docs(sdk): add comprehensive documentation for all 12 SDKs

Add README.md documentation for: - Main SDK overview with quick start guides - JavaScript/TypeScript SDK - Python SDK - Go SDK - Rust SDK - Java SDK - Kotlin SDK - Swift SDK - Flutter/Dart SDK - C SDK - C++ SDK - C#/.NET SDK - Ruby SDK Each README includes: - Installation instructions - Quick start examples - Tensor operations - Matrix operations (matmul, conv2d, attention) - LLM inference (single and streaming) - Configuration options - Error handling - Type definitions
2026-01-11 18:05:03 +05:30 · 2026-01-11 18:05:03 +05:30 · 162227dc71
commit 162227dc71
parent e2a3b66123
13 changed files with 2820 additions and 0 deletions
--- a/sdk/README.md
+++ b/sdk/README.md
@ -0,0 +1,166 @@
+# Synor Compute SDKs
+
+Access distributed heterogeneous compute resources (CPU, GPU, TPU, NPU, LPU, FPGA, DSP, WebGPU, WASM) at 90% cost reduction compared to traditional cloud.
+
+## Available SDKs
+
+| Language | Package | Status |
+|----------|---------|--------|
+| [JavaScript/TypeScript](./js) | `synor-compute` | Production |
+| [Python](./python) | `synor-compute` | Production |
+| [Go](./go) | `github.com/synor/compute-sdk-go` | Production |
+| [Flutter/Dart](./flutter) | `synor_compute` | Production |
+| [Java](./java) | `io.synor:compute-sdk` | Production |
+| [Kotlin](./kotlin) | `io.synor:compute-sdk-kotlin` | Production |
+| [Swift](./swift) | `SynorCompute` | Production |
+| [Rust](./rust) | `synor-compute` | Production |
+| [C](./c) | `libsynor-compute` | Production |
+| [C++](./cpp) | `synor-compute` | Production |
+| [C#/.NET](./csharp) | `SynorCompute` | Production |
+| [Ruby](./ruby) | `synor_compute` | Production |
+
+## Features
+
+- **Matrix Operations**: MatMul, Conv2D, Pooling, BatchNorm
+- **AI/ML**: Flash Attention, FFT, Inference (LLMs, Vision, Embeddings)
+- **Multi-Precision**: FP64, FP32, FP16, BF16, INT8, INT4
+- **Automatic Routing**: Cost, Speed, Energy, or Balanced optimization
+- **Streaming**: SSE-based streaming for LLM inference
+- **Job Management**: Async job submission with status polling
+
+## Quick Start
+
+### JavaScript/TypeScript
+
+```typescript
+import { SynorCompute } from 'synor-compute';
+
+const client = new SynorCompute('your-api-key');
+
+// Matrix multiplication
+const result = await client.matmul(a, b, {
+  precision: 'fp16',
+  processor: 'gpu'
+});
+
+// LLM inference with streaming
+for await (const chunk of client.inferenceStream('llama-3-70b', prompt)) {
+  process.stdout.write(chunk);
+}
+```
+
+### Python
+
+```python
+from synor_compute import SynorCompute, Tensor
+
+client = SynorCompute('your-api-key')
+
+# Matrix multiplication
+a = Tensor.random((512, 512))
+b = Tensor.random((512, 512))
+result = await client.matmul(a, b, precision='fp16', processor='gpu')
+
+# LLM inference with streaming
+async for chunk in client.inference_stream('llama-3-70b', prompt):
+    print(chunk, end='')
+```
+
+### Go
+
+```go
+import "github.com/synor/compute-sdk-go"
+
+client := synor.NewClient("your-api-key")
+
+// Matrix multiplication
+result, err := client.MatMul(ctx, a, b, synor.WithPrecision(synor.FP16))
+
+// LLM inference
+response, err := client.Inference(ctx, "llama-3-70b", prompt)
+```
+
+### Rust
+
+```rust
+use synor_compute::{SynorCompute, Tensor, Precision, ProcessorType};
+
+let client = SynorCompute::new("your-api-key");
+
+// Matrix multiplication
+let result = client.matmul(&a, &b)
+    .precision(Precision::FP16)
+    .processor(ProcessorType::GPU)
+    .send()
+    .await?;
+
+// LLM inference with streaming
+let mut stream = client.inference_stream("llama-3-70b", prompt).await?;
+while let Some(token) = stream.next().await {
+    print!("{}", token?);
+}
+```
+
+## API Endpoints
+
+All SDKs connect to the Synor Compute API:
+
+- **Production**: `https://api.synor.io/compute/v1`
+- **Local (Docker)**: `http://localhost:17250`
+
+## Processor Types
+
+| Type | Description |
+|------|-------------|
+| `cpu` | General-purpose CPU computation |
+| `gpu` | NVIDIA/AMD GPU acceleration |
+| `tpu` | Google TPU for ML workloads |
+| `npu` | Neural Processing Units |
+| `lpu` | Language Processing Units (Groq) |
+| `fpga` | Field-Programmable Gate Arrays |
+| `dsp` | Digital Signal Processors |
+| `webgpu` | Browser-based GPU |
+| `wasm` | WebAssembly runtime |
+| `auto` | Automatic selection (default) |
+
+## Precision Levels
+
+| Level | Bits | Use Case |
+|-------|------|----------|
+| `fp64` | 64 | Scientific computing |
+| `fp32` | 32 | General purpose (default) |
+| `fp16` | 16 | AI/ML training |
+| `bf16` | 16 | Large language models |
+| `int8` | 8 | Quantized inference |
+| `int4` | 4 | Extreme quantization |
+
+## Balancing Strategies
+
+| Strategy | Priority |
+|----------|----------|
+| `speed` | Minimize latency |
+| `cost` | Minimize cost |
+| `energy` | Minimize carbon footprint |
+| `latency` | Real-time requirements |
+| `balanced` | Optimal tradeoff (default) |
+
+## Local Development with Docker
+
+Deploy the compute infrastructure locally:
+
+```bash
+cd /path/to/Blockchain.cc
+docker-compose -f docker-compose.compute.yml up -d
+```
+
+Services available:
+- **Compute API**: `http://localhost:17250`
+- **CPU Workers**: `http://localhost:17260-17261`
+- **WASM Worker**: `http://localhost:17262`
+- **Spot Market**: `http://localhost:17270`
+- **Redis**: `localhost:17280`
+- **Prometheus**: `http://localhost:17290`
+
+## License
+
+MIT License - see individual SDK packages for details.
--- a/sdk/c/README.md
+++ b/sdk/c/README.md
@ -0,0 +1,258 @@
+# Synor Compute SDK for C
+
+Access distributed heterogeneous compute at 90% cost reduction.
+
+## Installation
+
+### Using CMake
+
+```cmake
+find_package(SynorCompute REQUIRED)
+target_link_libraries(your_app PRIVATE synor_compute)
+```
+
+### Manual Installation
+
+```bash
+git clone https://github.com/synor/compute-sdk-c
+cd compute-sdk-c
+mkdir build && cd build
+cmake ..
