Pinned Loading

1. alfred-abliterate alfred-abliterate Public

   Residual-stream abliteration toolkit for MoE models (Qwen3.5-397B-A10B) on Apple Silicon. Removes PRC-aligned content policies from local inference. Tested on Mac Studio M3 Ultra 512GB.

   Python

2. alfred-infra alfred-infra Public

   AI-infrastructure hardening kit for multi-machine local-LLM clusters: monitoring (node_exporter, DCGM, Prometheus), Grafana dashboards, cold backups, network-binding audits

   Shell 1

3. alfred-rag alfred-rag Public

   Hybrid RAG stack: dense (LanceDB) + BM25 (Tantivy) + RRF + Qwen3-Reranker-8B. With sentence-transformers fine-tuning recipe for Qwen3-Embedding-8B.

   Python 1

4. blockops-proxy blockops-proxy Public

   HTTP proxy that translates text-format tool calls from local LLMs into OpenAI tool_calls. SSE streaming, concurrency gating, context truncation. Works with MLX/vLLM/llama.cpp backends.

   Python

5. context-bench context-bench Public

   Context-window benchmark for local LLM serving endpoints (MLX, vLLM, llama.cpp). Measures throughput and memory pressure across 4k-128k context sizes.

   Python 3

6. hermes-agent hermes-agent Public

   Forked from NousResearch/hermes-agent

   The agent that grows with you

   Python