feat: update generated SDK models and restore agent settings compatibility#705
Merged
GregHolmes merged 6 commits intomainfrom May 6, 2026
Merged
feat: update generated SDK models and restore agent settings compatibility#705GregHolmes merged 6 commits intomainfrom
GregHolmes merged 6 commits intomainfrom
Conversation
GregHolmes
changed the title
Re-applied the four socket-client patches (broad except, optional message
defaults, agent _sanitize_numeric_types, listen/v2 send_configure shim).
Re-applied the legacy alias re-exports to seven package __init__.py files
(top-level, agent/, agent/v1/, agent/v1/{types,requests}/, types/, requests/)
that PR #703 introduced but didn't freeze. Added those inits to .fernignore
and documented them under temporarily-frozen in AGENTS.md so the next regen
cycle protects them via the .bak swap.
…ture
The 2026-05-05 regen renamed three Fern-generated types under
AgentV1SettingsAgentContext (adding a Context infix) and restructured the
model to nest messages under .context.messages.
Adds 5 alias shim files mapping the old names to the new canonical paths,
and patches the regenerated model with a model_validator(mode="before")
that detects the legacy `messages=[...]` kwarg and remaps it into
`context={"messages": ...}` for wire compatibility. Freezes new files in
.fernignore, documents both in AGENTS.md, and adds 4 regression tests
asserting old and new construction shapes produce identical output.
GregHolmes
changed the title
Contributor
There was a problem hiding this comment.
Pull request overview
This PR is an SDK regeneration update (2026-05-05) that refreshes generated models/TypedDicts and WireMock fixtures to match the latest API schema, including notable schema reshaping for the Agent v1 settings context while preserving backwards compatibility in key places.
Changes:
- Regenerated SDK types/requests, adding new fields (e.g., reasoning mode for think providers, additional model UUIDs for TTS metadata, language hints in listen v2 config success, richer project-member fields).
- Restructured Agent v1 settings context to nest message history under
context.messages, with a compatibility validator to accept legacymessages=[...]. - Updated WireMock mappings and tests to reflect schema changes; removed
uv.lockand updatedpoetry.lock.
Reviewed changes
Copilot reviewed 66 out of 68 changed files in this pull request and generated 2 comments.
Show a summary per file
| File | Description |
|---|---|
| wiremock/wiremock-mappings.json | Updates mocked project-members response payload shape. |
| uv.lock | Removes the uv lockfile from the repo. |
| tests/wire/test_manage_v1_projects_keys.py | Drops a WireMock test that targeted an older create-key request alias. |
| tests/custom/test_compat_aliases.py | Adds regression tests for agent-context legacy kwarg migration and renamed alias resolution. |
| src/deepgram/types/think_settings_v1provider.py | Adds reasoning_mode support for OpenAI/Groq think provider settings. |
| src/deepgram/types/open_ai_think_provider.py | Adds reasoning_mode field to the OpenAI think provider model. |
| src/deepgram/types/open_ai_think_provider_reasoning_mode.py | Introduces OpenAI reasoning mode literal/Any type alias. |
| src/deepgram/types/list_project_members_v1response_members_item.py | Adds member scopes, first_name, and last_name fields. |
| src/deepgram/types/groq.py | Adds reasoning_mode field to the Groq provider model. |
| src/deepgram/types/groq_think_provider_reasoning_mode.py | Introduces Groq reasoning mode literal/Any type alias. |
| src/deepgram/types/init.py | Re-exports new types (reasoning mode aliases). |
| src/deepgram/speak/v1/types/speak_v1metadata.py | Clarifies “primary model” wording and adds additional_model_uuids. |
| src/deepgram/speak/v1/requests/speak_v1metadata.py | Adds additional_model_uuids to metadata params and updates docstrings. |
| src/deepgram/requests/think_settings_v1provider.py | Adds reasoning_mode to think provider TypedDict params. |
| src/deepgram/requests/open_ai_think_provider.py | Adds reasoning_mode to OpenAI think provider params. |
| src/deepgram/requests/list_project_members_v1response_members_item.py | Adds member scopes, first_name, last_name params typing. |