+make
+sudo make install
+```
+
+## Quick Start
+
+```c
+#include <synor_compute.h>
+#include <stdio.h>
+
+int main() {
+    // Initialize client
+    synor_client_t* client = synor_client_create("your-api-key");
+    if (!client) {
+        fprintf(stderr, "Failed to create client\n");
+        return 1;
+    }
+
+    // Create tensors
+    size_t shape[] = {512, 512};
+    synor_tensor_t* a = synor_tensor_random(shape, 2, SYNOR_FP32);
+    synor_tensor_t* b = synor_tensor_random(shape, 2, SYNOR_FP32);
+
+    // Matrix multiplication on GPU
+    synor_matmul_options_t opts = {
+        .precision = SYNOR_FP16,
+        .processor = SYNOR_PROCESSOR_GPU,
+        .priority = SYNOR_PRIORITY_NORMAL
+    };
+
+    synor_result_t* result = synor_matmul(client, a, b, &opts);
+    if (result && result->status == SYNOR_STATUS_COMPLETED) {
+        printf("Time: %ldms\n", result->execution_time_ms);
+        printf("Cost: $%.6f\n", result->cost);
+    }
+
+    // Cleanup
+    synor_result_free(result);
+    synor_tensor_free(a);
+    synor_tensor_free(b);
+    synor_client_free(client);
+
+    return 0;
+}
+```
+
+## Tensor Operations
+
+```c
+// Create tensors
+size_t shape[] = {3, 3};
+synor_tensor_t* zeros = synor_tensor_zeros(shape, 2, SYNOR_FP32);
+synor_tensor_t* ones = synor_tensor_ones(shape, 2, SYNOR_FP32);
+synor_tensor_t* random = synor_tensor_random(shape, 2, SYNOR_FP32);
+
+// From data
+float data[] = {1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f};
+size_t data_shape[] = {2, 3};
+synor_tensor_t* tensor = synor_tensor_create(data, 6, data_shape, 2, SYNOR_FP32);
+
+// Get tensor info
+size_t ndim = synor_tensor_ndim(tensor);
+size_t size = synor_tensor_size(tensor);
+const size_t* tensor_shape = synor_tensor_shape(tensor);
+
+// Get data pointer
+const float* tensor_data = synor_tensor_data(tensor);
+```
+
+## Matrix Operations
+
+```c
+// Matrix multiplication
+synor_matmul_options_t matmul_opts = {
+    .precision = SYNOR_FP16,
+    .processor = SYNOR_PROCESSOR_GPU
+};
+synor_result_t* result = synor_matmul(client, a, b, &matmul_opts);
+
+// 2D Convolution
+synor_conv2d_options_t conv_opts = {
+    .stride_h = 1, .stride_w = 1,
+    .padding_h = 1, .padding_w = 1,
+    .precision = SYNOR_FP32
+};
+synor_result_t* conv = synor_conv2d(client, input, kernel, &conv_opts);
+
+// Attention
+synor_attention_options_t attn_opts = {
+    .num_heads = 8,
+    .flash = true,
+    .precision = SYNOR_FP16
+};
+synor_result_t* attn = synor_attention(client, query, key, value, &attn_opts);
+```
+
+## LLM Inference
+
+```c
+// Single response
+synor_inference_options_t opts = {
+    .max_tokens = 512,
+    .temperature = 0.7f,
+    .top_p = 0.9f
+};
+synor_inference_result_t* response = synor_inference(
+    client, "llama-3-70b", "Explain quantum computing", &opts
+);
+if (response) {
+    printf("%s\n", response->result);
+    synor_inference_result_free(response);
+}
+
+// Streaming with callback
+void on_chunk(const char* chunk, void* user_data) {
+    printf("%s", chunk);
+    fflush(stdout);
+}
+
+synor_inference_stream(client, "llama-3-70b", "Write a poem",
+    &opts, on_chunk, NULL);
+```
+
+## Configuration
+
+```c
+synor_config_t config = {
+    .api_key = "your-api-key",
+    .base_url = "https://api.synor.io/compute/v1",
+    .default_processor = SYNOR_PROCESSOR_GPU,
+    .default_precision = SYNOR_FP16,
+    .timeout_secs = 30,
+    .debug = false
+};
+
+synor_client_t* client = synor_client_create_with_config(&config);
+```
+
+## Error Handling
+
+```c
+synor_result_t* result = synor_matmul(client, a, b, &opts);
+
+if (!result) {
+    const char* error = synor_get_last_error();
+    fprintf(stderr, "Error: %s\n", error);
+} else if (result->status == SYNOR_STATUS_FAILED) {
+    fprintf(stderr, "Job failed: %s\n", result->error_message);
+}
+
+// Check for specific errors
+synor_error_t err = synor_get_error_code();
+switch (err) {
+    case SYNOR_ERROR_NETWORK:
+        fprintf(stderr, "Network error\n");
+        break;
+    case SYNOR_ERROR_AUTH:
+        fprintf(stderr, "Authentication failed\n");
+        break;
+    case SYNOR_ERROR_INVALID_ARG:
+        fprintf(stderr, "Invalid argument\n");
+        break;
+}
+```
+
+## Types
+
+```c
+// Processor types
+typedef enum {
+    SYNOR_PROCESSOR_CPU,
+    SYNOR_PROCESSOR_GPU,
+    SYNOR_PROCESSOR_TPU,
+    SYNOR_PROCESSOR_NPU,
+    SYNOR_PROCESSOR_LPU,
+    SYNOR_PROCESSOR_FPGA,
+    SYNOR_PROCESSOR_AUTO
+} synor_processor_t;
+
+// Precision
+typedef enum {
+    SYNOR_FP64,
+    SYNOR_FP32,
+    SYNOR_FP16,
+    SYNOR_BF16,
+    SYNOR_INT8,
+    SYNOR_INT4
+} synor_precision_t;
+
+// Job status
+typedef enum {
+    SYNOR_STATUS_PENDING,
+    SYNOR_STATUS_RUNNING,
+    SYNOR_STATUS_COMPLETED,
+    SYNOR_STATUS_FAILED,
+    SYNOR_STATUS_CANCELLED
+} synor_status_t;
+```
+
+## Memory Management
+
+All Synor objects must be freed:
+
+```c
+synor_tensor_free(tensor);
+synor_result_free(result);
+synor_inference_result_free(response);
+synor_client_free(client);
+```
+
+## Thread Safety
+
+The client is thread-safe. Each thread can share the same client instance.
+
+## Requirements
+
+- C99 or later
+- libcurl for HTTP
+- OpenSSL for TLS
+
+## Building
+
+```bash
+mkdir build && cd build
+cmake .. -DCMAKE_BUILD_TYPE=Release
+make
+```
+
+## Testing
+
+```bash
+cd build
+ctest
+```
+
+## License
+
+MIT
--- a/sdk/cpp/README.md
+++ b/sdk/cpp/README.md
@ -0,0 +1,274 @@
+# Synor Compute SDK for C++
+
+Access distributed heterogeneous compute at 90% cost reduction.
+
+## Installation
+
+### Using CMake
+
+```cmake
+find_package(SynorCompute REQUIRED)
+target_link_libraries(your_app PRIVATE synor::compute)
+```
+
+### vcpkg
+
+```bash
+vcpkg install synor-compute
+```
+
+### Conan
+
+```
+[requires]
+synor-compute/0.1.0
+```
+
+## Quick Start
+
+```cpp
+#include <synor/compute.hpp>
+#include <iostream>
+
+int main() {
+    synor::Client client("your-api-key");
+
+    // Matrix multiplication on GPU
+    auto a = synor::Tensor::random({512, 512});
+    auto b = synor::Tensor::random({512, 512});
+
+    auto result = client.matmul(a, b)
+        .precision(synor::Precision::FP16)
+        .processor(synor::ProcessorType::GPU)
+        .execute();
+
+    if (result.isSuccess()) {
+        std::cout << "Time: " << result.executionTimeMs() << "ms\n";
+        std::cout << "Cost: $" << result.cost() << "\n";
+    }
+
+    return 0;
+}
+```
+
+## Modern C++ Features
+
+### Auto Type Deduction
+
+```cpp
+auto tensor = synor::Tensor::random({10, 10});
+auto result = client.matmul(a, b).execute();
+```
+
+### Structured Bindings (C++17)
+
+```cpp
+auto [success, data, error] = client.matmul(a, b).execute();
+if (success) {
+    std::cout << "Result shape: " << data.shape() << "\n";
+}
+```
+
+### std::optional Results
+
+```cpp
+if (auto time = result.executionTimeMs()) {
+    std::cout << "Execution time: " << *time << "ms\n";
+}
+```
+
+## Tensor Operations
+
+```cpp
+// Create tensors
+auto zeros = synor::Tensor::zeros({3, 3});
+auto ones = synor::Tensor::ones({2, 2});
+auto random = synor::Tensor::random({10, 10});
+auto randn = synor::Tensor::randn({100});
+auto eye = synor::Tensor::eye(3);
+
+// From std::vector
+std::vector<float> data = {1, 2, 3, 4, 5, 6};
+auto tensor = synor::Tensor(data, {2, 3});
+
+// From initializer list
+auto tensor2 = synor::Tensor({1.0f, 2.0f, 3.0f}, {3});
+
+// Operations
+auto reshaped = tensor.reshape({3, 2});
+auto transposed = tensor.transpose();
+
+// Math
+float mean = tensor.mean();
+float sum = tensor.sum();
+float std_dev = tensor.std();
+```
+
+## Builder Pattern API
+
+```cpp
+// Matrix multiplication
+auto result = client.matmul(a, b)
+    .precision(synor::Precision::FP16)
+    .processor(synor::ProcessorType::GPU)
+    .priority(synor::Priority::High)
+    .strategy(synor::Strategy::Speed)
+    .execute();
+
+// 2D Convolution
+auto conv = client.conv2d(input, kernel)
+    .stride(1, 1)
+    .padding(1, 1)
+    .execute();
+
+// Attention
+auto attention = client.attention(query, key, value)
+    .numHeads(8)
+    .flash(true)
+    .execute();
+```
+
+## Async API with std::future
+
+```cpp
+#include <future>
+
+auto future = client.matmul(a, b)
+    .precision(synor::Precision::FP16)
+    .executeAsync();
+
+// Do other work...
