Releases: griptape-ai/griptape
Releases · griptape-ai/griptape
v0.18.2
v0.18.1
v0.18.0
🆕 New Features
- Added new embedding driver for Amazon Bedrock:
BedrockTitanEmbeddingDriver
. - Added new prompt driver for Amazon Bedrock:
AmazonBedrockPromptDriver
. - Added new Amazon Bedrock prompt models:
BedrockClaudePromptModelDriver
, andBedrockJurassicPromptModelDriver
,BedrockTitanPromptModelDriver
. - Added new Amazon Bedrock tokenizers:
BedrockClaudeTokenizer
,BedrockJurassicTokenizer
, andBedrockTitanTokenizer
. - Added
ignored_exception_types
toExponentialBackoffMixin
to ignore certain failures and not retry. - Added extraction engines for extracting data into JSON objects and CSV rows:
JsonExtractionEngine
andCsvExtractionEngine
. - Added
ExtractionTask
to integrate extraction engines into workflows and pipelines. - Added the
GoogleDriveClient
tool to access and search files in Google Drive. - Added the
GoogleDocsClient
tool to access and edit Google Docs documents. - Added
PgVectorVectorStoreDriver
to store and retrieve embeddings in Postgres.
🔧 Improvements
- Added
InvalidRequestError
toignored_exception_types
in OpenAI drivers. - Updated all relevant tools to use Griptape loaders.
- Abstracted
TextToolMemory
andToolOutputProcessor
activities intoTextMemoryActivitiesMixin
. - Updated the
Calculator
tool to use numexpr to improve security. - Improved base prompts.
- Updated
GoogleCalendarClient
andGoogleGmailClient
to support theowner_email
property. - Fixed OpenAI GPT-4 throttling issues in prompt drivers.
v0.17.1
v0.17.0
🚨 Breaking Changes
- Removed
schema_template_args
in all tools. The tool instance can now be accessed directly via{{ _self }}
in Jinja activity description templates.
🆕 New Features
- Added
FileLoader
to load arbitrary files into artifacts. - Added
DataFrameLoader
to load PandasDataFrame
s into artifacts. - Added task-level
Ruleset
s for more precise LLM steering inside structures. - Added the ability to optionally specify
Rule
s in place of structure and task-levelRuleset
s. This is useful when the user only needs one set of rules per task/structure. - Added the ability to let the LLM load content directly (not just from memory) into
FileManager
andAwsS3Client
tools. - Added the ability to map custom loaders based on loaded file extensions in
FileManager
. - Added
enable_activities()
anddisable_activities()
toActivityMixin
to quickly enable or disable all activities.
🔧 Improvements
- Added
BaseMultiModelPromptDriver
to better abstract driver methods relevant to handling LLM hosting platforms such as Amazon SageMaker and Hugging Face. - Added a default
Structure.embedding_driver
that automatically propagates down to relevant default dependencies. - Added the ability to specify output file encoding in
FileManager
. - Improved action matching in
ActionSubtask
to handle cases where the LLM wraps action output into markdown or prepends it with extra text.
v0.16.2
v0.16.1
v0.16.0
🆕 New Features
- Added
AmazonSageMakerPromptDriver
to easily connect open source SageMaker-hosted LLMs. - Introduced prompt model drivers that abstract input, output, and tokenization for hosted models. In this release we added support for:
- Llama models with
SageMakerLlamaPromptModelDriver
. - Falcon models with
SageMakerFalconPromptModelDriver
.
- Llama models with
- Added `OpenSearchVectorDriver to integrate OpenSearch vector store and querying capabilities.
- Added
AmazonOpenSearchVectorDriver
to support hosted OpenSearch. - Added
TextSummaryTask
for quick text summarization in Griptape structures. - Added
ToolTask
that allows developers to integrate any tools as a task without having to use chain-of-thought. - Added
TextQueryTask
for text querying in Griptape structures. - Added
CsvLoader
for loading and pre-processing CSV files.
🔧 Improvements
- Merged the
core
namespace intoutils
and moved all mixins into a separatemixins
namespace. - Multiple bug fixes.