HedgehogMemoryv0.1.1

A support agent that must remember every past ticket, resolution, and customer preference across months of interactions. HedgehogMemory keeps the full history in a compressed radial structure, loading all L0 one-liners at session start so the agent instantly knows the customer's relationship and past issues.

A developer assistant that persists project decisions, architectural tradeoffs, and debugging sessions across a codebase's lifetime. When a similar bug resurfaces weeks later, the agent queries its memory, drills from L1 to L4, and recovers the verbatim original context instead of guessing.

A medical intake and follow-up agent that must retain patient visit summaries, medication changes, and lab notes over long treatment periods without losing detail. The never-delete guarantee ensures the verbatim clinical note is always recoverable at L4 for audit or referral, while L0 summaries keep the daily context light.

A sales copilot that tracks every call, objection, and stakeholder note for long-cycle deals spanning quarters. The radial compression keeps thousands of interactions navigable, and a rep can pull up the exact original transcript of an early discovery call months into the deal.

A researcher's agent that accumulates summaries, citations, and notes across a multi-year thesis project. Keyword-based radial navigation surfaces relevant past reading by query, while the L0 origin overview ensures the agent never forgets which papers it has already processed.

The library stays free and self-hostable for individual developers, while teams pay for a hosted, backed-up 'origin.json' service with multi-agent sync, access controls, and observability dashboards. This mirrors the classic open-core playbook, converting hobbyist traction into enterprise revenue.

Since summarization quality depends on an LLM backend, offer a metered API that performs the 5-level radial compression on behalf of the agent. Developers point their ContextWindowManager at the hosted summarizer endpoint instead of managing their own OpenAI keys and rate limits.

Package HedgehogMemory into domain-specific memory products (clinical, legal, sales) with pre-built schemas, compliance logging, and integrations into existing systems like EHRs or CRMs. The generic library becomes a moat-fed platform sold as a solution, not infrastructure.