+
+auto result = future.get();
+```
+
+## LLM Inference
+
+```cpp
+// Single response
+auto response = client.inference("llama-3-70b", "Explain quantum computing")
+    .maxTokens(512)
+    .temperature(0.7)
+    .execute();
+
+std::cout << response.result().value_or("") << "\n";
+
+// Streaming with callback
+client.inferenceStream("llama-3-70b", "Write a poem",
+    [](std::string_view chunk) {
+        std::cout << chunk << std::flush;
+    });
+```
+
+## Configuration
+
+```cpp
+synor::Config config;
+config.apiKey = "your-api-key";
+config.baseUrl = "https://api.synor.io/compute/v1";
+config.defaultProcessor = synor::ProcessorType::GPU;
+config.defaultPrecision = synor::Precision::FP16;
+config.timeout = std::chrono::seconds(30);
+config.debug = true;
+
+synor::Client client(config);
+```
+
+## Error Handling
+
+```cpp
+try {
+    auto result = client.matmul(a, b).execute();
+} catch (const synor::ApiError& e) {
+    std::cerr << "API Error " << e.statusCode() << ": " << e.what() << "\n";
+} catch (const synor::NetworkError& e) {
+    std::cerr << "Network error: " << e.what() << "\n";
+} catch (const synor::InvalidArgumentError& e) {
+    std::cerr << "Invalid argument: " << e.what() << "\n";
+}
+
+// Or with std::expected (C++23)
+auto result = client.matmul(a, b).tryExecute();
+if (result) {
+    std::cout << "Success!\n";
+} else {
+    std::cerr << "Error: " << result.error().message() << "\n";
+}
+```
+
+## Types
+
+```cpp
+// Processor types
+enum class ProcessorType {
+    CPU, GPU, TPU, NPU, LPU, FPGA, Auto
+};
+
+// Precision
+enum class Precision {
+    FP64, FP32, FP16, BF16, INT8, INT4
+};
+
+// Job status
+enum class JobStatus {
+    Pending, Running, Completed, Failed, Cancelled
+};
+```
+
+## RAII Memory Management
+
+All Synor objects use RAII:
+
+```cpp
+{
+    auto tensor = synor::Tensor::random({100, 100});
+    auto result = client.matmul(tensor, tensor).execute();
+} // Automatic cleanup
+```
+
+## Move Semantics
+
+Efficient moves for large tensors:
+
+```cpp
+auto tensor = synor::Tensor::random({1000, 1000});
+auto moved = std::move(tensor);  // No copy
+```
+
+## Thread Safety
+
+The client is thread-safe. Use shared_ptr for multi-threaded access:
+
+```cpp
+auto client = std::make_shared<synor::Client>("your-api-key");
+
+// Multiple threads can use client safely
+std::thread t1([&client]() { client->matmul(a, b).execute(); });
+std::thread t2([&client]() { client->matmul(c, d).execute(); });
+```
+
+## Requirements
+
+- C++17 or later
+- CMake 3.16+
+- libcurl
+- nlohmann/json
+
+## Building
+
+```bash
+mkdir build && cd build
+cmake .. -DCMAKE_BUILD_TYPE=Release
+cmake --build .
+```
+
+## Testing
+
+```bash
+cd build
+ctest --output-on-failure
+```
+
+## License
+
+MIT
--- a/sdk/csharp/README.md
+++ b/sdk/csharp/README.md
@ -0,0 +1,250 @@
+# Synor Compute SDK for C#/.NET
+
+Access distributed heterogeneous compute at 90% cost reduction.
+
+## Installation
+
+### NuGet
+
+```bash
+dotnet add package SynorCompute
+```
+
+### Package Manager Console
+
+```powershell
+Install-Package SynorCompute
+```
+
+## Quick Start
+
+```csharp
+using SynorCompute;
+
+var client = new SynorClient("your-api-key");
+
+// Matrix multiplication on GPU
+var a = Tensor.Random(512, 512);
+var b = Tensor.Random(512, 512);
+
+var result = await client.MatMulAsync(a, b, new MatMulOptions
+{
+    Precision = Precision.FP16,
+    Processor = ProcessorType.GPU
+});
+
+if (result.IsSuccess)
+{
+    Console.WriteLine($"Time: {result.ExecutionTimeMs}ms");
+    Console.WriteLine($"Cost: ${result.Cost}");
+}
+```
+
+## Tensor Operations
+
+```csharp
+// Create tensors
+var zeros = Tensor.Zeros(3, 3);
+var ones = Tensor.Ones(2, 2);
+var random = Tensor.Random(10, 10);
+var randn = Tensor.Randn(100);
+var eye = Tensor.Eye(3);
+
+// From array
+float[,] data = { { 1, 2, 3 }, { 4, 5, 6 } };
+var tensor = Tensor.FromArray(data);
+
+// From 1D with shape
+var data1d = new float[] { 1, 2, 3, 4, 5, 6 };
+var tensor1d = new Tensor(data1d, new[] { 2, 3 });
+
+// Operations
+var reshaped = tensor.Reshape(3, 2);
+var transposed = tensor.Transpose();
+
+// Math
+var mean = tensor.Mean();
+var sum = tensor.Sum();
+var std = tensor.Std();
+```
+
+## Async/Await API
+
+```csharp
+// Matrix multiplication
+var result = await client.MatMulAsync(a, b, new MatMulOptions
+{
+    Precision = Precision.FP16,
+    Processor = ProcessorType.GPU,
+    Strategy = BalancingStrategy.Speed
+});
+
+// 2D Convolution
+var conv = await client.Conv2DAsync(input, kernel, new Conv2DOptions
+{
+    Stride = (1, 1),
+    Padding = (1, 1)
+});
+
+// Attention
+var attention = await client.AttentionAsync(query, key, value, new AttentionOptions
+{
+    NumHeads = 8,
+    Flash = true
+});
+```
+
+## LLM Inference
+
+```csharp
+// Single response
+var response = await client.InferenceAsync("llama-3-70b", "Explain quantum computing",
+    new InferenceOptions
+    {
+        MaxTokens = 512,
+        Temperature = 0.7f
+    });
+Console.WriteLine(response.Result);
+
+// Streaming with IAsyncEnumerable
+await foreach (var chunk in client.InferenceStreamAsync("llama-3-70b", "Write a poem"))
+{
+    Console.Write(chunk);
+}
+```
+
+## Configuration
+
+```csharp
+var config = new SynorConfig
+{
+    ApiKey = "your-api-key",
+    BaseUrl = "https://api.synor.io/compute/v1",
+    DefaultProcessor = ProcessorType.GPU,
+    DefaultPrecision = Precision.FP16,
+    Timeout = TimeSpan.FromSeconds(30),
+    Debug = true
+};
+
+var client = new SynorClient(config);
+```
+
+## Dependency Injection
+
+```csharp
+// In Startup.cs or Program.cs
+services.AddSynorCompute(options =>
+{
+    options.ApiKey = Configuration["Synor:ApiKey"];
+    options.DefaultProcessor = ProcessorType.GPU;
+});
+
+// In your service
+public class ComputeService
+{
+    private readonly ISynorClient _client;
+
+    public ComputeService(ISynorClient client)
+    {
+        _client = client;
+    }
+
+    public async Task<Tensor> ComputeAsync(Tensor a, Tensor b)
+    {
+        var result = await _client.MatMulAsync(a, b);
+        return result.Data;
+    }
+}
+```
+
+## Error Handling
+
+```csharp
+try
+{
+    var result = await client.MatMulAsync(a, b);
+}
+catch (SynorApiException ex)
+{
+    Console.WriteLine($"API Error {ex.StatusCode}: {ex.Message}");
+}
+catch (SynorNetworkException ex)
+{
+    Console.WriteLine($"Network error: {ex.Message}");
+}
+catch (SynorException ex)
+{
+    Console.WriteLine($"Error: {ex.Message}");
+}
+```
+
+## LINQ Integration
+
+```csharp
+// Process multiple tensors
+var tensors = new[] { a, b, c, d };
+var results = await Task.WhenAll(
+    tensors.Select(t => client.MatMulAsync(t, identity))
+);
+```
+
+## Types
+
+```csharp
+// Processor types
+public enum ProcessorType
+{
+    CPU, GPU, TPU, NPU, LPU, FPGA, Auto
+}
+
+// Precision
+public enum Precision
+{
+    FP64, FP32, FP16, BF16, INT8, INT4
+}
+
+// Job status
+public enum JobStatus
+{
+    Pending, Running, Completed, Failed, Cancelled
+}
+
+// Balancing strategy
+public enum BalancingStrategy
+{
+    Speed, Cost, Energy, Latency, Balanced
+}
+```
+
+## Cancellation Support
+
+```csharp
+var cts = new CancellationTokenSource();
+
+// Cancel after 10 seconds
+cts.CancelAfter(TimeSpan.FromSeconds(10));
+
+try
+{
+    var result = await client.MatMulAsync(a, b, cancellationToken: cts.Token);
+}
+catch (OperationCanceledException)
+{
+    Console.WriteLine("Operation was cancelled");
+}
+```
+
+## Requirements
+
+- .NET 6.0 or later
+- System.Text.Json
+
+## Testing
+
+```bash
+dotnet test
+```
+
+## License
+
+MIT
--- a/sdk/flutter/README.md
+++ b/sdk/flutter/README.md
@ -0,0 +1,249 @@
+# Synor Compute SDK for Flutter/Dart
+
+Access distributed heterogeneous compute at 90% cost reduction.
