Memory Spine API v0.4.0

Store a new memory with content, tags, and metadata. This is the primary way to persist information. Memories are automatically vectorized for semantic search.

Search memories using semantic similarity. Returns memories ranked by relevance to the query. Supports filtering by tags and minimum score threshold.

Get a specific memory by its ID. Returns the full memory object including content, metadata, and version history.

Get the most recently created or updated memories. Useful for quick context checks and session awareness.

Update an existing memory's content or metadata. Creates a new version while preserving history.

Delete a memory by ID. This is a soft delete - the memory can be recovered within 30 days.

CRITICAL: Use at the START of every task. Builds an optimized context window for LLM queries by combining relevant memories, pinned context, and recent activity.

Build task-specific context from relevant memories. Similar to llm_context_window but more focused on specific memory retrieval.

Generate a concise summary from multiple memories matching a query. Useful for condensing large amounts of information.

Calculate priority scores for memories based on relevance, recency, and importance. Helps rank memories for context building.

Extract patterns, trends, and insights from memory data. Analyzes memory usage, common topics, and relationships.

Get detailed analytics on memory usage including storage statistics, access patterns, and performance metrics.

Pin critical information for persistent, always-available access. Pinned content is automatically included in context windows.

Retrieve pinned information by key. Returns null if the pin doesn't exist.

Remove a pinned memory. The content is not deleted, just unpinned.

Build a knowledge graph from existing memories. Automatically discovers relationships based on semantic similarity and explicit links.

Query the knowledge graph for relationships. Supports finding neighbors, paths, and clusters.

Create an explicit relationship between two memories. Useful for manual knowledge graph construction.

Perform deep analysis of a codebase. Returns structure, dependencies, patterns, and quality metrics.

Use before ANY code edit. Generates task-specific context from a codebase including relevant files, functions, and dependencies.

Get AI-powered improvement suggestions for a codebase. Identifies refactoring opportunities, bugs, and best practice violations.

Generate new code based on codebase context and task description. Follows existing patterns and conventions.

Find functions, classes, and other symbols in a codebase. Useful for understanding code structure.

Analyze import/dependency relationships in a codebase. Maps which files depend on which.

Begin a new tracked conversation. Maintains message history for context.

Add a message turn to the current conversation. Supports user, assistant, and system roles.

Get the current conversation context including all message history up to token limit.

Persist the current conversation to memory for later retrieval.

Critical for multi-agent workflows. Transfer conversation context to another agent including recent memories and state.

Store multiple memories in a single operation. More efficient than individual stores.

Add or remove tags from multiple memories at once.

Delete multiple memories in a single operation. Requires confirmation.

Execute complex queries using Domain Specific Language. Supports field matching, boolean operators, and ranges.

Advanced search with boolean operators and fuzzy matching. Supports AND, OR, NOT, and proximity search.

Find memories semantically similar to a given memory. Useful for discovery and deduplication.

Get memories in chronological order within a time range.

Group memories into clusters by topic or tags. Useful for organization and analysis.

Find duplicate or near-duplicate memories based on content similarity.

Check if content would be a duplicate before storing. Prevents redundant memories.

Alternative duplicate check using text similarity algorithms.

Automatically generate relevant tags for a memory using AI analysis.

Adjust the importance level of a memory. Affects priority in search results and context building.

Merge multiple related memories into a single consolidated memory while preserving key information.

Merge identified duplicate memories, keeping the best version.

Automatically find and merge all duplicate memories in the system.

Get the complete version history of a memory including all changes and timestamps.

Revert a memory to a previous version. Creates a new version with the old content.

Compress a memory's content to reduce storage while preserving key information.

Archive a memory to cold storage. Archived memories are searchable but not included in default context.

Get comprehensive system statistics including vector count, storage usage, and performance metrics.

Detailed statistics breakdown including per-tag counts, temporal distribution, and access patterns.

Perform a comprehensive health check on the Memory Spine system. Reports issues and recommendations.

Consolidate old memories by merging related content and removing redundancy. Reduces storage and improves search.

Run all maintenance tasks: cleanup, consolidation, deduplication, and optimization.

Manage automatic lifecycle rules for memory retention and archiving.

Run performance benchmarks on memory operations. Useful for optimization and monitoring.

Generate a summary of memory operations in the current session including stores, searches, and modifications.

Register a webhook to receive notifications for memory events.

Export memories matching a query to JSON format for backup or migration.

PERMANENT. Completely remove memories with no recovery option. Use for GDPR compliance or sensitive data.