Common

This is a workaround for a limitation in Protocol Buffers where a map field cannot be directly placed in a oneof. By wrapping the map in this message, it can be used in oneof fields for update strategies, such as replacing vs. merging labels in update RPCs.

Field	Type	Description
`labels`	`...odmem.v1.StringMap.LabelsEntry`	The string-to-string map

StringMap.LabelsEntry

Field	Type	Description
`key`	`string`
`value`	`string`

GoodMemStatus

Represents a non-fatal status or warning message for GoodMem operations.

This is typically used in streaming RPCs to provide progress updates or report partial failures without stopping the stream.

Field	Type	Description
`code`	`goodmem.v1.GoodMemStatusCode`	A programmatic status code for client-side handling
`message`	`string`	A human-readable description of the status
`details`	`....v1.GoodMemStatus.DetailsEntry`	Additional context (e.g., `embedder_id`, `space_id`)

GoodMemStatus.DetailsEntry

Field	Type	Description
`key`	`string`
`value`	`string`

ChunkingConfig

Defines the strategy for splitting a single document into multiple chunks.

This is a top-level configuration that can be set as a default for a Space or specified for an individual Memory, overriding the space's default. Exactly one chunking strategy must be selected.

Field	Type	Description
`none`	`...m.v1.ChunkingConfig.NoChunking`	Do not split the document
`recursive`	`...unkingConfig.RecursiveChunking`	Use the recursive splitting strategy
`sentence`	`...hunkingConfig.SentenceChunking`	Use the sentence-based splitting strategy

ChunkingConfig.RecursiveChunking

A strategy that splits text recursively based on a list of separators.

This method iterates through a list of separators (e.g., "\n\n", "\n", " ", ""), splitting the text at each level until the chunks are smaller than the specified size.

Field	Type	Description
`chunk_size`	`uint32`	The target size for each chunk (e.g., 1024). Must be > 0
`chunk_overlap`	`uint32`	The number of characters or tokens to overlap between chunks
`separators`	`string`	Hierarchical list of separators to split on, in order of preference
`keep_strategy`	`...ngConfig.SeparatorKeepStrategy`	How to handle separators after splitting
`separator_is_regex`	`bool`	Whether the separator strings are to be treated as regular expressions
`length_measurement`	`...unkingConfig.LengthMeasurement`	The method for measuring chunk length

ChunkingConfig.SentenceChunking

A strategy that splits text based on sentence boundaries.

This method is generally more semantically aware than recursive splitting but may be less predictable in chunk size.

Field	Type	Description
`max_chunk_size`	`uint32`	The maximum size of a chunk. Must be > 0
`min_chunk_size`	`uint32`	The minimum size of a chunk before creating a new one
`enable_language_detection`	`bool`	Whether to detect the language for more accurate sentence segmentation
`length_measurement`	`...unkingConfig.LengthMeasurement`	The method for measuring chunk length

ChunkingConfig.NoChunking

A strategy that performs no chunking, treating the document as a single chunk.

Enums

ChunkingConfig.SeparatorKeepStrategy

Defines how to handle separators during recursive chunking.

Name	Value	Description
`SEPARATOR_KEEP_STRATEGY_UNSPECIFIED`	0	The default value. Invalid on writes
`KEEP_NONE`	1	Drop the separator text entirely
`KEEP_START`	2	Prepend the separator to the following chunk
`KEEP_END`	3	Append the separator to the preceding chunk (LangChain default)

ChunkingConfig.LengthMeasurement

Defines the method for measuring the length of a chunk.

Name	Value	Description
`LENGTH_MEASUREMENT_UNSPECIFIED`	0	The default value. Invalid on writes
`CHARACTER_COUNT`	1	Measure length by number of characters (default)
`TOKEN_COUNT`	2	Measure length by number of tokens (requires a server-side tokenizer)
`CUSTOM`	3	Reserved for custom, server-side-only length functions

SortOrder

Specifies the sort order for list operations.

Name	Value	Description
`SORT_ORDER_UNSPECIFIED`	0	The default value. This is invalid on writes and will be rejected.
`ASCENDING`	1	Sorts results in ascending order (e.g., A-Z, 0-9)
`DESCENDING`	2	Sorts results in descending order (e.g., Z-A, 9-0)

ProviderType

Defines the backend provider for an embedder or reranker model.

