Module auto_context 

Auto-Context — assemble a compact, always-fresh session snapshot.

Built once at session start (or any time the agent asks “what’s my current state?”). Combines:

• timeline[last_session] — top-N chronological stores
• open_loops — top-N unresolved tasks
• emotional_field — current mood centroid + energy
• system_state — block count + depth distribution

The output is plain text formatted for direct inclusion in an LLM prompt. Every MCP client gets the same format because the assembly happens server-side.

Structs

AutoContext

Assembled auto-context, ready to format for an LLM prompt.

CoreMemory

A core memory entry with full text content.

Constants

AUTO_CONTEXT_HEADER

Short header label used to embed auto-context into other tool outputs.

CORE_MEMORIES_MIN_IMP

Importance threshold for core memories (0-10).

CORE_MEMORIES_TOP_N

Max core memories to embed (full text, not just metadata).

CORE_MEMORY_MAX_CHARS

Max chars per core memory.

LOOPS_TOP_N

TIMELINE_TOP_N

How many items to include in the auto-context snapshot.

Functions

build

Build the auto-context from disk + in-memory state.

render

Render the auto-context as a compact text block (good for LLM prompts).

render_compact

Compact version of the auto-context (for embedding in recall output).