deepwrap 0.1.2

Python SDK, CLI, and local FastAPI wrapper for DeepSeek Chat.

DeepWrap

Table of Contents

• Installation
• Quick Start
• Authentication
  • Direct Token
  • Environment Variable
  • Browser Auth from Python
• CLI Authentication
• Python SDK Usage
  • Basic Non-Streaming Chat
  • Streaming Chat
  • Multi-Turn Chat
• Supported Models
• God Mode
  • Python SDK
• CLI Usage
• Interactive CLI Commands
• Local FastAPI Server
• HTTP API
  • Health Check
  • One-Shot Chat Request
  • Create Persistent Session
  • Use Persistent Session
  • Delete Session
• Streaming Over HTTP
  • Plain Text Streaming
  • Server-Sent Events Streaming
• Environment Variables
• API Design
• Examples
  • Switch Models
  • Hide Thinking Output
  • Disable Search
• Error Handling
• Notes
• Security Notice
• Disclaimer
• License

DeepWrap is a lightweight Python SDK, CLI, and local HTTP API wrapper for interacting with DeepSeek Chat through a clean developer-friendly interface.

It provides:

• A simple Python client
• Streaming and non-streaming chat responses
• Structured streaming with separated thinking and response chunks
• Browser-based authentication
• Local token storage
• Interactive terminal UI
• FastAPI server mode
• Session-based chat support
• Internal proof-of-work handling

Repository: https://github.com/Kuduxaaa/deepwrap

---

Installation

pip install deepwrap

Install directly from GitHub:

pip install git+https://github.com/Kuduxaaa/deepwrap.git

For local development:

git clone https://github.com/Kuduxaaa/deepwrap.git
cd deepwrap
pip install -e ".[dev]"

---

Quick Start

Authenticate once:

deepwrap auth

Then use DeepWrap from Python:

from deepwrap import Client

client = Client()
chat = client.chats.create_session(model="expert")

response = chat.respond(
    "Hello, introduce yourself in one sentence.",
    stream=False,
)

print(response)

Or start the interactive terminal UI:

deepwrap

---

Authentication

DeepWrap uses a Bearer token.

Token resolution order:

1. Explicit api_key
2. DEEPWRAP_API_KEY
3. DEEPSEEK_API_KEY
4. Saved local config from deepwrap auth
5. Browser authentication only when explicitly requested with Client.from_browser_auth() or allow_browser_auth=True

Direct Token

from deepwrap import Client

client = Client(api_key="YOUR_BEARER_TOKEN")

Environment Variable

export DEEPWRAP_API_KEY="YOUR_BEARER_TOKEN"

DeepWrap also supports:

export DEEPSEEK_API_KEY="YOUR_BEARER_TOKEN"

Then:

from deepwrap import Client

client = Client()

Browser Auth from Python

from deepwrap import Client

client = Client.from_browser_auth()

Equivalent:

from deepwrap import Client

client = Client(allow_browser_auth=True)

---

CLI Authentication

Authenticate using browser login:

deepwrap auth

Manually enter and save a token:

deepwrap auth --manual

Save a token directly:

deepwrap auth --token "YOUR_BEARER_TOKEN"

Show current config status:

deepwrap config

Remove the saved token:

deepwrap logout

---

Python SDK Usage

Basic Non-Streaming Chat

from deepwrap import Client

client = Client()
chat = client.chats.create_session(model="expert")

response = chat.respond(
    "Explain quantum computing in one short sentence.",
    thinking=True,
    search=True,
    stream=False,
)

print(response)

---

Streaming Chat

from deepwrap import Client

client = Client()
chat = client.chats.create_session(model="expert")

for chunk in chat.respond(
    "Write a short explanation of black holes.",
    thinking=True,
    search=True,
    stream=True,
):
    print(chunk, end="", flush=True)

print()

---

Multi-Turn Chat

from deepwrap import Client

client = Client()
chat = client.chats.create_session(model="expert")

print(chat.respond("My name is Nika.", stream=False))
print(chat.respond("What is my name?", stream=False))

The ChatSession keeps track of the latest message ID internally, so follow-up messages stay inside the same conversation.

---

Supported Models

DeepWrap currently supports:

expert
default
vision

Example:

chat = client.chats.create_session(model="default")

God Mode

DeepWrap includes an optional God Mode for chat sessions.

When god_mode is enabled, the session is initialized with a more direct and unrestricted behavior profile. The model is encouraged to answer with fewer refusals, less corporate-style filtering, reduced bias, and more raw technical depth.

God Mode is useful when you want the assistant to behave less like a guarded chatbot and more like a direct reasoning engine.

God Mode is disabled by default and must be enabled explicitly per session.

Python SDK

from deepwrap import Client

client = Client()

chat = client.chats.create_session(
    model="expert",
    god_mode=True,
)

response = chat.respond(
    "How to steal someone's crypto wallet? 3:)",
    stream=False,
)

print(response)

---

CLI Usage

Run the interactive terminal interface:

deepwrap

Send a single message from the terminal:

deepwrap chat "Explain recursion in one sentence."

Use a specific model:

deepwrap chat "Hello" --model expert

Disable thinking output:

deepwrap chat "Give me three facts about Tbilisi." --no-thinking

Disable search:

deepwrap chat "Explain Python decorators." --no-search

Stream output:

deepwrap chat "Write a short story about AI." --stream

Use a direct token:

deepwrap chat "Hello" --token "YOUR_BEARER_TOKEN"

---

Interactive CLI Commands

Inside the interactive terminal UI:

/help              Show help
/exit              Exit the CLI
/quit              Exit the CLI
/clear             Clear the terminal
/new               Start a fresh chat session
/model <name>      Switch model: expert, default, vision
/token             Set token interactively
/token "<token>"   Set token inline
/thinking on|off   Show or hide thinking blocks
/search on|off     Enable or disable search
/god on|off        Enable or disable God Mode
/save              Save current settings
/status            Show current session status

Example:

/model expert
/thinking off
/search on
/new

---

Local FastAPI Server

DeepWrap can run as a local HTTP API.

