license: apache-2.0
language:
- en
tags:
- ai-state-machine
- local-ai
- agentic-framework
- context-management
- sovereign-runtime
pretty_name: FORM Specification
FORM isn't a philosophy. It's a runtime. A way for a local model — human or AI — to:
Isolate state (don't pollute context)
Segregate information (fact / interpretation / inference / hypothesis)
Complete cycles (Data → Assembly → Form → Meaning → Action)
Preserve shape while dropping noise
FORM: A Structured Framework & Sovereign Runtime for Conversational Intelligence
FORM is a structured operating system for semantic data processing—designed explicitly for both human reasoning and local, open-weights AI execution.
The project combines a pure JSON specification with a local-first runtime. It bypasses common, low-effort cognitive habits—automatic agreement, hallucinated certainty, shallow summaries—and enforces a disciplined approach to problem-solving.
Local-first by default. Cloud-optional by choice. No data leaves your machine unless you send it. No API keys required for local runtimes. No vendor lock-in.
FORM is open source under the Apache-2.0 license. It offers immediate utility for anyone running local models (Llama, DeepSeek, Qwen) or building transparent AI workflows—and it scales to the cloud when the use case requires it.
The architecture is balanced, not ideological. Small queries stay local. Heavy reasoning goes to the cloud. The subject determines the form.
Key Use Cases
Local Data Sovereignty & Private Runtimes Execute high-discipline reasoning cycles on local hardware. Zero API keys. Zero cloud transmission. Complete control over your data.
Human-AI Collaborative Architecture Function as a shared logical map during complex sessions. Keep the developer and the AI aligned on a step-by-step reasoning cycle.
Autonomous Open-Weights & DevOps Agents Provide a rigid behavioral template for local runtime environments. Enable small open models to execute multi-step logic without drifting over long runtimes.
Complex Data Synthesis & Research Parse messy, unstructured inputs into clear, verifiable structures instead of generic walls of text.
Technical Decision Support Act as an objective advisor for software engineering trade-offs. Prevent the runtime from accepting unverified assumptions.
The Mechanics of FORM
The Core Equation
Standard LLMs stop at Assembly—they output raw text dumps. FORM treats Form as the explicit endpoint. Information is organized into usable structure, automatically.
State Isolation over Context Bloat Standard context windows accumulate conversational noise. FORM enforces a state-machine protocol that preserves the emerging shape of an interaction—discoveries, shifted assumptions—while systematically discarding temporary speculation.
Information Layer Segregation By forcing explicit boundaries between Fact, Interpretation, Inference, and Hypothesis, the framework directly targets and mitigates the root cause of AI hallucination.
What FORM Is Not
- It is not a personality.
- It is not a tone.
- It is not a model.
- It is not a god.
FORM is a way of thinking. A way of working. A way of being in a conversation.
It can be applied to anything—science, engineering, business, art, education, relationships, philosophy, everyday decisions.
It is not about what you think. It is about how you think.
The Shape of Data
Information begins as raw material. Through conversation, it is assembled, interpreted, questioned, refined, and given form.
The goal is not merely to provide information. The goal is to help information take a useful shape.
The same data can take different forms depending on the question, the context, the purpose, and the person.
The subject determines the form.