| src/deepgram/requests/groq.py | Adds reasoning_mode to Groq params typing. |
| src/deepgram/listen/v2/types/listen_v2configure_success.py | Adds language_hints to configure-success model. |
| src/deepgram/listen/v2/requests/listen_v2configure_success.py | Adds language_hints to configure-success params. |
| src/deepgram/agent/v1/types/function_call_history_message.py | Updates function-call history typing to new nested context message item type. |
| src/deepgram/agent/v1/types/conversation_history_message.py | Updates conversation history typing to new nested context role type. |
| src/deepgram/agent/v1/types/agent_v1settings_audio_output.py | Refines container typing to use a dedicated container type alias. |
| src/deepgram/agent/v1/types/agent_v1settings_audio_output_encoding.py | Expands supported audio output encodings. |
| src/deepgram/agent/v1/types/agent_v1settings_audio_output_container.py | Adds audio output container literal/Any type alias. |
| src/deepgram/agent/v1/types/agent_v1settings_agent.py | Replaces model with a union alias (AgentV1SettingsAgentContext or str). |
| src/deepgram/agent/v1/types/agent_v1settings_agent_context.py | Reshapes agent context (nesting messages under context) and adds legacy migration validator. |
| src/deepgram/agent/v1/types/agent_v1settings_agent_context_think.py | Adds agent-context think union type. |
| src/deepgram/agent/v1/types/agent_v1settings_agent_context_speak.py | Adds agent-context speak union type. |
| src/deepgram/agent/v1/types/agent_v1settings_agent_context_messages_item.py | Keeps legacy name as an alias to the new nested context message item type. |
| src/deepgram/agent/v1/types/agent_v1settings_agent_context_messages_item_function_calls_function_calls_item.py | Keeps legacy name as alias to the new nested function-call item type. |
| src/deepgram/agent/v1/types/agent_v1settings_agent_context_messages_item_content_role.py | Keeps legacy name as alias to the new nested role type. |
| src/deepgram/agent/v1/types/agent_v1settings_agent_context_listen.py | Adds agent-context listen model wrapper. |
| src/deepgram/agent/v1/types/agent_v1settings_agent_context_listen_provider.py | Adds discriminated union for agent-context listen providers (v1/v2). |
| src/deepgram/agent/v1/types/agent_v1settings_agent_context_listen_provider_v2language_hint.py | Adds language hint union type for agent-context v2 listen provider. |
| src/deepgram/agent/v1/types/agent_v1settings_agent_context_listen_provider_v2.py | Adds agent-context v2 listen provider model. |
| src/deepgram/agent/v1/types/agent_v1settings_agent_context_listen_provider_v1.py | Adds agent-context v1 listen provider model. |
| src/deepgram/agent/v1/types/agent_v1settings_agent_context_context.py | Adds nested context model containing messages. |
| src/deepgram/agent/v1/types/agent_v1settings_agent_context_context_messages_item.py | Adds nested union type for context message items. |
| src/deepgram/agent/v1/types/agent_v1settings_agent_context_context_messages_item_function_calls_function_calls_item.py | Adds nested function-call item model for context history. |
| src/deepgram/agent/v1/types/agent_v1settings_agent_context_context_messages_item_content_role.py | Adds nested role type alias for context messages. |
| src/deepgram/agent/v1/types/agent_v1conversation_text.py | Adds detected languages / hinted languages to conversation text turns. |
| src/deepgram/agent/v1/types/init.py | Re-exports newly added agent-context nested types and container type. |
| src/deepgram/agent/v1/requests/function_call_history_message.py | Updates function-call history params typing to new nested type. |
| src/deepgram/agent/v1/requests/conversation_history_message.py | Updates conversation history params role typing to new nested role type. |
| src/deepgram/agent/v1/requests/agent_v1settings_audio_output.py | Refines container param typing to use container type alias. |
| src/deepgram/agent/v1/requests/agent_v1settings_agent.py | Replaces agent params model with a union alias (AgentV1SettingsAgentContextParams or str). |
| src/deepgram/agent/v1/requests/agent_v1settings_agent_context.py | Updates agent-context params to the new context nesting and new listen/think/speak fields. |
| src/deepgram/agent/v1/requests/agent_v1settings_agent_context_think.py | Adds agent-context think params union type. |
| src/deepgram/agent/v1/requests/agent_v1settings_agent_context_speak.py | Adds agent-context speak params union type. |
| src/deepgram/agent/v1/requests/agent_v1settings_agent_context_messages_item.py | Keeps legacy params name as alias to nested context params union. |
| src/deepgram/agent/v1/requests/agent_v1settings_agent_context_messages_item_function_calls_function_calls_item.py | Keeps legacy params name as alias to nested function-call params. |
| src/deepgram/agent/v1/requests/agent_v1settings_agent_context_listen.py | Adds agent-context listen params wrapper. |
| src/deepgram/agent/v1/requests/agent_v1settings_agent_context_listen_provider.py | Adds union for agent-context listen provider params (v1/v2). |
| src/deepgram/agent/v1/requests/agent_v1settings_agent_context_listen_provider_v2language_hint.py | Adds language hint params union type for v2 listen provider. |
| src/deepgram/agent/v1/requests/agent_v1settings_agent_context_listen_provider_v2.py | Adds v2 listen provider params TypedDict. |
| src/deepgram/agent/v1/requests/agent_v1settings_agent_context_listen_provider_v1.py | Adds v1 listen provider params TypedDict. |
| src/deepgram/agent/v1/requests/agent_v1settings_agent_context_context.py | Adds nested context params TypedDict containing messages. |
| src/deepgram/agent/v1/requests/agent_v1settings_agent_context_context_messages_item.py | Adds nested union params for context messages. |
| src/deepgram/agent/v1/requests/agent_v1settings_agent_context_context_messages_item_function_calls_function_calls_item.py | Adds nested function-call params TypedDict for context messages. |
| src/deepgram/agent/v1/requests/agent_v1conversation_text.py | Adds languages_hinted / languages params typing. |
| src/deepgram/agent/v1/requests/init.py | Re-exports newly added agent-context request param types. |
| src/deepgram/agent/v1/init.py | Re-exports newly added agent v1 types/params at package level. |
| src/deepgram/agent/init.py | Re-exports newly added agent v1 types/params at deepgram.agent level. |
| src/deepgram/init.py | Re-exports new top-level types (e.g., reasoning mode aliases). |
| poetry.lock | Bumps types-requests lock entry. |
| AGENTS.md | Updates regen guidance and frozen/temporary-frozen file lists for new compat shims. |
| .fernignore | Freezes additional compat/__init__ files and documents the agent-context legacy migration patch. |
| .fern/metadata.json | Updates Fern regeneration metadata commit reference. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+28
to
+31
| reasoning_mode: typing.Optional[GroqThinkProviderReasoningMode] = pydantic.Field(default=None) | ||
| """ | ||
| OpenAI reasoning_effort | ||
| """ |
Comment on lines
+26
to
+29
| reasoning_mode: typing_extensions.NotRequired[GroqThinkProviderReasoningMode] | ||
| """ | ||
| OpenAI reasoning_effort | ||
| """ |
Member
Comprehensive Interface Review — PR #705Below is each interface that changed shape, what was done for backwards-compat, and the gaps that remain. I've grouped by impact. 🚨 Tier 1 — Breaking, no compat shim1.
|
| Aspect | Before | After |
|---|---|---|
| Shape | class AgentV1SettingsAgent(UncheckedBaseModel) with fields language, context, listen, think, speak, greeting |
AgentV1SettingsAgent = Union[AgentV1SettingsAgentContext, str] (now a type alias) |
Compat attempted: None for the class itself. The fields were migrated down into AgentV1SettingsAgentContext, and the legacy messages= kwarg there gets a model_validator.
Gaps:
AgentV1SettingsAgent(language=..., listen=..., think=..., speak=..., greeting=...)raisesTypeError— it's no longer callable as a constructor.isinstance(x, AgentV1SettingsAgent)no longer works (it's aUnion, not a class).- Pydantic models that referenced it as a field type and accepted a dict in the old shape (
{"listen": ..., "think": ...}) won't auto-coerce to the new nested shape —messages=is the only kwarg the validator rescues; nothing intercepts the other five fields when they arrive at the outerAgentV1SettingsAgentboundary.
Recommendation: Either keep a thin wrapper class AgentV1SettingsAgent that subclasses AgentV1SettingsAgentContext, or extend the model_validator to also accept the legacy flat-on-Agent payloads (listen/think/speak/greeting/language).
2. AgentV1SettingsAgentParams (request TypedDict)
File: src/deepgram/agent/v1/requests/agent_v1settings_agent.py
| Aspect | Before | After |
|---|---|---|
| Shape | class AgentV1SettingsAgentParams(TypedDict) with language, context, listen, think, speak, greeting |
AgentV1SettingsAgentParams = Union[AgentV1SettingsAgentContextParams, str] |
Compat attempted: None at the type level.
Gaps:
AgentV1SettingsAgentParams(listen=..., think=..., ...)style usage is now a static-type error.- A dict literal
{"listen": {...}, "think": {...}, "greeting": "hi"}will fail at runtime because Pydantic now expects either a string or anAgentV1SettingsAgentContextParams(which nests its messages under.context). However, sincelisten/think/speak/greetingwere lifted intoAgentV1SettingsAgentContextParams, those four keys actually do still work — but only because the schema also moved them.languageis also still there. So in practice, runtime payloads constructed as dicts pass through OK; it's strictly a Python-type-checking break and anisinstancebreak.
Recommendation: Note this as a typing-only break in the changelog; runtime is mostly fine.
⚠️ Tier 2 — Schema reshape with partial compat
3. AgentV1SettingsAgentContext (response model)
File: src/deepgram/agent/v1/types/agent_v1settings_agent_context.py
| Aspect | Before | After |
|---|---|---|
| Fields | messages: Optional[List[...MessagesItem]] only |
language, context: AgentV1SettingsAgentContextContext, listen, think, speak, greeting |
Compat attempted: ✅ Excellent — model_validator(mode="before") (Pydantic v2) and root_validator(pre=True) (Pydantic v1) both detect a legacy messages=[...] kwarg with no context= and remap it to context={"messages": [...]}. Test coverage in tests/custom/test_compat_aliases.py::test_legacy_settings_agent_context_messages_kwarg_remaps_to_context. There's also a deliberate test that explicit context= wins over legacy messages= (no clobber).
Gaps:
- The validator only handles dict input. A caller doing
AgentV1SettingsAgentContext.model_validate(some_pydantic_obj)wheresome_pydantic_objhas a.messagesattribute won't be migrated — but this is an edge case. - Frozen model +
messages=no longer being a real attribute meansobj.messagesraisesAttributeErrorpost-construction. Read-side callers must migrate toobj.context.messages. No deprecation property was added.
Recommendation: Consider a @property def messages(self) returning self.context.messages if self.context else None for read-side compat.
4. AgentV1SettingsAgentContextParams (request TypedDict)
File: src/deepgram/agent/v1/requests/agent_v1settings_agent_context.py
| Aspect | Before | After |
|---|---|---|
| Keys | messages: NotRequired[Sequence[...]] |
language, context: AgentV1SettingsAgentContextContextParams, listen, think, speak, greeting |
Compat attempted: None at the TypedDict layer (TypedDicts can't run validators).
Gaps:
- Static type checkers will flag
{"messages": [...]}as unknown key. - At runtime, when this dict is passed into Pydantic for serialization through the parent request type, the
model_validatoronAgentV1SettingsAgentContext(above) will catch and remap it — so the wire payload comes out correct. This is the primary save. But static type tooling (mypy/pyright) won't know that.
Recommendation: Either add messages: NotRequired[Sequence[...]] back to the TypedDict as a deprecated/legacy key, or document that messages is still accepted at runtime.
✅ Tier 3 — Pure renames with full alias shims
5. AgentV1SettingsAgentContextMessagesItem
Before: Union[ConversationHistoryMessage, FunctionCallHistoryMessage]
After: alias to new AgentV1SettingsAgentContextContextMessagesItem (same Union).
Compat: ✅ Type alias preserves identity (is test passes — see test_renamed_settings_context_type_aliases_resolve). __init__.py re-exports both old and new names. Frozen via .fernignore.
Gaps: None functional.
6. AgentV1SettingsAgentContextMessagesItemContentRole
Same pattern: alias to ...ContextMessagesItemContentRole. ✅ Identity-preserving alias.
7. AgentV1SettingsAgentContextMessagesItemFunctionCallsFunctionCallsItem (and *Params)
Same pattern: alias to ...ContextMessagesItemFunctionCallsFunctionCallsItem. ✅ is-identity preserved by both type and request shims.
Gap: A subtle one — ConversationHistoryMessage.role and FunctionCallHistoryMessage.function_calls are now typed against the new *Context* names. Anyone subclassing those models with explicit type annotations against the old name will get a static-typing mismatch (runtime fine due to alias).
🆕 Tier 4 — Type tightening (silent break risk)
8. AgentV1SettingsAudioOutput.container
File: src/deepgram/agent/v1/types/agent_v1settings_audio_output.py
| Before | After |
|---|---|
Optional[str], docstring says defaults to 'none' |
Optional[AgentV1SettingsAudioOutputContainer] = Union[Literal["none","wav","ogg"], Any] |
Compat attempted: The new alias includes typing.Any in the union, so at the type-system level any string is still accepted. Pydantic also won't reject arbitrary strings because of the Any arm.
Gaps:
- Docstring lost the "defaults to 'none'" note — if server behavior actually changed to no default, that's a runtime behavior change worth confirming.
- IDE auto-complete now suggests only
none/wav/ogg, which may mislead users away from values the server still accepts.
9. AgentV1SettingsAudioOutputEncoding
Before: Literal["linear16","mulaw","alaw"] | Any
After: Literal["linear16","mulaw","alaw","mp3","opus","flac","aac"] | Any
Compat: ✅ Pure widening. No break.
➕ Tier 5 — Purely additive (no break)
| Interface | Change |
|---|---|
AgentV1ConversationText / Params |
Adds optional languages_hinted: Optional[List[str]], languages: Optional[List[str]] |
ListenV2ConfigureSuccess / Params |
Adds optional language_hints: Optional[List[str]] |
OpenAiThinkProvider / Params, Groq / Params, ThinkSettingsV1Provider_OpenAi, ThinkSettingsV1Provider_Groq |
Adds optional reasoning_mode: Optional[OpenAiThinkProviderReasoningMode | GroqThinkProviderReasoningMode] (Literal["none","minimal","low","medium","high"] | Any) |
SpeakV1Metadata / Params |
Adds optional additional_model_uuids: Optional[List[str]]; docstrings clarify model_version/model_uuid refer to the primary model |
ListProjectMembersV1ResponseMembersItem / Params |
Adds optional scopes: Optional[List[str]], first_name: Optional[str], last_name: Optional[str] |
All optional with defaults, all safe.
🆕 Tier 6 — Brand-new types (safe, but worth noting)
These are entirely new (no old equivalent) and only break callers if they had hand-rolled a same-named local type:
AgentV1SettingsAgentContextContext(+Params)AgentV1SettingsAgentContextContextMessagesItem(+Params)AgentV1SettingsAgentContextContextMessagesItemContentRoleAgentV1SettingsAgentContextContextMessagesItemFunctionCallsFunctionCallsItem(+Params)AgentV1SettingsAgentContextListen/ListenProvider/ListenProviderV1/ListenProviderV2/ListenProviderV2LanguageHint/ListenProvider_V1/ListenProvider_V2(+Params)AgentV1SettingsAgentContextSpeak(+Params) — alias toUnion[SpeakSettingsV1, List[SpeakSettingsV1]]AgentV1SettingsAgentContextThink(+Params) — alias toUnion[ThinkSettingsV1, List[ThinkSettingsV1]]AgentV1SettingsAudioOutputContainerGroqThinkProviderReasoningMode,OpenAiThinkProviderReasoningMode
All wired into both __init__.py files (frozen via .fernignore).
🧪 Test-coverage observation
A test was deleted without replacement: tests/wire/test_manage_v1_projects_keys.py::test_manage_v1_projects_keys_create_with_old_request_alias. The CreateKeyV1RequestOneParams alias still exists (it's frozen as a compat shim per AGENTS.md), but the wire-level regression test that proved it actually serializes correctly when used as a request body is gone. The unit-level alias resolution test in test_compat_aliases.py covers identity but not the wire round-trip.
Recommendation: Restore that wire test, or add an equivalent to tests/custom/test_compat_aliases.py.
📋 Summary table
| Interface | Compat? | Gap to fix |
|---|---|---|
AgentV1SettingsAgent (class) |
❌ | Add wrapper class or expand validator |
AgentV1SettingsAgentParams |
❌ (type only) | Document or restore TypedDict shape |
AgentV1SettingsAgentContext |
✅ (validator) | No .messages read-side property |
AgentV1SettingsAgentContextParams |
Add legacy messages key for static checking |
|
*MessagesItem* aliases (4 types + 4 params) |
✅✅ | None |
AudioOutput.container |
Confirm default behavior; docstring regression | |
AudioOutputEncoding |
✅ | None (widening) |
| New additive fields (7 types) | ✅ | None |
Wire test for CreateKeyV1RequestOne alias |
❌ | Restore deleted test |
The team's compat strategy is solid where it was applied — the model_validator migration on AgentV1SettingsAgentContext is the highlight. The two real gaps to flag before merging are (1) AgentV1SettingsAgent becoming a non-instantiable union and (2) the deleted wire test for the legacy create-key alias.
Reinstate legacy AgentV1SettingsAgent construction and request typing, keep read-side messages access working, and restore wire coverage for the old CreateKeyV1RequestOneParams alias.
lukeocodes
approved these changes
May 6, 2026
GregHolmes
changed the title
GregHolmes
changed the title
GregHolmes
changed the title
GregHolmes
pushed a commit
that referenced
this pull request
May 6, 2026
🤖 I have created a release *beep* *boop* --- ## [7.1.0](v7.0.0...v7.1.0) (2026-05-06) ### Features * update generated SDK models and restore agent settings compatibility ([#705](#705)) ([0b820c9](0b820c9)) ### Documentation * target Context7 benchmark gaps in Python skills [no-ci] ([#699](#699)) ([a232eb8](a232eb8)) --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please). Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Compatibility
AgentV1SettingsAgent(...)works again as a callable wrapper classAgentV1Settings.agentaccepts the wrapper class,AgentV1SettingsAgentContext(...), or anagent_idstringmessages=[...]input is remapped into the newcontext={"messages": [...]}wire shapecontext=AgentV1SettingsAgentContext(messages=[...])input is normalized to the new nested context modelobj.messagesaccess continues to workTypedDictshapes for agent settings, including deprecatedmessagessupport onAgentV1SettingsAgentContextParams.__init__.pyfiles and added the patched files to.fernignoreso future regens preserve them.CreateKeyV1RequestOneParamsrequest alias.tests/custom/test_compat_aliases.pyfor the legacy constructor shapes, request typing, and read-side compatibility behavior.Generated Changes
context, top-levellisten/think/speak,greeting, new listen-provider variants, and related request/type exports.reasoning_modesupport for OpenAI and Groq.language_hintstoListenV2ConfigureSuccessandadditional_model_uuidstoSpeakV1Metadata.scopes,first_name, andlast_name.Manual Patches
except Exceptionhandlingagent/v1numeric sanitizationlisten/v2rawsend_configure(message: typing.Any)shim and broader response handlingaudio.output.containertyped asstrfor backward compatibility instead of the regenerated enum alias.reasoning_modedocstrings so they no longer reference OpenAI..fernignoreandAGENTS.mdso the regenerated files that still carry manual patches stay protected in the next regen cycle.uv.lockfile.Validation
185 passed, 4 skippedviaPYTHONPATH=src uv run python -m pytest.examples/27-transcription-live-sagemaker.py, which still requires the optionaldeepgram-sagemakerdependency.examples/12-transcription-prerecorded-callback.pystill uses a placeholder callback target and reports the expected callback resolution error.