- Added a health check to the Jina client to verify server availability.
- Improved error handling and messaging for failed health checks.
- Enhanced client robustness by providing feedback on server status.
- Added a new `create_multi_vector` function to the Jina client
to support creating multi-vector embeddings.
- Added a new `multi_vector_api.v` file containing the
implementation for the multi-vector API.
- Updated the `jina.vsh` example to demonstrate the usage of the
new multi-vector API.
- Added `delete_classifier` function to delete a classifier by ID.
- Added corresponding unit tests for the new function.
- Updated the client example to demonstrate classifier deletion.
- Renamed `jina_client_test.v` to `api_test.v` for better organization.
- Renamed `model_embed.v` to `embeddings_api.v` for better organization.
- Refactored the embedding API to use enums for task and truncate types,
and added error handling for invalid inputs.
- Added a new function to list available classifiers.
- Extended the Jina client with `list_classifiers()` method.
- Added unit tests to verify the new functionality.
- Update `jina.vsh` example to showcase the new classification API
with support for both text and image inputs. This improves
the flexibility and usability of the client.
- Introduce new structs `TextDoc`, `ImageDoc`, `ClassificationInput`,
`ClassificationOutput`, `ClassificationResult`, and `LabelScore`
to represent data structures for classification requests and
responses. This enhances code clarity and maintainability.
- Implement the `classify` function in `jina_client.v` to handle
classification requests with support for text and image inputs,
model selection, and label specification. This adds a crucial
feature to the Jina client.
- Add comprehensive unit tests in `jina_client_test.v` to cover
the new `classify` function's functionality. This ensures the
correctness and robustness of the implemented feature.
- Remove redundant code related to old classification API and data
structures from `model_embed.v`, `model_rank.v`, and
`jina_client.v`. This streamlines the codebase and removes
obsolete elements.
- Added `train` function to the Jina client for training
classifiers.
- Added `ClassificationTrain` struct to define training
parameters.
- Added `TrainingExample` struct to represent training data.
- Added `ClassificationTrainOutput` struct for the training
response.
- Added a new `classification_api.v` module for classifier
training functionalities.
- Added a new `classify` function to the Jina client for
classification tasks (currently commented out).
- Added a new `rerank` function to the Jina client for reranking documents.
- Added a new `RerankParams` struct to define parameters for reranking.
- Added unit tests for the new `rerank` function.
- Updated the example script to demonstrate reranking.
- Improved error handling and added more comprehensive logging.
- Add `type_`, `truncate`, and `late_chunking` parameters to the
`create_embeddings` function for finer control over embedding
generation. This allows users to specify embedding type,
truncation method, and whether to apply late chunking.
- Rename model parameter to `model` for clarity and consistency.
- Improve model enum naming for better readability and API consistency.
- Add unit tests for the `create_embeddings` function to ensure
correct functionality and handle potential errors.
- Added a `create_embeddings` function to the Jina client to
generate embeddings for given input texts.
- Improved the `create_embeddings` function input parameters
for better flexibility and error handling.
- Updated `TextEmbeddingInput` struct to handle optional
parameters for embedding type, truncation type, and late
chunking. This improves the flexibility of the embedding
generation process.
- Fixed compilation issues and ensured the code builds successfully
- Created an example to test the client functionality
- Started implementing additional endpoints
- Adds a new mechanism to synchronize the database efficiently
by serializing updates using binary encoding. This improves
performance and reduces bandwidth usage compared to previous methods.
- Introduces `SyncRecord` struct to represent database updates
for easier handling and serialization.
- Implements `push_updates` to serialize database changes since a
given index, handling both initial and incremental syncs.
- Implements `sync_updates` to apply received serialized updates
to the database, robustly handling errors and deletions.
- Added a diagram explaining the architecture of the OurDB
syncer, clarifying the interaction between the HTTP server,
master, and worker nodes.
- Added a README file providing a comprehensive overview of
the OurDB syncer project, including its architecture,
features, prerequisites, installation instructions, and usage
examples.
- Removed outdated Mycelium_Streamer documentation as it's no
longer relevant to the current project structure.
- Created example scripts for running the database, master,
and worker components, simplifying the setup and execution of
the system.
- Added HTTP client and server documentation, clarifying their
functionalities and interaction with the OurDB system.
