Compression Strategies

memtomem-stm automatically compresses MCP tool responses by content type to save tokens. It provides 10 strategies that reduce response size while preserving the information the agent needs.

10 Compression Strategies

Strategy	Target content	Behavior
truncate	Small text	Length-limited truncation (default fallback)
hybrid	Markdown	Preserve structure + abbreviate non-essential sections
selective	General text	Keep only query-relevant portions
progressive	Large content	Cursor-based sequential delivery (zero information loss)
extract_fields	JSON dictionaries	Extract key fields only
schema_pruning	JSON arrays	Preserve schema + reduce samples
skeleton	API docs	Preserve structural skeleton only
llm_summary	Complex text	LLM-based summarization (Ollama/OpenAI)
auto	All types	Analyze content and auto-select optimal strategy
none	—	Pass through original without compression

Auto-Selection Logic

The auto strategy (default) analyzes content to pick the optimal strategy:

Content type	Selected strategy
JSON dictionary	extract_fields
Large JSON array	schema_pruning
Markdown document	hybrid
API documentation	skeleton
Small text (< threshold)	truncate
Other large text	selective

Query-Aware Budget Allocation

During compression, the agent’s current query is taken into account — relevant sections receive a larger token budget. For example, when compressing API documentation while the agent is asking about “authentication module,” auth-related endpoints get more space.

Zero Information Loss: Progressive Delivery

The progressive strategy delivers large content without any information loss:

1. First response delivers a table of contents (TOC) and the first chunk
2. Agent requests more → cursor-based delivery of subsequent chunks
3. Full content can be inspected sequentially

Fallback Ladder

When the compression ratio guardrail (default 65% preservation) is violated, a 3-tier fallback activates automatically:

progressive → hybrid → truncate

Each tier checks the guardrail — if satisfied, that strategy’s output is used.

Compression Budget Tuning

Agent feedback automatically adjusts per-tool compression budgets:

• Agent reports information loss → Increase preservation ratio for that tool
• Agent reports response too long → Decrease preservation ratio