Production-grade semantic block composition engine with 2,060+ verified deterministic blocks
Project description
Neurop Forge
2,700+ verified function blocks AI can execute but never modify.
Neurop Forge is an immutable function library for AI systems. AI agents can search, compose, and execute blocks — but they cannot write code, alter logic, or bypass trust rules.
This turns AI from unpredictable authors into controlled operators.
The Problem
AI systems that generate or modify code are:
- Unpredictable
- Impossible to audit
- Difficult to roll back
- Risky in production
The Solution
Neurop Forge enforces immutability at the architecture level:
| What AI Can Do | What AI Cannot Do |
|---|---|
| Search blocks by intent | Write new code |
| Compose block graphs | Modify existing blocks |
| Execute with full trace | Bypass verification |
| Reason over metadata | Skip trust tiers |
Every block is:
- Verified — passes automated validation before admission
- Immutable — hash-locked, cannot be altered
- Traced — full execution lineage on every run
- Scored — trust tier classification (A or B)
Trust Tiers
Blocks are classified into safety tiers for controlled AI composition:
| Tier | Count | What It Means |
|---|---|---|
| A | 2,060 | Deterministic, pure functions, no side effects, no external dependencies |
| B | 680 | Context-dependent, may require specific inputs, explicit opt-in required |
Tier-A blocks are safe for unrestricted AI use.
Tier-B blocks require explicit permission — AI cannot silently use them.
Quick Example
AI executes blocks. AI does not write code.
from neurop_forge import NeuropForge
forge = NeuropForge()
# AI executes a verified block
result = forge.execute_block("reverse_string", {"s": "hello"})
print(result) # {'result': 'olleh', 'success': True}
# AI composes a workflow from verified blocks
workflow = forge.run_workflow("text_normalization", {"text": "Hello World"})
# AI queries the library
blocks = forge.list_verified_blocks(category="string", tier="A", limit=10)
The AI used three blocks. It modified zero lines of code.
Installation
pip install neurop-forge
Neurop Forge is designed to sit under AI agents, not replace them.
CLI
neurop-forge stats # Library overview
neurop-forge execute reverse_string -i '{"s":"hi"}' # Execute a block
neurop-forge list --tier A --limit 10 # List Tier-A blocks
neurop-forge info reverse_string # Block details
neurop-forge workflows # Available workflows
What's Included
2,740 verified blocks across 175 source modules and 30+ categories:
- String manipulation (trim, split, join, replace, case conversion)
- Collection operations (flatten, unique, sort, chunk, filter)
- Data validation (email, URL, phone, patterns, schemas)
- Type conversions (int, float, string, boolean, parsing)
- Date/time utilities (formatting, calculation, comparison)
- Math & statistics (sqrt, median, variance, trigonometry)
- Encoding & hashing (MD5, SHA256, base64, URL encode)
- Security utilities (sanitize, mask, injection detection)
- HTTP/API helpers (status codes, headers, cookies, MIME types)
- Database patterns (SQL building, migrations, transactions)
- And more...
Who This Is For
Good fit:
- Teams deploying AI agents in production
- Enterprises needing auditable AI execution
- Platforms requiring deterministic AI behavior
- Developers building AI-powered automation
Not a fit:
- Projects needing AI to generate arbitrary code
- Research on unconstrained AI capabilities
- Applications where auditability doesn't matter
V2.0.0 Features
- Verification Gate — blocks must pass validation to enter the library
- Tier Classification — automated A/B safety classification
- Deduplication Engine — removes duplicate blocks automatically
- Parameter Standardization — canonical names with alias support
- Full Execution Trace — every run is logged and auditable
v2.0.0 establishes the execution and trust layer. Policy engines and orchestration remain external by design.
License
Dual License:
- Non-Commercial: Free with attribution
- Commercial: Contact for licensing
Copyright (c) 2024-2026 Lourens Wasserman. All rights reserved.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file neurop_forge-2.0.0.tar.gz.
File metadata
- Download URL: neurop_forge-2.0.0.tar.gz
- Upload date:
- Size: 449.5 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.11.9
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
d3b89c189a3f553523ddf9a8706157d2f3fd76ac4e11d5a153e85141188751e2
|
|
| MD5 |
4bc6b7fd30a763010df4b595e9964201
|
|
| BLAKE2b-256 |
0020c95fa24cd65363b53789e9c3d9895909a2b361ae266696657f42814a5a1e
|
File details
Details for the file neurop_forge-2.0.0-py3-none-any.whl.
File metadata
- Download URL: neurop_forge-2.0.0-py3-none-any.whl
- Upload date:
- Size: 571.0 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.11.9
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
f5ab8d478ea189452ed1d21fa4343c9fabe6ee7741eb4249212b62761a8e93a6
|
|
| MD5 |
667e9e207636936637e13c7dfee53ffb
|
|
| BLAKE2b-256 |
fe3605f73ee1c93eaacb124974b6f1542e2114ac1db9486c8600b1b208379688
|