependency to the “all” distribution. |
| agentscope-core/src/main/java/io/agentscope/core/model/ModelUtils.java | Wraps timeouts with a TimeoutException cause for retry predicates to detect. |
| agentscope-core/src/main/java/io/agentscope/core/model/ModelContextWindows.java | Adds context window mappings for new OpenAI/GLM model names. |
Review details
Suppressed comments (2)
agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official/src/main/java/io/agentscope/extensions/model/openaiofficial/OpenAIResponsesChatModel.java:175
- validateConnectionFields also treats a blank baseUrl as an override mismatch (null vs ""), but OpenAISdkClientFactory already normalizes blank baseUrl to “use SDK default”. This can incorrectly fail-fast on semantically equivalent values.
String effectiveBaseUrl = effectiveOptions.getBaseUrl();
if (effectiveBaseUrl != null && !Objects.equals(effectiveBaseUrl, baseUrl)) {
throw new OpenAIOfficialModelException(
agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-e2e-tests/src/test/java/io/agentscope/core/e2e/providers/OpenAIOfficialResponsesProvider.java:91
- Same mismatch here: the Javadoc says “GPT-5.4-mini” but the provider uses "gpt-5.4".
/** GPT-5.4-mini with Multi-Agent Formatter. */
- Files reviewed: 46/46 changed files
- Comments generated: 4
- Review effort level: Lite
💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.
|
LGTM |
97b8f26 to
30a9821
Compare
Add agentscope-extensions-model-openai-official module integrating OpenAI models via the official OpenAI Java SDK (com.openai:openai-java) against the Responses API. Includes request mapping, response parsing, streaming assembly, error translation, credential type, SPI registration, e2e test provider, bilingual docs, and distribution/BOM updates. Resolve documentation conflicts from Mintlify migration (#3081): - Replace deleted _toc.yml with docs.json entries - Adapt agent.md/index.md/overview.md to Mintlify syntax - Add Mintlify front matter to openai-official.md pages
Aias00
left a comment
Aias00
left a comment
There was a problem hiding this comment.
已审。重点检查了新增 openai-official Responses 模块的请求映射、非流式解析、流式事件拼装、usage/metadata/finishReason、工具调用回放、错误转换、provider SPI、BOM/distribution 接入和相关测试覆盖。\n\n本地验证：mvn -pl agentscope-extensions/agentscope-extensions-model/agentscope-extensions-model-openai-official -am test -DskipITs 通过，274 tests, 0 failures/errors, 2 skipped。\n\n未发现需要阻塞合并的问题。
This comment was marked as duplicate.
This comment was marked as duplicate.
Sorry, something went wrong.
This comment was marked as duplicate.
This comment was marked as duplicate.
Sorry, something went wrong.
This comment was marked as duplicate.
This comment was marked as duplicate.
Sorry, something went wrong.
Aias00
left a comment
Aias00
left a comment
There was a problem hiding this comment.
Requesting changes after re-checking PR head 0bcf019. I found release-blocking correctness issues that should be fixed before merge.
[HIGH] Multi-agent formatter drops encrypted reasoning replay
OpenAIResponsesChatModel sends all history through formatter::formatHistory when a ResponsesMultiAgentFormatter is configured (OpenAIResponsesChatModel.java:112-122). In that path, ordinary assistant messages are classified as AGENT_CONVERSATION unless they contain tool blocks (ResponsesMultiAgentFormatter.java:172-181), and mergeAgentConversation rewrites them into a user <history> message (ResponsesMultiAgentFormatter.java:209-237). processMessage then serializes ThinkingBlock as visible text (ResponsesMultiAgentFormatter.java:305-312) and never delegates to ResponsesRequestMapper.mapAssistantMessage, which is the only code path that emits the Responses reasoning replay item from openai.reasoning.encrypted_content metadata (ResponsesRequestMapper.java:433-445, ResponsesRequestMapper.java:477-493).
This breaks stateless reasoning replay exactly for the documented reasoning + multi-agent formatter usage. Please preserve assistant messages with encrypted reasoning metadata as Responses reasoning input items, or route them through a passthrough/delegation path, and add a regression test that configures OpenAIResponsesChatModel with new ResponsesMultiAgentFormatter() and asserts second-turn input contains the encrypted ResponseReasoningItem.
