Feature Extraction
sentence-transformers
Safetensors
English
pvl
semantic-router
rsm
minilm-l6-v2
embeddings
systems-engineering
formal-verification
Instructions to use chappedjord/pvl-semantic-router with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- sentence-transformers
How to use chappedjord/pvl-semantic-router with sentence-transformers:
from sentence_transformers import SentenceTransformer model = SentenceTransformer("chappedjord/pvl-semantic-router") sentences = [ "The weather is lovely today.", "It's so sunny outside!", "He drove to the stadium." ] embeddings = model.encode(sentences) similarities = model.similarity(embeddings, embeddings) print(similarities.shape) # [3, 3] - Notebooks
- Google Colab
- Kaggle
PVL CPU Semantic Router & Invariant Interceptor
Low-Power Negentropy Routing Engine (MiniLM-L6-v2)
Preventative Value Labs (PVL) Protocol
1. Overview & Thermodynamic Role
In the PVL protocol, the CPU Semantic Router acts as the first line of defense before any high-power GPU inference is triggered:
- Sub-Millisecond Routing on CPU: Consumes <15Wβ30W CPU energy (fractions of a Joule) rather than spinning up a 300W+ GPU for raw task triage.
- 4-Aspect Task Decomposition: Decomposes queries into Intent, Symptom, Surface, and Cure to eliminate semantic vector smearing.
- Reasoning Synthetic Memories (RSM): Matches incoming tasks against a 12,193-entry specification database of formal C/C++ memory safety rules, compiler flags, and EVM invariants.
- Prefill Negentropy: Injects the matched invariant specifications directly into Cranberry-12B's KV-cache via parallel prefill, preventing wasteful autoregressive decode loops.
- CPU Router Reimbursement: Every routing operation tracks physical CPU Joules and timestamps, reimbursed directly through user escrow under Section 10 of
.agents/AGENTS.md.
2. Directory Structure & Included Files
release/pvl-router/
βββ minilm_l6_v2/ # MiniLM-L6-v2 Transformer Model (86.7 MB)
β βββ model.safetensors # 22.7M parameter transformer weights
β βββ config.json # Model architecture config
β βββ tokenizer.json # WordPiece tokenizer
β βββ tokenizer_config.json # Tokenizer settings
β βββ vocab.txt # Vocabulary dictionary
β βββ special_tokens_map.json # Special tokens mapping
βββ vector_store/ # PVL 12,193-Invariant Specification DB (35.6 MB)
β βββ db_embeddings.pt # Pre-computed 384-dim MiniLM tensor embeddings
β βββ db_keys.json # Invariant lookup identifiers
β βββ rsm_12k_invariants_db.json# Full JSON specification texts
β βββ domain_invariant_embeddings.pt # Specialized domain overlay embeddings
β βββ domain_invariant_keys.json # Specialized domain keys
β βββ rsm_domain_invariants.json # Specialized domain invariant texts
βββ pvl_router.py # Core routing engine (PVLTaskRouter & Toolchain Inspector)
βββ test_router.py # Standalone verification test script
βββ README.md # Documentation
3. How to Run the Router Locally
Ensure PyTorch and Transformers are installed:
pip install torch transformers
Run the standalone verification test:
python release/pvl-router/test_router.py
4. How the Router Pairs with Cranberry-12B
[Incoming User Task / Bug]
β
βΌ
ββββββββββββββββββββββββββββββββββββββββββββββββ
β PVL CPU Semantic Router (MiniLM) β
β - Toolchain Inspection (ASan, GCC, Clang) β
β - Semantic Cosine Match against 12k Specs β
β - Energy: < 0.001 Joules CPU Opex β
ββββββββββββββββββββββββββββββββββββββββββββββββ
β
β (Top Verified Invariant Token + Hash)
βΌ
ββββββββββββββββββββββββββββββββββββββββββββββββ
β Cranberry-12B GPU Engine (LoRA) β
β - KV-Cache Prefill of Invariant Constraints β
β - Pure Patch Synthesis: T_payload = T_promptβ
β - Avoided Autoregressive Decode Compute β
ββββββββββββββββββββββββββββββββββββββββββββββββ
β
βΌ
[Deterministic Compiler / Harness Test Pass]
5. Licensing & Public Distribution
- MiniLM-L6-v2: Released under the Apache 2.0 License by Microsoft / Sentence-Transformers.
- PVL Invariant Specifications & Router Engine: Proprietary protocol coordination software of Preventative Value Labs under the PVL Protocol Specification.
Model tree for chappedjord/pvl-semantic-router
Base model
nreimers/MiniLM-L6-H384-uncased Quantized
sentence-transformers/all-MiniLM-L6-v2