Skip to content

persist-os/vector-native

Repository files navigation

Vector Native

A symbolic communication protocol that turns natural language into queryable, reusable knowledge.

Core Principle: Structured symbols eliminate ambiguity. Precision is the goal. Token savings are the side effect.


The Problem

Natural language is ambiguous and information disappears into text blobs. Whether it's an instruction, a policy document, or a strategic plan, you can't easily query it, reuse its parts, template it, or compose it with other information.

Vector Native solves this by turning text into structured data.


Quick Example

Before (Natural Language):

"Please analyze the Q4 2024 sales data and generate an executive summary report focusing on revenue trends."

Ambiguous: Which Q4? What format? What depth?

After (Vector Native):

●analyze|dataset:Q4_2024_sales|output:executive_summary|focus:revenue_trends|depth:high

Explicit: Every parameter is clear. No guessing.


What This Enables

1. Reusable Components

●campaign_architecture|channel:LinkedIn|theme:speed_to_value

Apply ●campaign_architecture across different products. Don't rewrite from scratch.

2. Queryable Knowledge

Search a database of structured notes: "show all ●finding with confidence:high"

3. Composable Workflows

Combine a ●targeting_strategy from one workflow with a ●budget_allocation from another.

4. Audit Trails

●update|section:timeline|field:deadline|old:Jan_15|new:Jan_20

Queryable trail of all changes. Not just text diffs.

5. Template-Driven Operations

Template says ●budget_allocation|total:$50K. Change to total:$100K. Everything else stays intact.

Natural Language Vector Native
Text blob Structured operations
One-time use Reusable components
Can't query Database-ready
Can't compose Mix and match
Lost after use Knowledge asset

The Protocol

Four Core Symbols

  • Core Operation/Entity (Do this, This is an entity, This is a policy)
  • |Parameter Separator
  • :Key-value Binding
  • Addition/Combination

Additional Symbols

  • — Sequential flow
  • — Background/secondary
  • — Block/reject

How It Works

English: "Please give this maximum attention and add these values" (10 words, ~20 tokens)

Vector Native: ●⊕ (2 symbols, ~4 tokens)

Why this works: LLMs have learned symbol associations from training data (mathematical notation, programming syntax, configuration files). Vector Native leverages these pre-trained associations for precision.


Use Cases

✅ Agent-to-Agent Communication (A2A)

Multi-agent coordination, internal APIs, background jobs

●execute|agent:analyzer|input:user_data|output:insights|priority:high

✅ Human-to-AI Instructions (H2A)

System prompts, tool configurations, workflow definitions

●assistant|mode:analytical|style:concise|depth:comprehensive

✅ AI-Mediated Documentation (H2H)

Knowledge bases, audit trails, team collaboration

●policy|type:remote_work|eligibility:all_employees|approval:manager

✅ Production Systems

  • Multi-agent systems needing precise coordination
  • Knowledge management systems (legal, medical, research)
  • Business operations with template-driven workflows
  • System integrations requiring composable definitions

❌ Not Suitable For

  • Casual ChatGPT questions
  • Creative writing
  • Emotional support
  • Exploratory conversations
  • User-facing messages

Rule: If you need precision and reusability, use Vector Native. If a human reads it for the first time, use natural language.


Token Savings: A Side Effect, Not The Goal

Early testing (gpt-4o-mini, 5 scenarios):

Variant Compliance Token Reduction
STRICT 80% 88.8%
BALANCED 40% 95.4%
MINIMAL 40% 95.7%

When token savings matter:

  • ✅ Production systems with thousands of agent messages
  • ✅ Multi-agent coordination (compound costs)
  • ✅ Background jobs (1000s of operations/day)
  • ❌ One-off casual queries

Key insight: Token reduction (85-95%) is a side effect of precision. It matters for production cost scaling: $10,000/month vs $1,000/month.

Primary value: Precision, reusability, composability. Token savings are the bonus.