[MEDIUM] Schema-level strict mode is ignored for JSON schema structured output
JsonSchema exposes strict (JsonSchema.java:91-96), but ResponsesRequestMapper only writes schemaBuilder.strict(...) from the model-level strictJsonSchema flag (ResponsesRequestMapper.java:178-187). A caller setting ResponseFormat.json_schema with JsonSchema.strict(true) silently gets a request without strict: true unless they also configure the model builder flag.
Please resolve strict as schema-level first, then builder-level fallback, e.g. schema.getStrict() != null ? schema.getStrict() : strictJsonSchema, and add tests for schema-level override plus builder-level fallback.
[HIGH] all-in-one shaded artifact can lose ModelProvider service registration
The new module is added to agentscope-all (agentscope-all/pom.xml:111-116) and has its own SPI file (agentscope-extensions-model-openai-official/src/main/resources/META-INF/services/io.agentscope.core.model.spi.ModelProvider:1). But the shade config only includes artifacts and has no ServicesResourceTransformer or equivalent service merge (agentscope-all/pom.xml:449-466). There are multiple model-provider service descriptors across extension modules, so the shaded jar can keep only one descriptor and drop OpenAIOfficialModelProvider, making ModelRegistry discovery fail for users depending on the all-in-one agentscope artifact.
Please add a service resource transformer to the shade plugin and a packaging regression check that the shaded jar's META-INF/services/io.agentscope.core.model.spi.ModelProvider contains io.agentscope.extensions.model.openaiofficial.OpenAIOfficialModelProvider.
CI is green, but these are behavioral/package correctness issues not covered by the current checks.
Aias00
left a comment
Aias00
left a comment
There was a problem hiding this comment.
One additional blocking issue affects the default formatter path as well.
[HIGH] The production ReActAgent path discards encrypted reasoning metadata before the next turn
ResponsesResponseParser stores openai.reasoning.encrypted_content only in the terminal ChatResponse.metadata (lines 105-106), and the streaming path does the same in ResponsesStreamingAssembler (lines 261-263). However, the production consumer, ReasoningContext.processChunk, never reads chunk.getMetadata(); buildFinalMessage() creates a fresh metadata map containing only MessageMetadataKeys.CHAT_USAGE. The resulting assistant Msg therefore cannot contain the key that ResponsesRequestMapper.mapAssistantMessage looks up at lines 434-444.
As a result, encrypted reasoning replay is unreachable through a normal ReActAgent call even without ResponsesMultiAgentFormatter. This is particularly important for stateless reasoning/tool-call loops, where the reasoning item from the prior response must be sent back with the function-call output. The current cross-turn tests miss this because they manually construct an AssistantMessage with openai.reasoning.encrypted_content already present instead of exercising ChatResponse -> ReasoningContext -> Msg -> next request.
Please preserve the terminal response metadata in the final assistant message (or place the replay data on a content block and ensure its accumulator preserves it), and add a ReActAgent-level two-iteration regression test that captures the second ResponseCreateParams and asserts it contains the prior encrypted reasoning item.
|
Preview deployment for your docs. Learn more about Mintlify Previews.
💡 Tip: Enable Automations to automatically generate PRs for you. |
| if (current instanceof OpenAIIoException | ||
| || current instanceof OpenAIRetryableException) { | ||
| return true; | ||
| } | ||
|
|
||
| if (current instanceof IOException) { | ||
| return true; | ||
| } | ||
|
|
||
| if (current instanceof TimeoutException) { | ||
| return true; | ||
| } |
This comment was marked as abuse.
This comment was marked as abuse.
Sorry, something went wrong.
oss-maintainer
left a comment
oss-maintainer
left a comment
There was a problem hiding this comment.
Summary
Introduces a new agentscope-extensions-model-openai-official module implementing the OpenAI Responses API (as distinct from the existing chat-completions extension): SDK-backed client factory, request mapper, response parser, a streaming assembler over the SDK's StreamResponse, an error translator, credential type, ModelProvider SPI registration, an e2e ProviderFactory provider, and docs in both locales. 46 files, ~9.3k diff lines, of which the substantive new module is roughly 1.6k lines of main and the rest tests and docs.
