session-continuity-kit · packages

Manages multi-turn AI conversation state, token budgets, and context compression through a `SessionManager` class. It requires an implementation of `IStorageAdapter` and a `TokenCounter` to persist and track conversation history across different backends.

Provides a DynamoDB storage adapter for the `@reaatech/session-continuity` library, implementing the `IStorageAdapter` interface for session and message persistence. It requires an AWS SDK v3 `DynamoDBDocumentClient` and utilizes a single-table design with pre-configured GSIs for efficient querying.

Provides a Google Cloud Firestore implementation of the `IStorageAdapter` interface for the `@reaatech/session-continuity` library. It exports a `FirestoreAdapter` class that manages session metadata and message subcollections using the `@google-cloud/firestore` client.

Provides an in-memory `IStorageAdapter` implementation for the `@reaatech/session-continuity` library. It exposes a `MemoryAdapter` class that uses a `Map` to store session data with optional TTL expiration, suitable for development and testing environments.

Provides a Redis-backed storage implementation for the `@reaatech/session-continuity` library by exposing a `RedisAdapter` class. It requires a `node-redis` client instance to persist session data, message history, and user indices using native Redis hashes and sorted sets.

Provides classes for calculating exact or heuristic token counts for OpenAI and Anthropic models, implementing the `TokenCounter` interface from `@reaatech/session-continuity`. It includes a factory for automatic model-based selection and requires `@anthropic-ai/tokenizer` as an optional peer dependency for Anthropic support.