+
+## Installation
+
+Add to `pubspec.yaml`:
+
+```yaml
+dependencies:
+  synor_compute: ^0.1.0
+```
+
+Then run:
+
+```bash
+flutter pub get
+```
+
+## Quick Start
+
+```dart
+import 'package:synor_compute/synor_compute.dart';
+
+void main() async {
+  final client = SynorCompute('your-api-key');
+
+  // Matrix multiplication on GPU
+  final a = Tensor.random([512, 512]);
+  final b = Tensor.random([512, 512]);
+
+  final result = await client.matmul(a, b,
+    precision: Precision.fp16,
+    processor: ProcessorType.gpu,
+  );
+
+  if (result.isSuccess) {
+    print('Time: ${result.executionTimeMs}ms');
+    print('Cost: \$${result.cost}');
+  }
+}
+```
+
+## Tensor Operations
+
+```dart
+// Create tensors
+final zeros = Tensor.zeros([3, 3]);
+final ones = Tensor.ones([2, 2]);
+final random = Tensor.random([10, 10]);
+final randn = Tensor.randn([100]);
+final eye = Tensor.eye(3);
+
+// From list
+final data = [1.0, 2.0, 3.0, 4.0, 5.0, 6.0];
+final tensor = Tensor(data, shape: [2, 3]);
+
+// From typed data (efficient)
+final float32List = Float32List.fromList(data);
+final tensor2 = Tensor.fromTypedData(float32List, shape: [2, 3]);
+
+// Operations
+final reshaped = tensor.reshape([3, 2]);
+final transposed = tensor.transpose();
+
+// Math
+final mean = tensor.mean();
+final sum = tensor.sum();
+final std = tensor.std();
+```
+
+## Matrix Operations
+
+```dart
+// Matrix multiplication
+final result = await client.matmul(a, b,
+  precision: Precision.fp16,
+  processor: ProcessorType.gpu,
+  strategy: BalancingStrategy.speed,
+);
+
+// 2D Convolution
+final conv = await client.conv2d(input, kernel,
+  stride: (1, 1),
+  padding: (1, 1),
+);
+
+// Attention
+final attention = await client.attention(query, key, value,
+  numHeads: 8,
+  flash: true,
+);
+```
+
+## LLM Inference
+
+```dart
+// Single response
+final response = await client.inference(
+  'llama-3-70b',
+  'Explain quantum computing',
+  maxTokens: 512,
+  temperature: 0.7,
+);
+print(response.result);
+
+// Streaming
+await for (final chunk in client.inferenceStream(
+  'llama-3-70b',
+  'Write a poem',
+)) {
+  stdout.write(chunk);
+}
+```
+
+## Configuration
+
+```dart
+final config = SynorConfig(
+  apiKey: 'your-api-key',
+  baseUrl: 'https://api.synor.io/compute/v1',
+  defaultProcessor: ProcessorType.gpu,
+  defaultPrecision: Precision.fp16,
+  timeout: Duration(seconds: 30),
+  debug: true,
+);
+
+final client = SynorCompute.withConfig(config);
+```
+
+## Flutter Widget Integration
+
+```dart
+import 'package:flutter/material.dart';
+import 'package:synor_compute/synor_compute.dart';
+
+class ComputeWidget extends StatefulWidget {
+  @override
+  State<ComputeWidget> createState() => _ComputeWidgetState();
+}
+
+class _ComputeWidgetState extends State<ComputeWidget> {
+  final client = SynorCompute('your-api-key');
+  String? result;
+  bool isLoading = false;
+
+  Future<void> compute() async {
+    setState(() => isLoading = true);
+    try {
+      final response = await client.inference(
+        'llama-3-70b',
+        'Hello',
+      );
+      setState(() => result = response.result);
+    } catch (e) {
+      setState(() => result = 'Error: $e');
+    } finally {
+      setState(() => isLoading = false);
+    }
+  }
+
+  @override
+  Widget build(BuildContext context) {
+    return Column(
+      children: [
+        if (isLoading)
+          CircularProgressIndicator()
+        else if (result != null)
+          Text(result!),
+        ElevatedButton(
+          onPressed: compute,
+          child: Text('Compute'),
+        ),
+      ],
+    );
+  }
+}
+```
+
+## Riverpod Integration
+
+```dart
+import 'package:flutter_riverpod/flutter_riverpod.dart';
+
+final synorProvider = Provider((ref) => SynorCompute('your-api-key'));
+
+final inferenceProvider = FutureProvider.family<String, String>((ref, prompt) async {
+  final client = ref.watch(synorProvider);
+  final result = await client.inference('llama-3-70b', prompt);
+  return result.result ?? '';
+});
+```
+
+## Error Handling
+
+```dart
+try {
+  final result = await client.matmul(a, b);
+} on SynorException catch (e) {
+  print('API Error: ${e.message} (${e.statusCode})');
+} catch (e) {
+  print('Unexpected error: $e');
+}
+```
+
+## Types
+
+```dart
+// Processor types
+enum ProcessorType {
+  cpu, gpu, tpu, npu, lpu, fpga, auto
+}
+
+// Precision
+enum Precision {
+  fp64, fp32, fp16, bf16, int8, int4
+}
+
+// Job status
+enum JobStatus {
+  pending, running, completed, failed, cancelled
+}
+
+// Balancing strategy
+enum BalancingStrategy {
+  speed, cost, energy, latency, balanced
+}
+```
+
+## Platform Support
+
+| Platform | Status |
+|----------|--------|
+| Android | Supported |
+| iOS | Supported |
+| Web | Supported |
+| macOS | Supported |
+| Windows | Supported |
+| Linux | Supported |
+
+## Testing
+
+```bash
+flutter test
+```
+
+## License
+
+MIT
--- a/sdk/go/README.md
+++ b/sdk/go/README.md
@ -0,0 +1,174 @@
+# Synor Compute SDK for Go
+
+Access distributed heterogeneous compute at 90% cost reduction.
+
+## Installation
+
+```bash
+go get github.com/synor/compute-sdk-go
+```
+
+## Quick Start
+
+```go
+package main
+
+import (
+    "context"
+    "fmt"
+    "log"
+
+    synor "github.com/synor/compute-sdk-go"
+)
+
+func main() {
+    client := synor.NewClient("your-api-key")
+
+    // Create tensors
+    a := synor.NewTensor(data, []int{512, 512}, synor.FP32)
+    b := synor.Zeros([]int{512, 512}, synor.FP32)
+
+    // Matrix multiplication
+    ctx := context.Background()
+    result, err := client.MatMul(ctx, a, b,
+        synor.WithPrecision(synor.FP16),
+        synor.WithProcessor(synor.GPU),
+    )
+    if err != nil {
+        log.Fatal(err)
+    }
+
+    fmt.Printf("Execution time: %.2fms\n", result.Metrics.ExecutionTimeMs)
+}
+```
+
+## Configuration
+
+```go
+config := synor.Config{
+    APIKey:    "your-api-key",
+    Endpoint:  "https://api.synor.io/compute/v1",
+    Strategy:  synor.Balanced,
+    Precision: synor.FP16,
+    Timeout:   30 * time.Second,
+}
+
+client := synor.NewClientWithConfig(config)
+```
+
+## Tensor Operations
+
+```go
+// Create tensors
+zeros := synor.Zeros([]int{3, 3}, synor.FP32)
+ones := synor.Ones([]int{2, 2}, synor.FP32)
+
+// From slice
+data := []float32{1, 2, 3, 4, 5, 6}
+tensor := synor.NewTensor(data, []int{2, 3}, synor.FP32)
+
+// Serialize for API
+serialized := tensor.Serialize()
+```
+
+## Matrix Operations
+
+```go
+// Matrix multiplication
+result, err := client.MatMul(ctx, a, b,
+    synor.WithPrecision(synor.FP16),
+    synor.WithProcessor(synor.GPU),
+    synor.WithStrategy(synor.Speed),
+)
+
+// 2D Convolution
+conv, err := client.Conv2D(ctx, input, kernel,
+    synor.WithStride(1, 1),
+    synor.WithPadding(1, 1),
+)
+
+// Attention
+attention, err := client.Attention(ctx, query, key, value,
+    synor.WithNumHeads(8),
+    synor.WithFlash(true),
+)
+```
+
+## LLM Inference
+
+```go
+// Single response
+response, err := client.Inference(ctx, "llama-3-70b", "Explain quantum computing",
+    synor.WithMaxTokens(512),
+    synor.WithTemperature(0.7),
+)
+fmt.Println(response.Result)
+
+// Streaming (using channel)
+stream, err := client.InferenceStream(ctx, "llama-3-70b", "Write a poem")
+for chunk := range stream {
+    fmt.Print(chunk)
+}
+```
+
+## Job Management
+
+```go
+// Submit async job
+job, err := client.SubmitJob(ctx, "matmul", map[string]interface{}{
+    "a": a.Serialize(),
+    "b": b.Serialize(),
+})
+
+// Get status
+status, err := client.GetJobStatus(ctx, job.ID)
+
+// Cancel
+err = client.CancelJob(ctx, job.ID)
+```
+
+## Error Handling
+
+```go
+result, err := client.MatMul(ctx, a, b)
+if err != nil {
+    if synorErr, ok := err.(*synor.SynorError); ok {
+        fmt.Printf("API Error: %s (status: %d)\n",
+            synorErr.Message, synorErr.StatusCode)
+    }
+}
+```
+
+## Processor Types
+
+```go
+synor.CPU     // General-purpose CPU
+synor.GPU     // NVIDIA/AMD GPU
+synor.TPU     // Google TPU
+synor.NPU     // Neural Processing Unit
+synor.LPU     // Language Processing Unit
+synor.FPGA    // Field-Programmable Gate Array
+synor.WASM    // WebAssembly runtime
+synor.WebGPU  // Browser GPU
+```
+
+## Precision Levels
+
+```go
+synor.FP64  // 64-bit float
+synor.FP32  // 32-bit float (default)
+synor.FP16  // 16-bit float
+synor.BF16  // Brain float 16
+synor.INT8  // 8-bit integer
+synor.INT4  // 4-bit integer
+```
+
+## Testing
+
+```bash
+go test ./...