Note: This enum is shared between embedders and rerankers. A potential risk is that the list of supported providers may diverge. For LLMs, a separate LLMProviderType enum is used to avoid this issue.

Name	Value	Description
`PROVIDER_TYPE_UNSPECIFIED`	0	The default value. This is invalid on writes and will be rejected
`PROVIDER_TYPE_OPENAI`	1	Models provided by OpenAI (e.g., `text-embedding-ada-002`)
`PROVIDER_TYPE_VLLM`	2	Models served via a vLLM-compatible API endpoint
`PROVIDER_TYPE_TEI`	3	Models served via the HuggingFace Text-Embeddings-Inference (TEI) toolkit
`PROVIDER_TYPE_LLAMA_CPP`	4	Models served via a Llama.cpp-compatible API endpoint
`PROVIDER_TYPE_VOYAGE`	5	Models provided by Voyage AI (e.g., `rerank-2.5`)
`PROVIDER_TYPE_COHERE`	6	Models provided by Cohere's platform
`PROVIDER_TYPE_JINA`	7	Models provided by Jina AI or Jina-compatible endpoints

Modality

Defines the type of content a model can process.

Used by embedders, rerankers, and LLMs to declare their capabilities.

Name	Value	Description
`MODALITY_UNSPECIFIED`	0	The default value. This is invalid on writes and will be rejected
`MODALITY_TEXT`	1	Textual content
`MODALITY_IMAGE`	2	Image content
`MODALITY_AUDIO`	3	Audio content
`MODALITY_VIDEO`	4	Video content

GoodMemStatusCode

Defines granular, non-fatal status codes for GoodMem operations.

These codes are sent in the GoodMemStatus message, typically in streaming responses, to provide clients with programmatic insight into the progress or partial failures of a request without terminating the entire operation.

Name	Value	Description
`GOODMEM_STATUS_CODE_UNSPECIFIED`	0	The default value. Should not be used
`INVALID_ARGUMENT`	3	Standard gRPC status codes (mirrored for client independence) Invalid request parameters (mirrors gRPC INVALID_ARGUMENT)
`NOT_FOUND`	5	Requested resource not found (mirrors gRPC NOT_FOUND)
`PERMISSION_DENIED`	7	Access denied to resource (mirrors gRPC PERMISSION_DENIED)
`FAILED_PRECONDITION`	9	Operation requires specific preconditions (mirrors gRPC FAILED_PRECONDITION)
`EMBEDDER_FAILED`	1000	Embedder-related warnings Embedding generation failed for a specific embedder
`EMBEDDER_UNAVAILABLE`	1001	Embedder service is temporarily unavailable
`EMBEDDER_TIMEOUT`	1002	Embedding request timed out
`VECTOR_SEARCH_FAILED`	2000	Vector search warnings Vector search failed for a specific embedder/space
`VECTOR_SEARCH_PARTIAL`	2001	Vector search returned partial results
`VECTOR_SEARCH_TIMEOUT`	2002	Vector search timed out
`SPACE_INACCESSIBLE`	3000	Space access warnings User lacks permission to access a requested space
`SPACE_NOT_FOUND`	3001	Requested space does not exist
`SPACE_NO_EMBEDDERS`	3002	Space has no configured embedders
`CHUNK_NOT_FOUND`	4000	Memory/chunk warnings Referenced chunk was not found in database
`MEMORY_LOAD_FAILED`	4001	Failed to load memory data
`MEMORY_CONTENT_UNAVAILABLE`	4002	Memory content could not be retrieved
`RERANKING_FAILED`	5000	Processing warnings Post-processing reranking failed
`SUMMARIZATION_FAILED`	5001	Summary generation failed
`SUMMARIZATION_TIMEOUT`	5002	Summary generation timed out
`RATE_LIMITED`	6000	System warnings Request was rate limited
`RESOURCE_EXHAUSTED`	6001	System resources exhausted (memory, CPU, etc.)
`CONFIGURATION_ERROR`	6002	System configuration issue detected
`LLM_CAPABILITY_INFERRED`	7000	LLM-related warnings LLM capability was auto-filled by the system.
`FEATURE_DISABLED`	9000	Informational status messages (non-error) Feature disabled due to missing configuration