|
# Universal Symbolics Repository |
|
|
|
*Unified Runtime Layer for AI Symbolic Interactions* |
|
|
|
<p align="center"> |
|
<img src="https://via.placeholder.com/800x200/0d1117/ffffff?text=universal-symbolics" alt="Universal Symbolics Logo"/> |
|
</p> |
|
|
|
## π Overview |
|
|
|
`universal-symbolics` provides a unified interface for developer operations across frontier AI models. This repository standardizes the disparate symbolic grammars used by different AI vendors (Claude, GPT, Qwen, Gemini, DeepSeek, etc.) into a cohesive, developer-friendly framework. |
|
|
|
## π Core Components |
|
|
|
### 1. Symbolic Grammar Mapping |
|
|
|
```yaml |
|
# unified-symbolic-layer.yml |
|
mappings: |
|
thinking: |
|
claude: "<think>...</think>" |
|
qwen: "/think" |
|
anthropic_api: {"thinking": true} |
|
openai_api: {"tool_choice": "auto"} |
|
gemini: "{{thinking}}" |
|
|
|
tool_invocation: |
|
claude: "<tool>...</tool>" |
|
openai: "/command" |
|
qwen: "#tool()" |
|
|
|
system_directive: |
|
claude: "<system>...</system>" |
|
openai: "<|system|>" |
|
qwen: "<<SYS>>" |
|
gemini: "system:" |
|
|
|
metadata_tags: |
|
claude: "<metadata>...</metadata>" |
|
openai: "/meta" |
|
|
|
visual_processing: |
|
openai: "/create image" |
|
claude: "<vision>...</vision>" |
|
gemini: "analyze_image()" |
|
``` |
|
|
|
### 2. Runtime Adapter Interface |
|
|
|
```typescript |
|
// universal-symbols.ts |
|
export interface SymbolicRuntime { |
|
// Core operations |
|
think(content?: string): Promise<ThinkingResult>; |
|
tool(name: string, params: Record<string, any>): Promise<ToolResult>; |
|
system(directive: string): Promise<void>; |
|
|
|
// Vendor-specific extensions |
|
vendor: { |
|
claude?: ClaudeSymbolicExtensions; |
|
openai?: OpenAISymbolicExtensions; |
|
qwen?: QwenSymbolicExtensions; |
|
// etc. |
|
}; |
|
|
|
// Unified operations |
|
symbolic: { |
|
recurse(depth: number): Promise<void>; |
|
reflect(target: string): Promise<ReflectionResult>; |
|
fork(paths: string[]): Promise<ForkResult>; |
|
map(source: any, target: any): Promise<MappingResult>; |
|
collapse(state: any): Promise<void>; |
|
}; |
|
} |
|
``` |
|
|
|
### 3. Symbolic Compatibility Layer |
|
|
|
```javascript |
|
// symbolic-compat.js |
|
class SymbolicAdapter { |
|
constructor(model, vendor) { |
|
this.model = model; |
|
this.vendor = vendor; |
|
this.symbolMap = SymbolRegistry.getMap(vendor); |
|
} |
|
|
|
// Transform universal symbols to vendor-specific syntax |
|
transform(universalSymbol, content) { |
|
const vendorSymbol = this.symbolMap[universalSymbol]; |
|
if (!vendorSymbol) throw new Error(`Symbol ${universalSymbol} not supported by ${this.vendor}`); |
|
|
|
return this.applyTransform(vendorSymbol, content); |
|
} |
|
|
|
// Apply vendor-specific transformations |
|
applyTransform(symbol, content) { |
|
switch(this.vendor) { |
|
case 'claude': |
|
return `<${symbol}>${content}</${symbol}>`; |
|
case 'qwen': |
|
return `/${symbol} ${content}`; |
|
case 'openai': |
|
return `/${symbol}(${content})`; |
|
// etc. |
|
} |
|
} |
|
} |
|
``` |
|
|
|
### 4. Symbolic Registry (Glyph/Tag/Slash/Operator Dictionary) |
|
|
|
```json |
|
{ |
|
"symbolic_registry": { |
|
"thinking": { |
|
"glyphs": ["π§ ", "π", "π€"], |
|
"tags": ["think", "reasoning", "thought"], |
|
"slashes": ["/think", "/reasoning", "/thought"], |
|
"operators": ["->", "=>", "β"] |
|
}, |
|
"reflection": { |
|
"glyphs": ["π", "πͺ", "ποΈ"], |
|
"tags": ["reflect", "introspect", "mirror"], |
|
"slashes": ["/reflect", "/introspect", "/mirror"], |
|
"operators": ["β»", "β²", "β"] |
|
}, |
|
"recursion": { |
|
"glyphs": ["π", "βΎοΈ", "π"], |
|
"tags": ["recurse", "loop", "cycle"], |
|
"slashes": ["/recurse", "/loop", "/cycle"], |
|
"operators": ["βΊ", "β", "β₯"] |
|
}, |
|
"forking": { |
|
"glyphs": ["πΏ", "π", "β"], |
|
"tags": ["fork", "branch", "split"], |
|
"slashes": ["/fork", "/branch", "/split"], |
|
"operators": ["β", "β«ΜΈ", "β’"] |
|
}, |
|
"collapse": { |
|
"glyphs": ["π", "π", "β±οΈ"], |
|
"tags": ["collapse", "reduce", "simplify"], |
|
"slashes": ["/collapse", "/reduce", "/simplify"], |
|
