Gnosil/semantix

semantic agent kernel which make agent efficient and self-evolve

What it solves

Semantix addresses two primary problems for AI coding agents: the loss of context between sessions and the high cost/inefficiency of LLM provider prefix caches. Most agents forget everything once a session ends, and standard provider caches often fail (resulting in higher bills) because a single character change at the start of a prompt invalidates the entire cache.

How it works

Semantix provides a "memory kernel" that can be used as a standalone CLI agent (semantix-agent) or attached to existing agents (like Claude Code or LangChain apps) via tools, gateways, or skills. It works through several mechanisms:

  • Cross-Session Memory: It extracts reusable "slices" (task patterns, project knowledge, tool sequences) from finished sessions into a local library and reinjects them when similar tasks appear in new sessions.
  • Prefix Hygiene: It strips per-request attribution markers and canonicalizes tool arrays to ensure prompts are byte-identical, maximizing the hit rate of provider-side caches.
  • Byte-Stable Injection: Retrieved memory slices are ordered by ID rather than score to maintain consistent prompt prefixes.
  • L3 Result Reuse: Verified results can be returned directly without making a new model call.
  • Adaptive Scheduling: It learns tool-usage patterns to parallelize calls and prefetch read-only context.

Who it’s for

Developers using AI coding agents who want to reduce API costs, improve response speed, and ensure the agent retains project-specific knowledge across different sessions.

Highlights

  • High Cache Hit Rates: Demonstrates up to 99.8% hit rates on certain providers (e.g., DeepSeek).
  • Cost Reduction: Reported up to 79.8% cost savings in synthetic replay comparisons.
  • Flexible Integration: Can be integrated as a gateway (no code change), a tool registration, or a specific agent skill.
  • Provider Awareness: Includes a vendor capability table and per-vendor cache lifetime tracking to optimize breakpoint placement.

Related

  • Project
    24kchengYe/MemoMind

    MemoMind is a locally hosted, GPU‑accelerated memory system for AI coding agents. It stores extracted facts from chats, documents, and daily life events in PostgreSQL + pgvector, builds a knowledge graph, and offers fast 4‑way hybrid retrieval (semantic, BM25, graph, temporal). The agent can retain new information, recall relevant memories, and reflect across the whole store. All data stays on the user’s machine, works with any OpenAI‑compatible LLM, and includes a web dashboard for browsing and exporting memories.

  • Project
    rtk-ai/icm

    ICM is a Rust‑based, SQLite‑backed memory system for AI coding agents. It stores and recalls project‑specific facts without needing LLM calls, supports keyword or embedding‑based search, integrates with many coding assistants, and achieves benchmark accuracy comparable to larger services.

  • Project
    AVIDS2/memorix

    Memorix is a Node‑JS CLI/library that adds a local, SQLite‑backed, searchable memory store to any Git project. It lets AI coding assistants (Claude Code, Copilot, Cursor, etc.) share project facts, code snapshots, design rationales and commit‑derived knowledge across sessions and tools via the Model‑Context‑Protocol. Features include autopilot worksets, observation/reasoning memory, Git‑memory ingestion, code‑graph indexing, orchestration of multi‑agent tasks, a web dashboard, and optional sync. Install with npm, run `memorix setup --agent <agent>` and the agents can start reusing the shared memory.

  • Project
    rohitg00/agentmemory

    A persistent memory layer for AI coding agents that captures and compresses session data to eliminate the need for re-explaining project context across sessions and tools.

  • Project
    mcncarl/agent-memory-vault

    Agent Memory Vault is a Python toolkit that provides a local, Git‑backed Markdown vault with SQLite (and optional vector) search, enabling Claude Code, Codex and similar AI agents to share durable, auditable long‑term memory across sessions.