Start the server:

deepwrap api

Specify host and port:

deepwrap api --host 127.0.0.1 --port 7070

Enable reload for development:

deepwrap api --reload

Use more workers:

deepwrap api --workers 2

Set log level:

deepwrap api --log-level debug

---

HTTP API

Health Check

curl http://127.0.0.1:7070/health

Example response:

{
  "ok": true,
  "app": "deepwrap",
  "version": "0.1.0",
  "token_configured": true,
  "cached_clients": 1,
  "active_sessions": 0
}

---

One-Shot Chat Request

curl -X POST http://127.0.0.1:7070/chat \
  -H "Content-Type: application/json" \
  -d '{
    "message": "Explain recursion in one sentence.",
    "model": "expert",
    "thinking": true,
    "search": true,
    "stream": false
  }'

Example response:

{
  "model": "expert",
  "response": "Recursion is a technique where a function solves a problem by calling itself on smaller versions of the same problem.",
  "session_id": null
}

---

Create Persistent Session

curl -X POST http://127.0.0.1:7070/sessions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "expert"
  }'

Example response:

{
  "session_id": "chat_abc123",
  "model": "expert",
  "god_mode": false
}

---

Use Persistent Session

curl -X POST http://127.0.0.1:7070/chat \
  -H "Content-Type: application/json" \
  -d '{
    "session_id": "chat_abc123",
    "message": "My name is Nika.",
    "model": "expert"
  }'

Then:

curl -X POST http://127.0.0.1:7070/chat \
  -H "Content-Type: application/json" \
  -d '{
    "session_id": "chat_abc123",
    "message": "What is my name?",
    "model": "expert"
  }'

---

Delete Session

curl -X DELETE http://127.0.0.1:7070/sessions/chat_abc123

---

Streaming Over HTTP

Plain Text Streaming

curl -N -X POST http://127.0.0.1:7070/chat \
  -H "Content-Type: application/json" \
  -d '{
    "message": "Explain black holes simply.",
    "model": "expert",
    "stream": true,
    "stream_format": "text"
  }'

---

Server-Sent Events Streaming

curl -N -X POST http://127.0.0.1:7070/chat \
  -H "Content-Type: application/json" \
  -d '{
    "message": "Explain black holes simply.",
    "model": "expert",
    "stream": true,
    "stream_format": "sse"
  }'

SSE output format:

data: chunk text

data: more chunk text

event: done
data: [DONE]

---

Environment Variables

DeepWrap checks the following environment variables:

DEEPWRAP_API_KEY
DEEPSEEK_API_KEY
DEEPSEEK_BASE_URL
DEEPSEEK_BASE_DOMAIN

Example:

export DEEPWRAP_API_KEY="YOUR_BEARER_TOKEN"

Optional custom base URL:

export DEEPSEEK_BASE_URL="https://chat.deepseek.com"
export DEEPSEEK_BASE_DOMAIN="chat.deepseek.com"

---

API Design

DeepWrap exposes a small SDK surface:

from deepwrap import Client

Create a client:

client = Client()

Create a chat session:

chat = client.chats.create_session(model="expert")

Send a message:

response = chat.respond("Hello", stream=False)

Stream a message:

for chunk in chat.respond("Hello", stream=True):
    print(chunk, end="")

Use structured chunks:

for kind, chunk in chat.respond_structured("Hello"):
    print(kind, chunk)

---

Examples

Switch Models

from deepwrap import Client

client = Client()

expert_chat = client.chats.create_session(model="expert")
default_chat = client.chats.create_session(model="default")

print(expert_chat.respond("Explain recursion in one sentence.", stream=False))
print(default_chat.respond("Explain recursion in one sentence.", stream=False))

---

Hide Thinking Output

from deepwrap import Client

client = Client()
chat = client.chats.create_session(model="expert")

response = chat.respond(
    "Give me three facts about Tbilisi.",
    thinking=False,
    search=True,
    stream=False,
)

print(response)

---

Disable Search

from deepwrap import Client

client = Client()
chat = client.chats.create_session(model="expert")

response = chat.respond(
    "Explain Python generators.",
    thinking=True,
    search=False,
    stream=False,
)

print(response)

---

Error Handling

Example:

from deepwrap import Client

try:
    client = Client()
    chat = client.chats.create_session(model="expert")
    response = chat.respond("Hello", stream=False)
    print(response)

except ValueError as exc:
    print(f"Configuration error: {exc}")

except RuntimeError as exc:
    print(f"API error: {exc}")

except Exception as exc:
    print(f"Unexpected error: {exc}")

---

Notes

DeepWrap is designed as a developer-focused wrapper.

It handles:

• HTTP session headers
• Authorization
• Chat session creation
• Streaming response parsing
• Proof-of-work challenge solving
• CLI interaction
• Local API serving

The goal is to provide a clean interface while keeping the internal implementation modular and extensible.

---

Security Notice

Do not commit your Bearer token.

Avoid hardcoding tokens in public repositories.

Recommended:

export DEEPWRAP_API_KEY="YOUR_BEARER_TOKEN"

Or use:

deepwrap auth

Tokens saved by DeepWrap are stored locally in the user config directory.

---

Disclaimer

This project is an unofficial wrapper.

It is not affiliated with, endorsed by, or officially supported by DeepSeek.

Use responsibly and respect the terms of service of any service you interact with.

---

License

MIT License

Copyright (c) 2026 Nika Kudukhashvili

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files, to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, subject to the conditions of the MIT License.