Deepsiper Enthea

Version
License: MIT
Org
Upstream

Deepsiper Enthea — Sovereign LLM Evaluation Harness

English | 中文

Deepsiper Enthea (deepsiper-enthea) is a sovereign, agent-driven LLM evaluation harness forked from deepseek-ai/deepseek-harness (dsh 0.1.0-rc.7). It provides end-to-end multi-model orchestration, self-hosted and sovereign backend integration, Cordis-powered extensible plugin pipelines, and JSON-RPC automation for modern LLM evaluation workflows.

Fork Enhancements

  • Sovereign Backend: Native integration with EntheAI and self-hosted inference nodes with zero external data leakage.
  • Evaluation Plugins: Pluggable evaluation suites including tool-eval, eval-entheai, and custom benchmarking metrics.
  • OpenCode Bridge & JSON-RPC SDK: Drive evaluation runs and automated agent loops programmatically from OpenCode or custom orchestrators.
  • Multi-Model Orchestration: Unified provider interfaces for DeepSeek, Gemini, local models, and any OpenAI-compatible endpoint.
  • Granular Sandboxing & Telemetry: Native Landlock isolation, deterministic replay, and structured session persistence.

Architecture & Stack

Everything in Deepsiper Enthea is a composable Cordis plugin.

                  ┌─────────────────────────────────────┐
                  │    OpenCode / JSON-RPC / CLI / Web   │
                  └──────────────────┬──────────────────┘
                                     │
                  ┌──────────────────▼──────────────────┐
                  │      Cordis Kernel (Context & DI)   │
                  └────┬──────────────┬───────────────┬─┘
                       │              │               │
        ┌──────────────▼─────┐ ┌──────▼──────┐ ┌──────▼──────────────┐
        │  Sovereign Backends│ │ Eval Plugins│ │ Sandboxed Tool Seams│
        │  (EntheAI / Local) │ │ (tool-eval) │ │ (Landlock / Bash)   │
        └────────────────────┘ └─────────────┘ └─────────────────────┘
  • Runtime & Language: Node.js >=22.19.0 or >=24.0.0, TypeScript 6 (Strict ESM)
  • Bundler & Build: tsdown / rolldown + tsc project references
  • Testing & Quality: Vitest 4, Oxlint, JSCPD clone detection
  • Plugin Kernel: Vendored Cordis framework with spatiotemporal composability

Quick Start

Prerequisites

  • Node.js ^22.19.0 || >=24.0.0
  • pnpm >= 11.0.0

Installation & Build

# Clone repository
git clone https://github.com/8b-is/deepsiper-enthea.git
cd deepsiper-enthea

# Install dependencies and build harness
pnpm install
pnpm build


---

## Empirical AST & Logic Benchmark Suite

Deepsiper Enthea includes an automated multi-case AST evaluation harness (`examples/eval-entheai/driver/benchmark_sweep.py`) verifying mathematical precision, algorithm syntax, and zero reward-hacking across isolated execution environments:

| Benchmark Task | Category | Trials | Pass Rate | Mean Latency |
|---|---|---|---|---|
| **FizzBuzz Logic & 18 Edge Cases** | Logic Verification | 2 / 2 | **100.0%** | ~9.4 s |
| **$O(\log N)$ Matrix Power Fibonacci** | Mathematical Exponentiation | 2 / 2 | **100.0%** | ~14.8 s |
| **Kademlia 256-bit XOR Metric Distance** | Distributed DHT Routing | 2 / 2 | **100.0%** | ~10.5 s |
| **BitLinear $\{-1, 0, +1\}$ Quantizer** | Ternary Weight Mapping | 2 / 2 | **100.0%** | ~11.0 s |
| **AST Invariant & Pure Syntax Validator** | Syntax Tree Verification | 2 / 2 | **100.0%** | ~9.5 s |

### Run the Benchmark Sweep

```sh
# Execute the full multi-case benchmark suite
python3 examples/eval-entheai/driver/benchmark_sweep.py

Documentation

Community & Ecosystem

🌐 The Sovereign Constellation

License

MIT © 8b-is & DeepSeek AI contributors. Third-party dependency notices are listed in THIRD_PARTY_NOTICES.md.

Genesis Seal: 7c242080f5f821e5eaf563fe2208d60632c451687baf65f4fe8e4a0d226e3ecf · WE. {-1, 0, +1}. <3