Repository navigation
sinapsis-chatbots v0.4.0
·
27 commits
to main
since this release
The Sinapsis-chatbots monorepo now introduces two new packages, sinapsis-anthropic and sinapsis-mem0, designed to enhance your AI integration experience.
sinapsis-anthropic: Claude Model Support
Expand your AI capabilities with support for Anthropic's Claude models.
AnthropicTextGeneration: Streamline text and code generation tasks using Claude models via the Anthropic API.
Customize with parameters like llm_model_name, role, prompt, system_prompt, context_max_len, extended_thinking, web_search, and more.
AnthropicMultiModal: Empower multimodal chat processing with Claude models, offering the same flexibility and customization as AnthropicTextGeneration.
sinapsis-mem0: Advanced Memory Management
Enhance memory handling with Mem0's powerful tools:
Mem0Add: Organize AI interactions into structured memories, supporting both individual facts and full conversation histories.
Mem0Delete: Efficiently manage memories with selective or bulk deletion options.
Mem0Get: Retrieve memories, from individual entries to comprehensive conversation histories.
Mem0Reset: Clear memory storage entirely or within specific scopes (e.g., user, agent, run).
Mem0Search: Dynamically inject relevant memories into prompts, with configurable formatting options.
Common Attributes:
use_managed: Choose between the managed Mem0 API and self-hosted infrastructure.
memory_config: Configure MemoryClient or Memory with custom parameters.
Template-Specific Attributes:
Mem0Add: Use add_kwargs and generic_key for tailored memory addition.
Mem0Delete: Control deletion with delete_all and delete_kwargs.
Mem0Get: Retrieve memories efficiently with get_all and get_kwargs.
Mem0Search: Customize memory injection with enclosure and search_kwargs.
Additionally, we have improved Local Model Loading
Enhance flexibility with the ability to load local models, complementing existing support for HuggingFace Hub models. This update simplifies model deployment and extends capabilities for LLaMA models.