⚡ Qoder Workflow: Unified Suite & AI Gateway
Comprehensive, modular framework and high-performance OpenAI-compatible gateway integrating all essential capabilities of the Qoder ecosystem.
🏗️ Unified 5-Stage Architecture
Qoder Workflow seamlessly integrates all functional modules into an end-to-end framework:
┌──────────────────────────────────────────────────────────────────────────────────┐
│ UNIFIED QODER SUITE │
├──────────────────────────────────────────────────────────────────────────────────┤
│ │
│ [Stage 1: Provisioner & Patcher] ──► [Stage 2: Harvester] ──► [Stage 3: Decrypt] │
│ (Auth, SSO & Machine Patcher) (PAT Generator pt-*) (AES-CBC Session) │
│ │ │
│ ▼ │
│ [Stage 5: OpenAI & Proxy Bridge] ◄── [Stage 4: COSY Claim Engine] ◄───┘ │
│ (REST API & 15 Top Models) (Activity & Quota Check) │
│ │
└──────────────────────────────────────────────────────────────────────────────────┘
Module Capabilities
- Provisioner & Desktop Patcher (
src/provisioner/ & src/patcher/):
- Multi-platform identity patching (macOS, Windows, Linux) for machine IDs, MAC addresses, and application states.
- Google Workspace SSO handshake, mail verification flows, and account status initializations.
- PAT Harvester (
src/harvester/):
- Automated REST generation of official 64-character Personal Access Tokens (
pt-*).
- Session Decryptor (
src/decryptor/):
- Isolated sandbox runner for Qoder CLI and cryptographic AES-CBC-128 extraction of
.auth/user access tokens (jt-*).
- COSY Activity & Claim Engine (
src/claimer/):
- Hardware machine binding (
Cosy-MachineToken, Cosy-MachineCode) with dynamic GMT MD5 date-signatures (cosy&war, war never changes&<DATE>) for quota inspection and activity eligibility scanning.
- OpenAI & Anthropic Proxy Gateway (
src/gateway/ & src/proxy/):
- Drop-in HTTP REST API server (
/v1/chat/completions & /v1/models) supporting SSE streaming, fallback routing, and native model mappings for 15 top AI architectures.
🎯 Top Models Qoder PAT
Integrated model catalog available through the gateway:
| Model ID |
Model Name |
Upstream Architecture |
Context Window |
Category / Specialization |
qd/auto |
Auto |
Auto Dynamic Routing |
128k |
Dynamic multi-tier routing |
qd/ultimate |
Opus 4.7 |
Claude Opus 4.7 |
200k |
Deep reasoning & complex architecture |
qd/performance |
Sonnet 4.6 |
Claude Sonnet 4.6 |
200k |
High-speed coding & precision |
qd/efficient |
Haiku 4.5 |
Claude Haiku 4.5 |
128k |
Token efficiency & lightweight tasks |
qd/lite |
Lite |
Lite Engine |
64k |
High throughput basic parsing |
qd/qmodel_38max |
Qwen3.8-Max |
Qwen 3.8 Max |
128k |
Advanced coding & mathematics |
qd/qmodel_latest |
Qwen3.7-Max |
Qwen 3.7 Max |
128k |
Flagship coding model |
qd/qmodel |
Qwen3.7-Plus |
Qwen 3.7 Plus |
128k |
General software engineering |
qd/kmodel_latest |
Kimi-K3 |
Moonshot Kimi K3 |
200k |
Long-context & document analysis |
qd/kmodel |
Kimi-K2.7-Code |
Moonshot Kimi K2.7 Code |
128k |
Syntax & code refactoring |
qd/gm51model |
GLM-5.2 |
Zhipu GLM-5.2 |
128k |
Analytical logic & reasoning |
qd/dmodel |
DeepSeek-V4-Pro |
DeepSeek V4 Pro |
128k |
Complex algorithms & deep logic |
qd/dfmodel |
DeepSeek-V4-Flash |
DeepSeek V4 Flash |
64k |
Low-latency response generation |
qd/mmodel |
MiniMax-M3 |
MiniMax M3 |
128k |
Natural language dialogue & writing |
qd/cmodel |
Fable 5 |
Fable 5 |
128k |
Technical writing & synthesis |
📂 Repository Structure
qoder-workflow/
├── config/
│ ├── config.example.toml # Engine TOML Configuration
│ ├── config.example.json # Engine JSON Configuration
│ └── models.json # 15 Top Models Catalog Definition
├── src/
│ ├── provisioner/ # Provisioning & SSO Handshake
│ ├── harvester/ # PAT Token Harvester API
│ ├── patcher/ # Multi-Platform Identity Patcher
│ ├── decryptor/ # AES-CBC CLI Session Decryptor
│ ├── claimer/ # COSY Signature & Claim Engine
│ ├── proxy/ # OpenAI Model Router & Fallback
│ └── gateway/ # OpenAI REST API HTTP Server
├── test/
│ ├── test_decrypt.py # Unit tests for AES decryption
│ ├── test_cosy_sign.py # Unit tests for COSY MD5 signature
│ ├── test_patcher_router.py # Unit tests for Patcher & Model Router
│ └── test_gateway.js # Integration tests for OpenAI Gateway
├── main.py # Unified CLI Orchestrator
├── package.json # Node.js Package Specification
├── README.md # Complete Documentation
└── LICENSE # MIT License
🚀 Quick Start
1. Launch OpenAI-Compatible API Gateway
git clone https://github.com/d4ncboz/qoder-workflow.git
cd qoder-workflow
npm start
The gateway starts listening on http://localhost:20240.
2. Test Model Routing via cURL
curl http://localhost:20240/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "qd/ultimate",
"messages": [
{"role": "user", "content": "Hello world"}
],
"stream": false
}'
3. Run Verification & State Patcher via CLI
# Verify PAT & Check Live Quota
python3 main.py --action verify-pat --pat "pt-..."
# Scan Activity Eligibility
python3 main.py --action scan-eligibility --pat "pt-..."
- Community discussions, project updates, and ecosystem releases: BOZDROP CRYPTO.
- Support independent open-source development via SAWERIA.
📄 License
Licensed under the MIT License © 2026 D4NNBOZ.