This is a strong, unusually careful module. I specifically checked the areas this repo's review guidance calls out and they hold up:
- Credential handling —
OpenAIOfficialCredential.toString()redacts the key toapiKey=***, and per-requestapiKeyoverrides are explicitly rejected rather than silently honoured, which closes the usual leak-into-logs path. - Streaming reliability —
doFinally(signal -> closeQuietly(streamResponse))covers complete/error/cancel, terminalfailed/error/incompleteevents are translated rather than swallowed, andmaxRetries(0)delegates retry to AgentScope with amoduleRetryOn()predicate rather than double-retrying. - Plugin isolation — SPI registration via
META-INF/servicesplus aggregation/BOM wiring, andCheck Module Syncis green.
The one real defect is a narrow stream-leak path; the rest is a scoping question about core.
Findings
- [Warning]
OpenAIResponsesChatModel.java:154—StreamResponseis not closed ifassemble(...)throws before itsdoFinallyguard is attached (see inline comment for the mechanism and a fix). - [Warning]
ModelContextWindows.java:72— seven new model→context-window entries are added toagentscope-core, but they describe models served by this new extension and the sameOPENAItable is consumed by the existing OpenAI extension. Core changes cascade to harness/distribution/extensions, so an inflated value mis-sizes compaction silently. A citable source per value plus a resolution test would make the table trustworthy. - [Info]
ModelContextWindows.java:87— theglm-5.3entry is unrelated to this feature and would be easier to review and revert as its own change.
Suggestions
The stream cleanup is the only thing I would fix before merge; it is a few lines. If you prefer to keep the provider's model metadata with the provider, a per-extension context-window table that core consults through the SPI would also avoid the core edit entirely — not something to change in this PR, just worth a follow-up issue since every new provider module currently has to touch core.
Tests
Coverage is comprehensive (error translation, credential, SDK client factory, request mapper, response parser, streaming assembler, cross-turn, plus e2e wiring), and CI is green on ubuntu-latest with 8 checks passing. Note that a 9.3k-line diff means I reviewed the new module's main sources and the core/pom edits closely and sampled the tests and docs rather than auditing every assertion — the maintainers should still read the assembler state machine end to end.
Automated review by github-manager-bot
| try { | ||
| StreamResponse<ResponseStreamEvent> streamResponse = | ||
| client.responses().createStreaming(params); | ||
| return ResponsesStreamingAssembler.assemble( |
There was a problem hiding this comment.
[Warning] StreamResponse leak on the assembly failure path.
ResponsesStreamingAssembler.assemble(...) evaluates streamResponse.stream() eagerly as an argument to Flux.fromStream(...), and the doFinally(signal -> closeQuietly(streamResponse)) guard is only attached to the flux after that call returns. If stream() (or anything else in assemble before doFinally) throws, the exception propagates out of assemble, this catch (RuntimeException e) converts it to Flux.error(...), and the SDK stream is never closed — leaking the underlying HTTP connection for the remainder of the client's lifetime.
The window is narrow but it is exactly the path that fires when the upstream is already misbehaving, which is when connection hygiene matters most. Closing in the catch keeps the invariant local and obvious:
StreamResponse<ResponseStreamEvent> streamResponse = null;
try {
streamResponse = client.responses().createStreaming(params);
Flux<ChatResponse> assembled =
ResponsesStreamingAssembler.assemble(streamResponse, modelName, start);
streamResponse = null;
return assembled;
} catch (RuntimeException e) {
if (streamResponse != null) {
closeQuietly(streamResponse);
}
return Flux.error(OpenAIErrorTranslator.translate(e, modelName));
}Alternatively, move the doFinally attach inside assemble onto a local Flux built before anything else can throw.
Summary
This PR adds a new model extension module,
agentscope-extensions-model-openai-official, which integrates OpenAI models through the official OpenAI Java SDK (com.openai:openai-java) against the Responses API. This complements the existingagentscope-extensions-model-openaimodule, which uses a hand-rolled HTTP client against the Chat Completions API.The Responses API is OpenAI's newer interface with built-in support for reasoning models, encrypted reasoning replay across turns, structured outputs, and server-side conversation state. This module brings first-class support for those capabilities into AgentScope.