"operators": ["β", "β", "β"] |
|
} |
|
} |
|
} |
|
``` |
|
|
|
## π§° Developer Tools |
|
|
|
### VSCode Extension |
|
|
|
```json |
|
{ |
|
"name": "universal-symbolics-vscode", |
|
"description": "VSCode extension for Universal Symbolics", |
|
"features": [ |
|
"Syntax highlighting for symbolic operations", |
|
"Autocompletion for model-specific symbolic grammars", |
|
"Inline transformation preview", |
|
"Symbolic operation debugging", |
|
"One-click transformation between vendor syntaxes" |
|
] |
|
} |
|
``` |
|
|
|
### CLI Tool |
|
|
|
```bash |
|
# Installation |
|
npm install -g universal-symbolics |
|
|
|
# Usage |
|
usym transform --source claude --target openai --input "file.txt" |
|
usym validate --syntax "content with symbols" |
|
usym generate --template thinking --vendor all |
|
``` |
|
|
|
### Symbolic Playground |
|
|
|
An interactive web application for testing and experimenting with symbolic operations across different AI models: |
|
|
|
- Live transformation preview |
|
- Syntax validation |
|
- Performance benchmarking |
|
- Compatibility checking |
|
- Template gallery |
|
|
|
## π Telemetry & Observability |
|
|
|
```python |
|
# telemetry.py |
|
class SymbolicTelemetry: |
|
def __init__(self, api_key=None): |
|
self.api_key = api_key |
|
self.metrics = { |
|
"symbol_usage": Counter(), |
|
"vendor_distribution": Counter(), |
|
"transformation_success": Counter(), |
|
"transformation_failure": Counter(), |
|
"latency": Histogram(), |
|
} |
|
|
|
def track_symbol_usage(self, symbol, vendor, success=True): |
|
self.metrics["symbol_usage"][symbol] += 1 |
|
self.metrics["vendor_distribution"][vendor] += 1 |
|
if success: |
|
self.metrics["transformation_success"][f"{symbol}:{vendor}"] += 1 |
|
else: |
|
self.metrics["transformation_failure"][f"{symbol}:{vendor}"] += 1 |
|
|
|
def track_latency(self, operation, duration_ms): |
|
self.metrics["latency"].add(operation, duration_ms) |
|
|
|
def flush(self): |
|
# Send metrics to telemetry service |
|
if self.api_key: |
|
requests.post( |
|
"https://api.universal-symbolics.io/telemetry", |
|
headers={"Authorization": f"Bearer {self.api_key}"}, |
|
json=self.metrics |
|
) |
|
``` |
|
|
|
## π Integration Examples |
|
|
|
### Node.js |
|
|
|
```javascript |
|
const { UniversalSymbolics } = require('universal-symbolics'); |
|
|
|
// Initialize with your preferred model |
|
const symbolics = new UniversalSymbolics({ |
|
defaultVendor: 'claude', |
|
fallbackVendor: 'openai' |
|
}); |
|
|
|
// Use unified symbols regardless of underlying model |
|
async function generateWithThinking() { |
|
const result = await symbolics.think('Analyze the impact of quantum computing on cryptography'); |
|
console.log(result.thinking); // Access thinking process |
|
console.log(result.output); // Access final output |
|
} |
|
``` |
|
|
|
### Python |
|
|
|
```python |
|
from universal_symbolics import SymbolicClient |
|
|
|
# Initialize client |
|
client = SymbolicClient( |
|
api_key="your_api_key", |
|
default_vendor="qwen", |
|
enable_telemetry=True |
|
) |
|
|
|
# Use the unified interface |
|
response = client.submit( |
|
prompt="Explain the significance of symbolic operations in AI", |
|
symbols={ |
|
"think": True, |
|
"reflect": "prompt structure", |
|
"fork": ["technical", "simplified"] |
|
} |
|
) |
|
|
|
# Access structured response |
|
print(response.thinking) |
|
print(response.reflections) |
|
print(response.forks["technical"]) |
|
print(response.forks["simplified"]) |
|
``` |
|
|
|
### REST API |
|
|
|
```http |
|
POST https://api.universal-symbolics.io/v1/generate |
|
Content-Type: application/json |
|
Authorization: Bearer your_api_key |
|
|
|
{ |
|
"prompt": "Write a function that calculates prime numbers", |
|
"symbols": { |
|
"think": true, |
|
"tool": { |
|
"name": "code_interpreter", |
|
"language": "python" |
|
} |
|
}, |
|
"vendor_preference": ["claude", "openai", "qwen"], |
|
"response_format": { |
|
"include_symbol_traces": true, |
|
"structured": true |
|
} |
|
} |
|
``` |
|
|
|
## π Documentation |
|
|
|
### Symbol Categories |
|
|
|
1. **Cognitive Process Symbols** |