Examples

Agent Task Delegation

Before: "Can you create a presentation about our Q3 results? Include revenue charts, keep it concise."
After:  ●create|type:presentation|topic:Q3_results|include:revenue_charts|style:concise

System Instructions

Before: "You are a helpful assistant. Always provide detailed responses. When analyzing data, be thorough."
After:  ●assistant|mode:helpful|detail:high|reasoning:explicit

Document Updates

Before: "Please update the deadline in the project timeline section from January 15th to January 20th."
After:  ●update|section:timeline|field:deadline|old:Jan_15|new:Jan_20

Workflow Definition

●workflow|id:content_review
⊕step_1|action:draft|owner:writer|deadline:monday
⊕step_2|action:review|owner:editor|deadline:wednesday
⊕step_3|action:publish|owner:admin|deadline:friday

Try It Now

Live translator: Vector-Native Gem

Say anything in natural language. Watch it become structured, reusable data.

Note: The optimal translation depends on your use case. Experiment with your own prompts.

📖 Implementation guides: docs/quickstart.md


Applications

Multi-Agent Systems

Precise coordination with reusable patterns. Agents speak a common structured language.

Knowledge Management

Queryable research notes, legal documents, medical records. Structure makes information findable.

Business Operations

Template-driven project management, policy documentation, workflow definitions.

Domain-Specific

  • Legal: Machine-readable contracts, clause libraries
  • Medical: Structured clinical notes, treatment protocols
  • Research: Queryable experiment logs, hypothesis tracking
  • Engineering: Specification templates, requirement tracking

📖 Full catalog: docs/use-cases.md


Getting Started

1. Read the Spec

Understand the symbols and their meanings: LANGUAGE_SPEC.md

2. Try the Translator

Use the Gem to translate your own prompts and see what works.

3. Implement

Start with simple operations. Build your own symbol library over time.

4. Share

Open an issue with your use case, examples, and learnings.


Research & Development

Vector Native is an open experiment. There's no single "correct" translation - it depends on your domain, use case, and model.

We need your perspective. Every domain has unique patterns. Your experiments help define what this protocol should be.

Ways to Contribute

1. Share Translation Examples

  • Take a verbose prompt from your domain
  • Show your VN translation
  • Explain your choices
  • Share what you learned

2. Test in Your Domain

  • Try VN for your specific use case
  • Run experiments with different models
  • Share results (positive or negative)
  • Document what worked and what didn't

3. Build Variants

  • Create your own interpretation
  • Use different symbols or structures
  • Test with your team/system
  • Share your approach

How to Contribute

Simple: Open a GitHub issue with your examples, results, or ideas.

Code: See CONTRIBUTING.md for technical guidelines.

Discussion: Questions? Open a discussion thread.


What We're Learning

This is open research. We're discovering:

  • Which symbols work best for different operations
  • How much structure is optimal
  • Where VN excels and where it falls short
  • How different models interpret symbols
  • What makes information truly reusable
  • Domain-specific patterns and variations

Your contributions directly shape these answers.


Frequently Asked Questions

Is this just abbreviations?

No. VN leverages pre-trained symbol associations in LLMs. Symbols trigger statistical patterns from training data (math, programming, config files), not just shorter text.

Why not just use JSON?

JSON is verbose and LLMs aren't trained to "think" in JSON. VN uses symbols with strong semantic associations, achieving both precision and token efficiency.

Does this work with all LLMs?

Testing shows it works well with GPT-4, Claude, Gemini. Smaller models may need more explicit system prompts. Your testing helps us understand compatibility.

When should I NOT use VN?

Casual conversations, creative writing, user-facing content, emotional support. VN is for precision and reusability, not warmth.

Can I modify the symbols?

Yes! Build your own variant if your domain needs different symbols. Share your approach so others can learn.

📖 More questions: docs/faq.md


Learn More

📖 Language Spec — Complete symbol definitions
🎯 Use Cases — Domain-specific applications
📈 Token Savings — When efficiency matters
🧠 Why It Works — Training data mechanics
💬 FAQ — Common questions
🚀 Quickstart — Get started fast


Community

  • GitHub Issues: Bug reports, feature requests, examples
  • Discussions: Ideas, questions, research
  • Discord: [Coming soon]

License

MIT License - see LICENSE


Vector Native is fully open source. We're defining this protocol together.

Maintained by PersistOS

About

LLMs speaking their native language: vector operations, not English.

Resources

Contributing

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 5

Languages