Skip to content

MANIFESTO.md

Latest commit

 

History

History
163 lines (97 loc) · 5.23 KB

MANIFESTO.md

File metadata and controls

163 lines (97 loc) · 5.23 KB

The Local AI Manifesto

Why We Built ACF Local Edition

The cloud had its moment. Now it's time to bring AI home.

The Problem with Cloud AI

Every time you send code to a cloud API, you're making a decision:

  • Your proprietary code travels across the internet
  • Your business logic sits on someone else's servers
  • Your prompts become training data (or might)
  • Your compliance team has another vendor to audit
  • Your internet connection becomes a single point of failure

For hobby projects, this is fine. For production code at serious companies, it's a liability.

The Local AI Revolution

Something changed in 2024.

  • Ollama hit millions of downloads
  • LM Studio became standard on developer machines
  • Qwen, Llama, DeepSeek started matching cloud models
  • Apple Silicon made inference fast on laptops
  • Quantization made 70B models run on consumer hardware

Local AI isn't a compromise anymore. It's a choice.

Our Principles

1. Your Code Never Leaves Your Machine

Zero telemetry. Zero cloud sync. Zero "anonymous usage data."

When you run acf run "Build an API", the prompt goes to your local Ollama instance. The generated code writes to your local filesystem. The git commits stay in your local repo.

We don't know what you're building. We don't want to know.

2. Offline-First by Design

Internet down? Doesn't matter.

ACF Local works on airplanes, in secure facilities, in countries with unreliable connectivity. Your development workflow shouldn't depend on someone else's uptime.

3. AI Agents Watching AI

Here's the uncomfortable truth: LLMs hallucinate. They generate insecure code. They leak secrets into outputs. They ignore your coding standards.

Our solution: specialized agents that monitor the AI.

  • Policy Agent — Enforces your organization's coding standards
  • Secrets Scanner — Catches hardcoded credentials before commit
  • Security Agent — Scans for OWASP vulnerabilities
  • Consistency Checker — Ensures generated code matches your patterns

The AI generates. The agents verify. Humans approve.

4. Organizational Policy as Code

Every company has rules:

  • "Never use eval() in production"
  • "All API endpoints must have rate limiting"
  • "Database queries must use parameterized statements"
  • "No external HTTP calls without timeout"

These rules live in documentation that nobody reads. Or in the heads of senior engineers who review PRs.

ACF makes policy executable:

# .acf/policies/security.yaml
rules:
  - name: no-eval
    pattern: "eval\\("
    severity: critical
    message: "eval() is forbidden in production code"

  - name: require-rate-limit
    pattern: "@router\\.(get|post|put|delete)"
    must_contain: "RateLimiter"
    message: "All endpoints must use rate limiting"

Now your policies run automatically. Every generation. Every developer. Every time.

5. The Marketplace: Shared Expertise

Security scanning, framework templates, code retrieval — these are solved problems. But every team rebuilds them from scratch.

The ACF Marketplace changes this:

  • Buy battle-tested extensions from experts
  • Sell your own extensions and earn 80% of revenue
  • Share free extensions with the community

One team's security scanner becomes everyone's security scanner. Knowledge compounds.

Who This Is For

Enterprise Security Teams

You can't send proprietary code to OpenAI. You can't explain to compliance why your trade secrets are on Anthropic's servers. You need AI code generation that stays inside the firewall.

Regulated Industries

Healthcare. Finance. Defense. Government. If you're subject to HIPAA, SOX, ITAR, or FedRAMP, cloud AI is a compliance nightmare. Local AI is an audit checkbox.

Privacy-Conscious Developers

Maybe you're building a startup and don't want your ideas on someone else's servers. Maybe you just believe that what happens on your machine should stay on your machine.

Teams with Coding Standards

You've spent years developing conventions. Your codebase has patterns. You don't want AI generating code that looks nothing like the rest of your project. Policy agents ensure consistency.

Offline Environments

Secure facilities. Air-gapped networks. Remote locations. Unreliable internet. If you can't always reach the cloud, local AI is the only option.

The Future We're Building

We believe in a future where:

  1. AI is a local utility — Like your text editor, it runs on your machine
  2. Privacy is the default — Not a premium feature
  3. Agents govern agents — AI output is verified before it ships
  4. Expertise is shareable — Through an open marketplace of extensions
  5. Developers own their tools — Open source, extensible, hackable

The cloud will always exist. But it shouldn't be mandatory.

Join Us

Use it: pip install acf

Extend it: Build agents, profiles, RAG kits

Improve it: PRs welcome on GitHub

Spread it: Tell developers who care about privacy

The future of AI-assisted development is local.

ACF Local Edition is free and open source under the MIT license.

Built by AgentCodeFactory — also available as a cloud platform for teams who prefer managed infrastructure.