+```
+
+## License
+
+MIT
--- a/sdk/java/README.md
+++ b/sdk/java/README.md
@ -0,0 +1,195 @@
+# Synor Compute SDK for Java
+
+Access distributed heterogeneous compute at 90% cost reduction.
+
+## Installation
+
+### Maven
+
+```xml
+<dependency>
+    <groupId>io.synor</groupId>
+    <artifactId>compute-sdk</artifactId>
+    <version>0.1.0</version>
+</dependency>
+```
+
+### Gradle
+
+```groovy
+implementation 'io.synor:compute-sdk:0.1.0'
+```
+
+## Quick Start
+
+```java
+import io.synor.compute.*;
+
+public class Example {
+    public static void main(String[] args) {
+        SynorCompute client = new SynorCompute("your-api-key");
+
+        // Matrix multiplication on GPU
+        Tensor a = Tensor.random(512, 512);
+        Tensor b = Tensor.random(512, 512);
+
+        JobResult<Tensor> result = client.matmul(a, b)
+            .precision(Precision.FP16)
+            .processor(ProcessorType.GPU)
+            .execute();
+
+        if (result.isSuccess()) {
+            System.out.println("Time: " + result.getExecutionTimeMs() + "ms");
+            System.out.println("Cost: $" + result.getCost());
+        }
+    }
+}
+```
+
+## Tensor Operations
+
+```java
+// Create tensors
+Tensor zeros = Tensor.zeros(3, 3);
+Tensor ones = Tensor.ones(2, 2);
+Tensor random = Tensor.random(10, 10);
+Tensor randn = Tensor.randn(100);
+Tensor eye = Tensor.eye(3);
+
+// From array
+double[][] data = {{1, 2, 3}, {4, 5, 6}};
+Tensor tensor = Tensor.fromArray(data);
+
+// Operations
+Tensor reshaped = tensor.reshape(3, 2);
+Tensor transposed = tensor.transpose();
+
+// Math
+double mean = tensor.mean();
+double sum = tensor.sum();
+double std = tensor.std();
+```
+
+## Builder Pattern API
+
+```java
+// Matrix multiplication
+JobResult<Tensor> result = client.matmul(a, b)
+    .precision(Precision.FP16)
+    .processor(ProcessorType.GPU)
+    .priority(Priority.HIGH)
+    .execute();
+
+// 2D Convolution
+JobResult<Tensor> conv = client.conv2d(input, kernel)
+    .stride(1, 1)
+    .padding(1, 1)
+    .execute();
+
+// Attention
+JobResult<Tensor> attention = client.attention(query, key, value)
+    .numHeads(8)
+    .flash(true)
+    .execute();
+```
+
+## Async API with CompletableFuture
+
+```java
+import java.util.concurrent.CompletableFuture;
+
+CompletableFuture<JobResult<Tensor>> future = client.matmul(a, b)
+    .precision(Precision.FP16)
+    .executeAsync();
+
+future.thenAccept(result -> {
+    System.out.println("Completed: " + result.isSuccess());
+});
+```
+
+## LLM Inference
+
+```java
+// Single response
+InferenceResult response = client.inference("llama-3-70b", "Explain quantum computing")
+    .maxTokens(512)
+    .temperature(0.7)
+    .execute();
+System.out.println(response.getResult());
+
+// Streaming with callback
+client.inferenceStream("llama-3-70b", "Write a poem", chunk -> {
+    System.out.print(chunk);
+});
+```
+
+## Configuration
+
+```java
+SynorConfig config = SynorConfig.builder()
+    .apiKey("your-api-key")
+    .baseUrl("https://api.synor.io/compute/v1")
+    .defaultProcessor(ProcessorType.GPU)
+    .defaultPrecision(Precision.FP16)
+    .timeout(Duration.ofSeconds(30))
+    .debug(true)
+    .build();
+
+SynorCompute client = new SynorCompute(config);
+```
+
+## Error Handling
+
+```java
+try {
+    JobResult<Tensor> result = client.matmul(a, b).execute();
+} catch (SynorException e) {
+    System.err.println("API Error: " + e.getMessage());
+    System.err.println("Status: " + e.getStatusCode());
+}
+```
+
+## Enums
+
+```java
+// Processor types
+ProcessorType.CPU
+ProcessorType.GPU
+ProcessorType.TPU
+ProcessorType.NPU
+ProcessorType.LPU
+ProcessorType.FPGA
+ProcessorType.AUTO
+
+// Precision
+Precision.FP64
+Precision.FP32
+Precision.FP16
+Precision.BF16
+Precision.INT8
+Precision.INT4
+
+// Job status
+JobStatus.PENDING
+JobStatus.RUNNING
+JobStatus.COMPLETED
+JobStatus.FAILED
+```
+
+## Requirements
+
+- Java 11 or higher
+- Gson for JSON serialization
+- OkHttp for HTTP client
+
+## Testing
+
+```bash
+mvn test
+# or
+./gradlew test
+```
+
+## License
+
+MIT
--- a/sdk/js/README.md
+++ b/sdk/js/README.md
@ -0,0 +1,159 @@
+# Synor Compute SDK for JavaScript/TypeScript
+
+Access distributed heterogeneous compute at 90% cost reduction.
+
+## Installation
+
+```bash
+npm install synor-compute
+# or
+pnpm add synor-compute
+# or
+yarn add synor-compute
+```
+
+## Quick Start
+
+```typescript
+import { SynorCompute, Tensor } from 'synor-compute';
+
+const client = new SynorCompute('your-api-key');
+
+// Matrix multiplication on GPU
+const a = Tensor.random([512, 512]);
+const b = Tensor.random([512, 512]);
+const result = await client.matmul(a, b, {
+  precision: 'fp16',
+  processor: 'gpu'
+});
+
+console.log(`Execution time: ${result.executionTimeMs}ms`);
+console.log(`Cost: $${result.cost}`);
+```
+
+## Tensor Operations
+
+```typescript
+// Create tensors
+const zeros = Tensor.zeros([3, 3]);
+const ones = Tensor.ones([2, 2]);
+const random = Tensor.random([10, 10]);
+const randn = Tensor.randn([100]); // Normal distribution
+
+// From array
+const data = Tensor.from([1, 2, 3, 4, 5, 6], [2, 3]);
+
+// Operations
+const reshaped = data.reshape([3, 2]);
+const transposed = data.transpose();
+```
+
+## Matrix Operations
+
+```typescript
+// Matrix multiplication
+const result = await client.matmul(a, b, {
+  precision: 'fp16',
+  processor: 'gpu',
+  strategy: 'speed'
+});
+
+// 2D Convolution
+const conv = await client.conv2d(input, kernel, {
+  stride: [1, 1],
+  padding: [1, 1]
+});
+
+// Flash Attention
+const attention = await client.attention(query, key, value, {
+  numHeads: 8,
+  flash: true
+});
+```
+
+## LLM Inference
+
+```typescript
+// Single response
+const response = await client.inference('llama-3-70b', 'Explain quantum computing', {
+  maxTokens: 512,
+  temperature: 0.7
+});
+console.log(response.result);
+
+// Streaming response
+for await (const chunk of client.inferenceStream('llama-3-70b', 'Write a poem')) {
+  process.stdout.write(chunk);
+}
+```
+
+## Configuration
+
+```typescript
+const client = new SynorCompute({
+  apiKey: 'your-api-key',
+  baseUrl: 'https://api.synor.io/compute/v1', // or localhost:17250 for local
+  defaultProcessor: 'gpu',
+  defaultPrecision: 'fp16',
+  defaultStrategy: 'balanced',
+  timeout: 30000,
+  debug: false
+});
+```
+
+## Job Management
+
+```typescript
+// Submit async job
+const job = await client.submitJob('matmul', { a, b });
+
+// Poll for status
+const status = await client.getJobStatus(job.jobId);
+
+// Cancel job
+await client.cancelJob(job.jobId);
+```
+
+## Error Handling
+
+```typescript
+import { SynorError } from 'synor-compute';
+
+try {
+  const result = await client.matmul(a, b);
+} catch (error) {
+  if (error instanceof SynorError) {
+    console.error(`API Error: ${error.message} (${error.statusCode})`);
+  }
+}
+```
+
+## TypeScript Support
+
+Full TypeScript support with exported types:
+
+```typescript
+import type {
+  Tensor,
+  ProcessorType,
+  Precision,
+  BalancingStrategy,
+  JobStatus,
+  SynorConfig,
+  MatMulOptions,
+  InferenceOptions,
+  JobResult
+} from 'synor-compute';
+```
+
+## Testing
+
+```bash
+npm test
+# or
+pnpm test
+```
+
+## License
+
+MIT
--- a/sdk/kotlin/README.md
+++ b/sdk/kotlin/README.md
@ -0,0 +1,206 @@
+# Synor Compute SDK for Kotlin
+
+Access distributed heterogeneous compute at 90% cost reduction.
