Zleap

Enterprise token router and model-access appliance that centralizes model-provider configuration, member permissions, token budgets, policy-based routing and usage analytics. It connects providers including Claude, GPT, Gemini, DeepSeek, GLM and Kimi and uses an aluminum enclosure with a HiSilicon chip, 8-core CPU and 12 GB memory.

Private enterprise AI workspace/Agent OS providing managed employee desktops, multi-Harness orchestration, private application hosting, files and knowledge, internal communication, task scheduling and continued server-side agent execution. It is offered on a NAS configuration for roughly 5–10 simultaneous users and a higher-end appliance for roughly 20–30 users with support for 35B-class local models.

SQL-Retrieval Augmented Generation architecture and open-source knowledge-base application. It ingests documents and web sources, extracts events and entities, performs semantic and relational retrieval using query-time dynamic hyperedges, preserves source traceability, and exposes cited answers to people and Agents.

MIT-licensed Python knowledge-data processing engine/package for SAG, with embedded SQLite and LanceDB defaults, OpenAI-compatible model and embedding endpoints, optional Elasticsearch, MySQL/OceanBase and PostgreSQL/pgvector backends, document parsing, extraction, search and source tracing.