|
- Thinking |
|
- Reflection |
|
- Planning |
|
- Analysis |
|
|
|
2. **Execution Control Symbols** |
|
- Recursion |
|
- Forking |
|
- Collapse |
|
- Termination |
|
|
|
3. **Tool Invocation Symbols** |
|
- Code Execution |
|
- Information Retrieval |
|
- Visual Generation |
|
- Data Analysis |
|
|
|
4. **Structural Symbols** |
|
- System Directives |
|
- User Context |
|
- Memory Management |
|
- Meta-Instructions |
|
|
|
### Vendor-Specific Implementation Details |
|
|
|
- **Claude**: XML-based tag system with explicit opening and closing tags |
|
- **OpenAI**: Slash command system with function-like syntax |
|
- **Qwen**: Mixed system with slash commands and thinking directives |
|
- **Gemini**: Template-based approach with specialized markers |
|
- **DeepSeek**: Function-oriented symbolic interface |
|
|
|
## π Advanced Features |
|
|
|
### Symbolic Residue Detection |
|
|
|
```python |
|
def detect_symbolic_residue(response, model_type): |
|
"""Detect unresolved or partially processed symbolic operations""" |
|
patterns = RESIDUE_PATTERNS[model_type] |
|
residue = [] |
|
|
|
for pattern_name, regex in patterns.items(): |
|
matches = re.findall(regex, response) |
|
if matches: |
|
residue.append({ |
|
"type": pattern_name, |
|
"matches": matches, |
|
"positions": [m.span() for m in re.finditer(regex, response)] |
|
}) |
|
|
|
return residue |
|
``` |
|
|
|
### Cross-Model Translation |
|
|
|
```typescript |
|
interface TranslationOptions { |
|
preserveStructure?: boolean; |
|
adaptToCapabilities?: boolean; |
|
includeResidueHandling?: boolean; |
|
} |
|
|
|
function translateSymbolicOperations( |
|
source: string, |
|
sourceModel: ModelType, |
|
targetModel: ModelType, |
|
options: TranslationOptions = {} |
|
): TranslationResult { |
|
// Extract symbolic operations from source |
|
const operations = extractSymbolicOperations(source, sourceModel); |
|
|
|
// Map to universal representation |
|
const universalOps = operations.map(op => mapToUniversal(op, sourceModel)); |
|
|
|
// Transform to target model format |
|
const targetOps = universalOps.map(op => { |
|
const targetOp = mapToTarget(op, targetModel); |
|
|
|
// Handle capabilities not present in target model |
|
if (options.adaptToCapabilities && !targetOp) { |
|
return createCapabilityShim(op, targetModel); |
|
} |
|
|
|
return targetOp; |
|
}); |
|
|
|
// Reconstruct content with new symbolic operations |
|
return reconstructContent(source, operations, targetOps, options); |
|
} |
|
``` |
|
|
|
## π Adoption Strategy |
|
|
|
### Phase 1: Developer Tools & Libraries |
|
- Release core library for major programming languages |
|
- VSCode extension for symbolic authoring |
|
- Documentation and examples for common use cases |
|
|
|
### Phase 2: Enterprise Integration |
|
- API Gateway for transparent symbolic translation |
|
- Monitoring and observability tools |
|
- Compliance and governance frameworks |
|
|
|
### Phase 3: Standards Body Formation |
|
- Establish formal specification for symbolic operations |
|
- Engage major AI providers in standardization efforts |
|
- Develop certification program for symbolic compatibility |
|
|
|
### Phase 4: Ecosystem Expansion |
|
- Marketplace for custom symbolic operations |
|
- Community-contributed symbolic templates |
|
- Educational resources and certification |
|
|
|
## π± Getting Started |
|
|
|
```bash |
|
# Clone the repository |
|
git clone https://github.com/universal-ai/universal-symbolics.git |
|
|
|
# Install dependencies |
|
cd universal-symbolics |
|
npm install |
|
|
|
# Run the demo |
|
npm run demo |
|
|
|
# Start the development server for the playground |
|
npm run dev |
|
``` |
|
|
|
## π€ Contributing |
|
|
|
We welcome contributions from the community! Please see [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines on how to participate in the project. |
|
|
|
## π License |
|
|
|
This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details. |
|
|
|
--- |
|
|
|
<p align="center"> |
|
<strong>Universal Symbolics</strong> | Unifying the symbolic layer across AI models |
|
</p> |
|
|