⚡ Verdent 2API

OpenAI & Anthropic-compatible proxy for Verdent.ai — single Go binary.

Features • Quick Start • Usage • Configuration • Troubleshooting • Contact

English | Русский

---

Features

• 🚀 Single binary — HTTP server, Verdent client, and dashboard in one Go binary; zero external dependencies
• 🔓 Multi-protocol — OpenAI (/v1/chat/completions), Anthropic (/v1/messages), and Responses API (/v1/responses)
• 🧩 Tool calling — full function calling support for IDE integration; works out of the box with text-based contract
• 🔐 Account pooling — multi-JWT rotation with automatic failover and quota tracking
• 📊 Embedded dashboard — live HTML UI for account management, model selection, and OAuth login
• 🌐 Cross-platform — Windows, macOS, Linux
• 📦 Standalone — pure Go, no CGO dependencies

💡 Works with IDE out of the box! Perfect for AI-assisted development with Verdent models.

---

Quick Start

1. Download

Download the latest release for your platform:

• Windows (64-bit)
• Linux (64-bit)
• macOS Intel
• macOS Apple Silicon

Or build from source:

git clone https://github.com/D3-vin/Verdent2Api.git
cd Verdent2Api
go build -trimpath -ldflags="-s -w" -o verdent ./cmd/server

2. Configure

cp .env.example .env
# Edit .env — set VERDENT_TOKEN (JWT from Verdent desktop app)
# Get token from DevTools → Network → authorization header

3. Run

./verdent

Dashboard: http://localhost:5084

4. Test

curl -X POST http://localhost:5084/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer change-me" \
  -d '{"model":"deepseek-v4-flash-free","messages":[{"role":"user","content":"Hello!"}],"stream":false}'

---

Usage

Streaming

curl -N -X POST http://localhost:5084/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer change-me" \
  -d '{"model":"deepseek-v4-flash-free","stream":true,"messages":[{"role":"user","content":"Say hi"}]}'

Connect from an IDE

Setting	Value
Base URL	http://localhost:5084/v1
API Key	change-me (the AUTH_TOKEN from .env)
Model	deepseek-v4-flash-free, glm-5.3-flash-free, etc.

💡 IDE Integration: This API works perfectly with AI coding assistants! Just add it as a custom OpenAI endpoint.

Models

Free models available without account limits:

• deepseek-v4-flash-free — DeepSeek V4 Flash (fast, efficient)
• glm-5.3-flash-free — GLM-5.3 Flash (balanced)
• And more via /v1/models

Premium models require Verdent JWT:

• deepseek-v4 — DeepSeek V4 (full version)
• glm-4-plus — GLM-4 Plus
• claude-3-5-sonnet-20241022 — Claude 3.5 Sonnet
• Full list at http://localhost:5084/ or GET /v1/models

API Endpoints

Method	Path	Auth	Description
POST	/v1/chat/completions	yes	OpenAI-compatible chat (stream/non-stream, tools)
POST	/v1/messages	yes	Anthropic-compatible chat (Claude Desktop)
POST	/v1/responses	yes	Responses API (extended format)
GET	/v1/models	yes	List available models
POST	/prompt	yes	Legacy: single text prompt
GET	/status	no	Server status

Session headers:

Header	Purpose
X-Session-Id	Conversation ID (default default)
X-Fresh-Session	true — new conversation, empty history

Extra request fields: deepThink, search/webSearch, reasoning.

Tool Calling

Verdent has no native function calling. The proxy implements a text-based contract adapter:

• tools → rewritten as structured prompt → model emits JSON → parsed into tool_calls
• Multiple parsing strategies: JSON code blocks, agent markers, direct JSON
• Streaming support: live tool call interception
• Compatible with OpenAI and Anthropic formats

Account Management

Web dashboard at http://localhost:5084/:

• Add/remove JWT accounts
• View quota usage per account
• Login via OAuth (PKCE flow)
• Select default model
• View cooldown status

---

Configuration

Settings in .env (auto-loaded) or environment variables:

Variable	Default	Description
VERDENT_TOKEN	(empty)	Verdent JWT (seed for first run)
VERDENT_TOKENS_FILE	verdent_tokens.txt	Multi-account seed file (one JWT per line)
PORT	5084	Server port
HOST	0.0.0.0	Bind address
AUTH_TOKEN	change-me	Bearer token for API access
LOG_LEVEL	info	debug, info, warn, error
TIMEOUT	300000	Request timeout (milliseconds)
VERDENT_PROXY	(empty)	HTTP proxy for Verdent traffic
PERSIST_HISTORY	false	Persist conversation history
CLAUDE_DESKTOP	false	Enable Anthropic-compatible mode

Getting VERDENT_TOKEN:

1. Open Verdent desktop app
2. DevTools (F12) → Network tab
3. Look for requests with authorization header
4. Copy the JWT (starts with eyJ...)
5. Add to .env: VERDENT_TOKEN=eyJ...

Optional CLI flags:

./verdent  # All config from .env

---

Troubleshooting

Symptom	Cause	Fix
No accounts available	Missing or invalid JWT	Add valid VERDENT_TOKEN in .env or via dashboard
401 Unauthorized	Wrong AUTH_TOKEN	Check Bearer token matches AUTH_TOKEN in .env
Empty response	Network/timeout error	Set LOG_LEVEL=debug, check logs
Model not available	Model not in catalog	Check /v1/models for available models
Quota exhausted	All accounts depleted	Wait for reset or add more accounts

---

Project Structure

verdent/
├── cmd/server/          # Entry point (calls app.Main)
├── internal/app/        # Core implementation
│   ├── verdent.go       # Verdent client (AES-GCM, SSE streaming)
│   ├── verdentmeta.go   # Model catalog + quota tracking
│   ├── verdentoauth.go  # OAuth PKCE login
│   ├── verdentpool.go   # JWT account pool + config.json
│   ├── models.go        # Model catalog + aliases + cooldowns
│   ├── main.go          # HTTP server, OpenAI routes
│   ├── anthropic.go     # Anthropic API compatibility
│   ├── responses.go     # Responses API
│   ├── tools.go         # Tool calling adapter
│   ├── toolstream.go    # Streaming tool call parser
│   └── dashboard.go     # Embedded dashboard + API
├── .env.example         # Configuration template
├── go.mod               # Go module (no external deps)
└── BUILD.md             # Build instructions

---

Contact

• GitHub: https://github.com/D3-vin/Verdent2Api
• Telegram: @D3_vin
• Author: @D3vin_dev

---

License

Provided as-is for educational and interoperability purposes. Use responsibly and in accordance with Verdent's terms of service.