What's included
架构图
流式时序图
Module:
agentscope-extensions-model-openai-official(14 production classes, ~2,500 LOC)The module is structured around a clean separation of concerns:
OpenAIResponsesChatModel— theChatModelBaseimplementation. Owns the streaming/non-streaming dispatch, retry classification, and builder API. SDK retry is disabled (maxRetries = 0); AgentScope owns all retry logic via a module-levelretryOnpredicate that inspectsx-should-retryheaders and SDK exception types.OpenAIOfficialModelProvider— SPI provider registered viaModelProviderservice loader. Supports theopenai-official:<model>model-id convention and resolvesOPENAI_API_KEY/OPENAI_BASE_URLfrom the environment.ResponsesRequestMapper— maps AgentScopeMsgobjects,ToolSchemadefinitions, andGenerateOptionsto SDKResponseCreateParams. Handles reasoning replay (encrypted content from previous turns), structured output (JSON object / JSON schema), tool-call mapping with strict mode, and fail-fast validation for unsupported fields.ResponsesResponseParser— parses a non-streamingResponseinto a singleChatResponse, assembling content blocks in fixed order (ThinkingBlock -> TextBlock -> ToolUseBlock) with full metadata extraction.ResponsesStreamingAssembler— assembles aFlux<ResponseStreamEvent>into incrementalChatResponsechunks. Routes text deltas, reasoning summary deltas, function-call argument deltas, and terminal events (completed/incomplete/failed/error) with proper resource cleanup viadoFinally.ResponsesMultiAgentFormatter— multi-agent conversation formatter. Groups messages by type (SYSTEM, TOOL_SEQUENCE, AGENT_CONVERSATION, BYPASS), merges agent conversation messages into<history>-tagged user messages, and passes through tool sequences and system messages unchanged. Supports a customizable conversation history prompt.OpenAIErrorTranslator— normalizes SDK exceptions (OpenAIServiceException,OpenAIIoException,OpenAIRetryableException,OpenAIInvalidDataException,TimeoutException) intoOpenAIOfficialModelExceptionwith HTTP status codes preserved.OpenAISdkClientFactory— the single production entry point forOpenAIClientcreation. SetsmaxRetries = 0, injects builder-level additional headers, and fail-fasts on missing API key.OpenAIOfficialCredential— JSON-serializable credential type (openai_official_credential) for use with AgentScope's credential system.OpenAIOfficialConstants— shared constants for metadata namespace keys (openai.*) and theadditionalBodyParamswhitelist.Key capabilities:
low/medium/high/minimal) with encrypted reasoning content automatically replayed across turns viaMsg.metadata— no manual management neededadditionalBodyParams:reasoning.summary,reasoning.context,reasoning.mode,service_tier,prompt_cache_key,prompt_cache_options,max_tool_calls,safety_identifierOpenAIOfficialResponsesProvider)Supporting changes
Several improvements to core and harness modules were needed to support this integration:
ModelContextWindows— added context window sizes for GPT-5.x, GPT-6, and GLM-5.3 models.ModelUtils— timeout exceptions now wrap aTimeoutExceptionas cause, so module-levelretryOnpredicates can detect timeouts in the cause chain.agentscope-allandagentscope-bomupdated to include the new module.Testing
The module includes 13 test files totaling ~5,600 lines, covering:
Not yet supported
The following are intentionally out of scope for this initial PR. They are tracked for follow-up work:
storeparameter — hardcoded tofalseinResponsesRequestMapper. OpenAI's server-side conversation storage is not used; AgentScope manages conversation history client-side. Making this configurable is a straightforward future enhancement.TextBlock,ImageBlock(URL and base64), and image-typeDataBlockare supported.AudioBlockandVideoBlockare not mapped; the request mapper fail-fasts on unsupported block types, and the multi-agent formatter silently skips them. The existing OpenAI and DashScope modules handle audio and video blocks.ProxyConfig— not supported. This is a real gap for enterprise and restricted-network environments. The official SDK'sOpenAIOkHttpClient.Buildersupports proxy configuration, so wiring is feasible.MultiModalToolimplementation. The OpenAI module providesOpenAIMultiModalTool(text-to-image, image-to-text, text-to-audio, audio-to-text) and the DashScope module providesDashScopeMultiModalTool. A corresponding tool for this module is optional but would improve feature parity.Usage
Via explicit builder:
With reasoning and multi-agent formatter: