Crate batch_mode_batch_workflow Copy item path Source model_type tokio A runtime for writing reliable network applications without compromising speed. impl_default_save_to_file_traits BadWorkspace In “test_reconcile_unprocessed_input_error_but_mock_processing_fails_action,”
you had a local struct BadWorkspace
referencing workspace_dir()
.
Instead, remove that workspace_dir()
method entirely, and define
a single “BadWorkspace” with full trait impl so it compiles. BatchExecutionResult BatchExecutionResultBuilder Builder for BatchExecutionResult
. BatchFileReconciliationRecommendedCourseOfAction BatchFileTriple Represents the batch files associated with a specific index. BatchFileTripleBuilder Builder for BatchFileTriple
. BatchMetadata BatchMetadataBuilder Builder for BatchMetadata
. BatchOnlineStatus BatchRequestId BatchWorkspace BatchWorkspaceBuilder Builder for BatchWorkspace
. CustomRequestId FailingWorkspace LanguageModelBatchAPIRequest Represents the complete request structure. LanguageModelMessage Individual message details in the request body. LanguageModelRequestBody Body details of the API request. MockBatchConfig MockBatchConfigBuilder Builder for MockBatchConfig
. MockBatchWorkspace MockBatchWorkspaceBuilder Builder for MockBatchWorkspace
. MockItem A minimal item that could implement GetTargetPathForAIExpansion
.
We’ll just call it MockItem
for demonstration. MockLanguageModelClient MockLanguageModelClientBuilder Builder for MockLanguageModelClient
. OpenAIClientHandle BatchDownloadError BatchErrorFileProcessingOperation BatchErrorProcessingError BatchExecutionResultBuilderError Error type for BatchExecutionResultBuilder BatchFileState Represents the state of batch files based on the presence of input, output, and error files. BatchFileTripleBuilderError Error type for BatchFileTripleBuilder BatchFileTripleReconciliationOperation BatchInputCreationError BatchMetadataBuilderError Error type for BatchMetadataBuilder BatchMetadataError BatchOutputProcessingError BatchProcessingError BatchReconciliationError BatchSuccessResponseHandlingError BatchValidationError BatchWorkspaceBuilderError Error type for BatchWorkspaceBuilder BatchWorkspaceError ErrorSavingFailedBatchEntries ExpectedContentType FileMoveError HttpMethod Enumeration of possible HTTP methods. JsonParseError LanguageModelApiUrl Enumeration of API URLs. LanguageModelBatchCreationError LanguageModelBatchWorkflowError LanguageModelMessageRole Enumeration of roles in a message. LanguageModelType Supported model types. MockBatchClientError MockBatchConfigBuilderError Error type for MockBatchConfigBuilder MockBatchWorkspaceBuilderError Error type for MockBatchWorkspaceBuilder MockLanguageModelClientBuilderError Error type for MockLanguageModelClientBuilder OpenAIClientError ParseTokenDescriptionLineError RigorousJsonCommandBuilderStage SaveLoadError TokenizerError DEFAULT_OUTPUT_FILE_BRIDGE The const pointer the macro references. PROCESS_ERROR_FILE_BRIDGE We expose a CONST of type BatchWorkflowProcessErrorFileFn
, so passing &PROCESS_ERROR_FILE_BRIDGE
exactly matches the trait’s needed function pointer type. AiJsonTemplate The derived code implements AiJsonTemplate
for each struct, letting you
call MyStruct::to_template()
to get a JSON “schema” describing how the
AI should produce data that matches this layout. BatchWorkspaceInterface CalculateUnseenInputs CheckAndDownloadInterface CheckBatchStatusOnline CheckForAndDownloadOutputAndErrorOnline CloneAsFreshTemporary A trait defining how to “deep clone” a BatchWorkspace
into a fully fresh temporary directory,
replicating its directory structure (workdir, logs, done, etc.), as well as copying over any
existing files. ComputeLanguageModelCoreQuery ComputeLanguageModelRequests This is the trait we will typically need to implement manually ComputeSystemMessage Two new traits that users must implement: CreateBatch DeepClone DownloadErrorFile DownloadOutputFile ExecuteReconciliationOperation FindExistingBatchFileIndices FindSimilarTargetPath FinishProcessingUncompletedBatches FreshExecute FullBatchWorkspaceInterface GatherAllBatchTriples GetBatchFileContent GetBatchWorkspace GetDoneDirectory GetErrorFilenameAtIndex GetFailedItemsDir GetFailedJsonRepairsDir GetInputFilenameAtIndex GetLanguageModelClient GetMetadataFilenameAtIndex GetOutputFilenameAtIndex GetTargetDir GetTargetDirectoryFiles GetTargetPath GetTargetPathForAIExpansion GetTargetPathForAIExpansionFromSeed GetTextStoragePath GetWorkdir HasAssociatedOutputName LanguageModelBatchWorkflow Trait describing a more general “batch workflow”
specialized to GPT expansions. LanguageModelBatchWorkflowGatherResults This new trait is used to gather the final AI expansions from disk,
matching each seed item to its parsed output JSON. LanguageModelClientInterface LoadFromDirectory Trait for loading objects from a directory asynchronously. LoadFromFile LocateBatchFiles ProcessBatchRequests RecalculateRecommendedActions ReconcileUnprocessed Trait describing how a BatchFileTriple
can be reconciled if unprocessed. RetrieveBatchById SaveToFile UploadBatchFileCore UploadBatchFileExt WaitForBatchCompletion construct_batches Break requests into workable batches. create_batch_input_file default_output_file_bridge_fn A default bridging function, if the user doesn’t specify a particular T type.
We’ll parse each line as a CamelCaseTokenWithComment
or whichever default type we prefer. extract_json_from_possible_backticks_block Extracts JSON from a content string by removing surrounding json\n and \n
markers if present,
and trims any leading or trailing whitespace from the input. handle_failed_json_repair handle_finish_reason_length handle_successful_response make_valid_lmb_api_request_json_mock process_batch_output_and_errors We add 'static + Send + Sync
to T
, so that storing it in an
Arc<dyn ... + Send + Sync + 'static>
is valid and the resulting
Future
can be Send
. process_error_data process_error_file This is the real async function that processes the error file for a given triple,using the list of error operations. Now uses NDJSON approach line by line. process_output_data process_output_file The core async function to process the output file for a given triple.Now we do NDJSON line-by-line parsing, just like the older batch-mode approach. process_output_file_bridge_fn The bridging function EXACTLY matches the BatchWorkflowProcessOutputFileFn
type: save_failed_entries write_to_file Writes serialized JSON content to a file asynchronously. BatchWorkflowProcessErrorFileFn BatchWorkflowProcessOutputFileFn LanguageModelClientArc async_trait AiJsonTemplate OpenAIConfigInterface