+
+## Installation
+
+### Gradle (Kotlin DSL)
+
+```kotlin
+implementation("io.synor:compute-sdk-kotlin:0.1.0")
+```
+
+### Gradle (Groovy)
+
+```groovy
+implementation 'io.synor:compute-sdk-kotlin:0.1.0'
+```
+
+## Quick Start
+
+```kotlin
+import io.synor.compute.*
+import kotlinx.coroutines.runBlocking
+
+fun main() = runBlocking {
+    val client = SynorCompute("your-api-key")
+
+    // Matrix multiplication on GPU
+    val a = Tensor.random(512, 512)
+    val b = Tensor.random(512, 512)
+
+    val result = client.matmul(a, b) {
+        precision = Precision.FP16
+        processor = ProcessorType.GPU
+    }
+
+    if (result.isSuccess) {
+        println("Time: ${result.executionTimeMs}ms")
+        println("Cost: $${result.cost}")
+    }
+}
+```
+
+## Kotlin Coroutines Support
+
+```kotlin
+// Suspend functions
+suspend fun compute() {
+    val result = client.matmul(a, b)
+    println(result.result)
+}
+
+// Flows for streaming
+client.inferenceStream("llama-3-70b", "Write a poem")
+    .collect { chunk ->
+        print(chunk)
+    }
+```
+
+## Tensor Operations
+
+```kotlin
+// Create tensors
+val zeros = Tensor.zeros(3, 3)
+val ones = Tensor.ones(2, 2)
+val random = Tensor.random(10, 10)
+val randn = Tensor.randn(100)
+val eye = Tensor.eye(3)
+
+// From array
+val data = arrayOf(
+    floatArrayOf(1f, 2f, 3f),
+    floatArrayOf(4f, 5f, 6f)
+)
+val tensor = Tensor.from(data)
+
+// Operations
+val reshaped = tensor.reshape(3, 2)
+val transposed = tensor.transpose()
+
+// Math (extension properties)
+val mean = tensor.mean
+val sum = tensor.sum
+val std = tensor.std
+```
+
+## DSL-Style API
+
+```kotlin
+// Matrix multiplication with DSL
+val result = client.matmul(a, b) {
+    precision = Precision.FP16
+    processor = ProcessorType.GPU
+    priority = Priority.HIGH
+    strategy = BalancingStrategy.SPEED
+}
+
+// Convolution
+val conv = client.conv2d(input, kernel) {
+    stride = 1 to 1
+    padding = 1 to 1
+}
+
+// Attention
+val attention = client.attention(query, key, value) {
+    numHeads = 8
+    flash = true
+}
+```
+
+## LLM Inference
+
+```kotlin
+// Single response
+val response = client.inference("llama-3-70b", "Explain quantum computing") {
+    maxTokens = 512
+    temperature = 0.7
+}
+println(response.result)
+
+// Streaming with Flow
+client.inferenceStream("llama-3-70b", "Write a poem")
+    .collect { chunk ->
+        print(chunk)
+    }
+```
+
+## Configuration
+
+```kotlin
+val config = SynorConfig(
+    apiKey = "your-api-key",
+    baseUrl = "https://api.synor.io/compute/v1",
+    defaultProcessor = ProcessorType.GPU,
+    defaultPrecision = Precision.FP16,
+    timeout = 30.seconds,
+    debug = true
+)
+
+val client = SynorCompute(config)
+```
+
+## Error Handling
+
+```kotlin
+try {
+    val result = client.matmul(a, b)
+} catch (e: SynorException) {
+    println("API Error: ${e.message} (${e.statusCode})")
+}
+
+// Or with Result type
+val result = runCatching {
+    client.matmul(a, b)
+}
+
+result.onSuccess { println("Success: ${it.result}") }
+       .onFailure { println("Failed: ${it.message}") }
+```
+
+## Extension Functions
+
+```kotlin
+// Operator overloading
+val c = a * b  // Matrix multiplication
+val d = a + b  // Element-wise addition
+
+// Infix functions
+val result = a matmul b
+```
+
+## Types
+
+```kotlin
+// Sealed classes for type safety
+sealed class ProcessorType {
+    object Cpu : ProcessorType()
+    object Gpu : ProcessorType()
+    object Tpu : ProcessorType()
+    object Auto : ProcessorType()
+}
+
+enum class Precision {
+    FP64, FP32, FP16, BF16, INT8, INT4
+}
+
+enum class JobStatus {
+    PENDING, RUNNING, COMPLETED, FAILED, CANCELLED
+}
+```
+
+## Requirements
+
+- Kotlin 1.9+
+- Kotlinx Coroutines
+- Kotlinx Serialization
+
+## Testing
+
+```bash
+./gradlew test
+```
+
+## License
+
+MIT
--- a/sdk/python/README.md
+++ b/sdk/python/README.md
@ -0,0 +1,187 @@
+# Synor Compute SDK for Python
+
+Access distributed heterogeneous compute at 90% cost reduction.
+
+## Installation
+
+```bash
+pip install synor-compute
+# or
+poetry add synor-compute
+```
+
+## Quick Start
+
+```python
+import asyncio
+from synor_compute import SynorCompute, Tensor
+
+async def main():
+    client = SynorCompute('your-api-key')
+
+    # Matrix multiplication on GPU
+    a = Tensor.random((512, 512))
+    b = Tensor.random((512, 512))
+    result = await client.matmul(a, b, precision='fp16', processor='gpu')
+
+    print(f"Execution time: {result.execution_time_ms}ms")
+    print(f"Cost: ${result.cost}")
+
+asyncio.run(main())
+```
+
+## NumPy Integration
+
+```python
+import numpy as np
+from synor_compute import Tensor
+
+# Create from NumPy
+arr = np.random.randn(100, 100).astype(np.float32)
+tensor = Tensor.from_numpy(arr)
+
+# Convert back to NumPy
+result_np = tensor.numpy()
+```
+
+## Tensor Operations
+
+```python
+# Create tensors
+zeros = Tensor.zeros((3, 3))
+ones = Tensor.ones((2, 2))
+random = Tensor.random((10, 10))
+randn = Tensor.randn((100,))  # Normal distribution
+
+# Operations
+reshaped = tensor.reshape((50, 200))
+transposed = tensor.T
+
+# Math operations
+mean = tensor.mean()
+std = tensor.std()
+```
+
+## Matrix Operations
+
+```python
+# Matrix multiplication
+result = await client.matmul(a, b,
+    precision='fp16',
+    processor='gpu',
+    strategy='speed'
+)
+
+# 2D Convolution
+conv = await client.conv2d(input_tensor, kernel,
+    stride=(1, 1),
+    padding=(1, 1)
+)
+
+# Flash Attention
+attention = await client.attention(query, key, value,
+    num_heads=8,
+    flash=True
+)
+```
+
+## LLM Inference
+
+```python
+# Single response
+response = await client.inference(
+    'llama-3-70b',
+    'Explain quantum computing',
+    max_tokens=512,
+    temperature=0.7
+)
+print(response.result)
+
+# Streaming response
+async for chunk in client.inference_stream('llama-3-70b', 'Write a poem'):
+    print(chunk, end='', flush=True)
+```
+
+## Configuration
+
+```python
+from synor_compute import SynorCompute, Config
+
+config = Config(
+    api_key='your-api-key',
+    base_url='https://api.synor.io/compute/v1',
+    default_processor='gpu',
+    default_precision='fp16',
+    default_strategy='balanced',
+    timeout=30.0,
+    debug=False
+)
+
+client = SynorCompute(config)
+```
+
+## Synchronous API
+
+For non-async contexts:
+
+```python
+from synor_compute import SynorComputeSync
+
+client = SynorComputeSync('your-api-key')
+result = client.matmul(a, b)  # Blocking call
+```
+
+## Job Management
+
+```python
+# Submit async job
+job = await client.submit_job('matmul', {'a': a, 'b': b})
+
+# Poll for status
+status = await client.get_job_status(job.job_id)
+
+# Wait for completion
+result = await client.wait_for_job(job.job_id, timeout=60.0)
+
+# Cancel job
+await client.cancel_job(job.job_id)
+```
+
+## Error Handling
+
+```python
+from synor_compute import SynorError
+
+try:
+    result = await client.matmul(a, b)
+except SynorError as e:
+    print(f"API Error: {e.message} (status: {e.status_code})")
+```
+
+## Type Hints
+
+Full type hint support:
+
+```python
+from synor_compute.types import (
+    ProcessorType,
+    Precision,
+    BalancingStrategy,
+    JobStatus,
+    MatMulOptions,
+    InferenceOptions,
+    JobResult
+)
+```
+
+## Testing
+
+```bash
+pytest
+# or
+python -m pytest tests/
+```
+
+## License
+
+MIT
--- a/sdk/ruby/README.md
+++ b/sdk/ruby/README.md
@ -0,0 +1,251 @@
+# Synor Compute SDK for Ruby
+
+Access distributed heterogeneous compute at 90% cost reduction.
+
+## Installation
+
+Add to `Gemfile`:
+
+```ruby
+gem 'synor_compute'
+```
+
+Then run:
+
+```bash
+bundle install
+```
+
+Or install directly:
+
+```bash
+gem install synor_compute
+```
+
+## Quick Start
+
+```ruby
+require 'synor_compute'
+
+client = SynorCompute::Client.new('your-api-key')
+
+# Matrix multiplication on GPU
+a = SynorCompute::Tensor.random([512, 512])
+b = SynorCompute::Tensor.random([512, 512])
+
+result = client.matmul(a, b,
+  precision: :fp16,
+  processor: :gpu
+)
+
+if result.success?
+  puts "Time: #{result.execution_time_ms}ms"
+  puts "Cost: $#{result.cost}"
+end
+```
+
+## Tensor Operations
+
+```ruby
+# Create tensors
+zeros = SynorCompute::Tensor.zeros([3, 3])
+ones = SynorCompute::Tensor.ones([2, 2])
+random = SynorCompute::Tensor.random([10, 10])
+randn = SynorCompute::Tensor.randn([100])
+eye = SynorCompute::Tensor.eye(3)
+
+# From array
+data = [1.0, 2.0, 3.0, 4.0, 5.0, 6.0]
+tensor = SynorCompute::Tensor.new(data, shape: [2, 3])
+
+# Operations
+reshaped = tensor.reshape([3, 2])
+transposed = tensor.transpose
+
+# Math
+mean = tensor.mean
+sum = tensor.sum
+std = tensor.std
+```
+
+## Matrix Operations
+
+```ruby
+# Matrix multiplication
+result = client.matmul(a, b,
+  precision: :fp16,
+  processor: :gpu,
+  strategy: :speed
+)
+
+# 2D Convolution
+conv = client.conv2d(input, kernel,
+  stride: [1, 1],
+  padding: [1, 1]
+)
+
+# Attention
+attention = client.attention(query, key, value,
+  num_heads: 8,
+  flash: true
+)
+```
+
+## LLM Inference
+
+```ruby
+# Single response
+response = client.inference('llama-3-70b', 'Explain quantum computing',
+  max_tokens: 512,
+  temperature: 0.7
+)
+puts response.result
+
+# Streaming with block
+client.inference_stream('llama-3-70b', 'Write a poem') do |chunk|
+  print chunk
+end
+
+# Streaming with Enumerator
+client.inference_stream('llama-3-70b', 'Write a poem').each do |chunk|
+  print chunk
+end
+```
+
+## Configuration
+
+```ruby
+config = SynorCompute::Config.new(
+  api_key: 'your-api-key',
+  base_url: 'https://api.synor.io/compute/v1',
+  default_processor: :gpu,
+  default_precision: :fp16,
+  timeout: 30,
+  debug: true
+)
+
+client = SynorCompute::Client.new(config)
+
+# Or with block
+SynorCompute.configure do |config|
+  config.api_key = 'your-api-key'
+  config.default_processor = :gpu
+end
+```
+
+## Rails Integration
+
+```ruby
+# config/initializers/synor_compute.rb
+SynorCompute.configure do |config|
+  config.api_key = Rails.application.credentials.synor[:api_key]
+  config.default_processor = :gpu
+end
+
+# In your controller/service
+class ComputeService
+  def self.client
+    @client ||= SynorCompute::Client.new
+  end
+
+  def self.compute(a, b)
+    client.matmul(a, b)
+  end
+end
+```
+
+## Error Handling
+
+```ruby
+begin
+  result = client.matmul(a, b)
+rescue SynorCompute::ApiError => e
+  puts "API Error #{e.status_code}: #{e.message}"
+rescue SynorCompute::NetworkError => e
+  puts "Network error: #{e.message}"
+rescue SynorCompute::Error => e
+  puts "Error: #{e.message}"
+end
+```
+
+## Types
+
+```ruby
+# Processor types (symbols)
+:cpu, :gpu, :tpu, :npu, :lpu, :fpga, :auto
+
+# Precision (symbols)
+:fp64, :fp32, :fp16, :bf16, :int8, :int4
+
+# Job status (symbols)
+:pending, :running, :completed, :failed, :cancelled
+
+# Balancing strategy (symbols)
+:speed, :cost, :energy, :latency, :balanced
+```
+
+## Job Management
+
+```ruby
+# Submit async job
+job = client.submit_job(:matmul, a: a, b: b)
+
+# Poll for status
+status = client.job_status(job.id)
+
+# Wait for completion
+result = client.wait_for_job(job.id, timeout: 60)
+
+# Cancel job
+client.cancel_job(job.id)
+```
+
+## Ruby-Specific Features
+
+### Method Chaining
+
+```ruby
+result = client
+  .matmul(a, b)
+  .tap { |r| puts "Computing..." }
+  .then { |r| r.data if r.success? }
+```
+
+### Duck Typing
+
+```ruby
+# Any object with #to_tensor method works
+class MyData
+  def to_tensor
+    SynorCompute::Tensor.new(@data, shape: @shape)
+  end
+end
+
+result = client.matmul(MyData.new, b)
+```
+
+### Frozen String Literals
+
+The gem is compatible with frozen string literals:
+
+```ruby
+# frozen_string_literal: true
+require 'synor_compute'
+```
+
+## Requirements
+
+- Ruby 3.0+
+- faraday gem for HTTP
+
+## Testing
+
+```bash
+bundle exec rake test
+# or
+bundle exec rspec
+```
+
+## License
+
+MIT
--- a/sdk/rust/README.md
+++ b/sdk/rust/README.md
@ -0,0 +1,224 @@
+# Synor Compute SDK for Rust
+
+Access distributed heterogeneous compute at 90% cost reduction.
+
+## Installation
+
+Add to `Cargo.toml`:
+
+```toml
+[dependencies]
+synor-compute = "0.1"
+tokio = { version = "1", features = ["full"] }
+```
+
+## Quick Start
+
+```rust
+use synor_compute::{SynorCompute, Tensor, Precision, ProcessorType};
+
+#[tokio::main]
+async fn main() -> Result<(), Box<dyn std::error::Error>> {
+    let client = SynorCompute::new("your-api-key");
+
+    // Matrix multiplication on GPU
+    let a = Tensor::rand(&[512, 512]);
+    let b = Tensor::rand(&[512, 512]);
+
+    let result = client.matmul(&a, &b)
+        .precision(Precision::FP16)
+        .processor(ProcessorType::Gpu)
+        .send()
+        .await?;
+
+    if result.is_success() {
+        println!("Time: {}ms", result.execution_time_ms.unwrap_or(0));
+        println!("Cost: ${}", result.cost.unwrap_or(0.0));
+    }
+
+    Ok(())
+}
+```
+
+## Tensor Operations
+
+```rust
+// Create tensors
+let zeros = Tensor::zeros(&[3, 3]);
+let ones = Tensor::ones(&[2, 2]);
+let random = Tensor::rand(&[10, 10]);    // Uniform [0, 1)
+let randn = Tensor::randn(&[100]);       // Normal distribution
+let eye = Tensor::eye(3);                // Identity matrix
+
+// From data
+let data = vec![1.0, 2.0, 3.0, 4.0, 5.0, 6.0];
+let tensor = Tensor::new(&[2, 3], data);
+
+// Ranges
+let range = Tensor::arange(0.0, 10.0, 1.0);
+let linspace = Tensor::linspace(0.0, 1.0, 100);
+
+// Operations
+let reshaped = tensor.reshape(&[3, 2]);
+let transposed = tensor.transpose();
+
+// Math
+let mean = tensor.mean();
+let sum = tensor.sum();
+let std = tensor.std();
+
+// Activations
+let relu = tensor.relu();
+let sigmoid = tensor.sigmoid();
+let softmax = tensor.softmax();
+```
+
+## Builder Pattern API
+
+```rust
+// Matrix multiplication with options
+let result = client.matmul(&a, &b)
+    .precision(Precision::FP16)
+    .processor(ProcessorType::Gpu)
+    .priority(Priority::High)
+    .send()
+    .await?;
+
+// 2D Convolution
+let result = client.conv2d(&input, &kernel)
+    .stride((1, 1))
+    .padding((1, 1))
+    .precision(Precision::FP32)
+    .send()
+    .await?;
+
+// Attention
+let result = client.attention(&query, &key, &value)
+    .num_heads(8)
+    .flash(true)
+    .precision(Precision::FP16)
+    .send()
+    .await?;
+```
+
+## LLM Inference
+
+```rust
+// Single response
+let response = client.inference("llama-3-70b", "Explain quantum computing")
+    .send()
+    .await?;
+println!("{}", response.result.unwrap_or_default());
+
+// Streaming with futures
+use futures::StreamExt;
+
+let mut stream = client.inference_stream("llama-3-70b", "Write a poem").await?;
+while let Some(token) = stream.next().await {
+    print!("{}", token?);
+}
+```
+
+## Configuration
+
+```rust
+use synor_compute::Config;
+
+let config = Config::new("your-api-key")
+    .base_url("https://api.synor.io/compute/v1")
+    .default_processor(ProcessorType::Gpu)
+    .default_precision(Precision::FP16)
+    .timeout_secs(30)
+    .debug(true);
+
+let client = SynorCompute::with_config(config);
+```
+
+## Error Handling
+
+```rust
+use synor_compute::{Error, Result};
+
+async fn compute() -> Result<()> {
+    let result = client.matmul(&a, &b).send().await?;
+
+    match result.status {
+        JobStatus::Completed => println!("Success!"),
+        JobStatus::Failed => {
+            if let Some(err) = result.error {
+                eprintln!("Job failed: {}", err);
+            }
+        }
+        _ => {}
+    }
+    Ok(())
+}
+
+// Pattern matching on errors
+match client.matmul(&a, &b).send().await {
+    Ok(result) => println!("Result: {:?}", result),
+    Err(Error::Api { status_code, message }) => {
+        eprintln!("API error {}: {}", status_code, message);
+    }
+    Err(Error::InvalidArgument(msg)) => {
+        eprintln!("Invalid argument: {}", msg);
+    }
+    Err(e) => eprintln!("Other error: {}", e),
+}
+```
+
+## Types
+
+```rust
+// Processor types
+ProcessorType::Cpu
+ProcessorType::Gpu
+ProcessorType::Tpu
+ProcessorType::Npu
+ProcessorType::Lpu
+ProcessorType::Fpga
+ProcessorType::Auto  // Automatic selection
+
+// Precision levels
+Precision::FP64
+Precision::FP32
+Precision::FP16
+Precision::BF16
+Precision::INT8
+Precision::INT4
+
+// Job status
+JobStatus::Pending
+JobStatus::Running
+JobStatus::Completed
+JobStatus::Failed
+JobStatus::Cancelled
+
+// Priority
+Priority::Low
+Priority::Normal
+Priority::High
+Priority::Critical
+```
+
+## Features
+
+Enable optional features in `Cargo.toml`:
+
+```toml
+[dependencies]
+synor-compute = { version = "0.1", features = ["serde", "rayon"] }
+```
+
+- `serde` - Serialization support (enabled by default)
+- `rayon` - Parallel tensor operations
+
+## Testing
+
+```bash
+cargo test
+```
+
+## License
+
+MIT
--- a/sdk/swift/README.md
+++ b/sdk/swift/README.md
@ -0,0 +1,227 @@
+# Synor Compute SDK for Swift
+
+Access distributed heterogeneous compute at 90% cost reduction.
+
+## Installation
+
+### Swift Package Manager
+
+Add to `Package.swift`:
+
+```swift
+dependencies: [
+    .package(url: "https://github.com/synor/compute-sdk-swift", from: "0.1.0")
+]
+```
+
+### Xcode
+
+File > Add Packages > Enter URL:
+`https://github.com/synor/compute-sdk-swift`
+
+## Quick Start
+
+```swift
+import SynorCompute
+
+let client = SynorCompute(apiKey: "your-api-key")
+
+// Matrix multiplication on GPU
+let a = Tensor.random(shape: [512, 512])
+let b = Tensor.random(shape: [512, 512])
+
+Task {
+    let result = try await client.matmul(a, b,
+        precision: .fp16,
+        processor: .gpu
+    )
+
+    if result.isSuccess {
+        print("Time: \(result.executionTimeMs ?? 0)ms")
+        print("Cost: $\(result.cost ?? 0)")
+    }
+}
+```
+
+## Tensor Operations
+
+```swift
+// Create tensors
+let zeros = Tensor.zeros(shape: [3, 3])
+let ones = Tensor.ones(shape: [2, 2])
+let random = Tensor.random(shape: [10, 10])
+let randn = Tensor.randn(shape: [100])
+let eye = Tensor.eye(size: 3)
+
+// From array
+let data: [Float] = [1, 2, 3, 4, 5, 6]
+let tensor = Tensor(data: data, shape: [2, 3])
+
+// Operations
+let reshaped = tensor.reshape(to: [3, 2])
+let transposed = tensor.transpose()
+
+// Math
+let mean = tensor.mean()
+let sum = tensor.sum()
+let std = tensor.std()
+```
+
+## Async/Await API
+
+```swift
+// Matrix multiplication
+let result = try await client.matmul(a, b,
+    precision: .fp16,
+    processor: .gpu,
+    strategy: .speed
+)
+
+// Convolution
+let conv = try await client.conv2d(input, kernel,
+    stride: (1, 1),
+    padding: (1, 1)
+)
+
+// Attention
+let attention = try await client.attention(query, key, value,
+    numHeads: 8,
+    flash: true
+)
+```
+
+## LLM Inference
+
+```swift
+// Single response
+let response = try await client.inference(
+    model: "llama-3-70b",
+    prompt: "Explain quantum computing",
+    maxTokens: 512,
+    temperature: 0.7
+)
+print(response.result ?? "")
+
+// Streaming with AsyncSequence
+for try await chunk in client.inferenceStream(
+    model: "llama-3-70b",
+    prompt: "Write a poem"
+) {
+    print(chunk, terminator: "")
+}
+```
+
+## Configuration
+
+```swift
+let config = SynorConfig(
+    apiKey: "your-api-key",
+    baseUrl: "https://api.synor.io/compute/v1",
+    defaultProcessor: .gpu,
+    defaultPrecision: .fp16,
+    timeout: 30,
+    debug: true
+)
+
+let client = SynorCompute(config: config)
+```
+
+## SwiftUI Integration
+
+```swift
+import SwiftUI
+import SynorCompute
+
+struct ComputeView: View {
+    @StateObject private var vm = ComputeViewModel()
+
+    var body: some View {
+        VStack {
+            if vm.isLoading {
+                ProgressView()
+            } else if let result = vm.result {
+                Text("Result: \(result)")
+            }
+            Button("Compute") {
+                Task { await vm.compute() }
+            }
+        }
+    }
+}
+
+@MainActor
+class ComputeViewModel: ObservableObject {
+    @Published var result: String?
+    @Published var isLoading = false
+
+    private let client = SynorCompute(apiKey: "your-api-key")
+
+    func compute() async {
+        isLoading = true
+        defer { isLoading = false }
+
+        do {
+            let response = try await client.inference(
+                model: "llama-3-70b",
+                prompt: "Hello"
+            )
+            result = response.result
+        } catch {
+            result = "Error: \(error.localizedDescription)"
+        }
+    }
+}
+```
+
+## Error Handling
+
+```swift
+do {
+    let result = try await client.matmul(a, b)
+} catch let error as SynorError {
+    switch error {
+    case .apiError(let statusCode, let message):
+        print("API Error \(statusCode): \(message)")
+    case .networkError(let underlying):
+        print("Network error: \(underlying)")
+    case .invalidArgument(let message):
+        print("Invalid argument: \(message)")
+    }
+} catch {
+    print("Unexpected error: \(error)")
+}
+```
+
+## Types
+
+```swift
+// Processor types
+enum ProcessorType: String, Codable {
+    case cpu, gpu, tpu, npu, lpu, fpga, auto
+}
+
+// Precision
+enum Precision: String, Codable {
+    case fp64, fp32, fp16, bf16, int8, int4
+}
+
+// Job status
+enum JobStatus: String, Codable {
+    case pending, running, completed, failed, cancelled
+}
+```
+
+## Requirements
+
+- iOS 15.0+ / macOS 12.0+ / tvOS 15.0+ / watchOS 8.0+
+- Swift 5.9+
+
+## Testing
+
+```bash
+swift test
+```
+
+## License
+
+MIT