• ✨ AI Chat Conversation
  1. Rich Format Compatibility
     • Supports Vision models, including direct image upload and image direct link or Base64 image input functions (like GPT-4 Vision Preview, Gemini Pro Vision, etc.)
     • Supports drawing with DALL-E models
     • Supports Imagine / Upscale / Variant / Reroll operations for Midjourney / Niji models
  2. Markdown Support / Theme Switching Support, supports light and dark modes, code highlighting, Mermaid, LaTeX formulas, tables, progress bars, Virtual Message, etc.
  3. Support for Message Menu, including re-answering, copying messages, using messages, editing messages, deleting messages, saving as a file, and more operations...
  4. Support for Multi-platform Adaptation, supports PWA apps, desktop platforms (desktop is based on Tauri).
  5. Dialogue Memory, cloud synchronization, native support for direct link sharing of site conversations, supports using shared conversations, saving shared conversations as images, and share management (including viewing, deleting shares, etc.).
  6. Native Support for Full Model File Parsing, supports pdf, docx, pptx, xlsx, images, and other formats parsing (for more details, see the project chatnio-blob-service).
  7. Supports Full-model DuckDuckGo Online Search (for details, refer to the duckduckgo-api project, needs to be set up on your own and configured in the internet settings in the system settings, thanks to the author @binjie09, enable web searching by prefixing the relay api model with web-.)
  8. Full-screen Large Text Editing, supports plain text editing, edit-preview mode, pure preview mode three mode switching.
  9. Model Marketplace, supports model search, supports drag-and-drop in sequence, includes model name, model description, model Tags, model avatar, automatic binding of model pricing, automatic binding of subscription quotas (models included in the subscription will be tagged with plus label)
  10. Support for Preset, supports custom presets and cloud synchronization features, supports preset cloning, preset avatar settings, preset introduction settings
  11. Support for Site Announcements Supports site announcements and notifications
  12. Support for Preference Settings, i18n multi-language support, custom maximum number of carry-over sessions, maximum reply tokens number, model parameters customization, reset settings, etc.
  13. Internationalization Support, support multi-language switching
  14. Additional (User group permissions for additional functions can be enabled and disabled through backend system settings)
  • [Discontinued] 🍎 AI project generator function, supports viewing of the generation process, supports TAR / ZIP format downloads (based on presets, may be unstable)
  • [Discontinued] 📂 Bulk article generation function, supports progress bar, one-click generation of DOCX documents in TAR / ZIP format download (requires a generation quantity higher than the highest concurrency number of the upstream model)
  • [Deprecated] 🥪 AI Card feature (deprecated), AI questions and answers are presented in the form of cards, can be directly embedded via image url. (Based on dynamic SVG generation)
• 🔔 Rich Admin System
  1. Rich and beautiful dashboard, including current day and month's crediting information, subscribers, model usage statistics line charts, pie charts, revenue statistics, user type statistics, model usage statistics, request counts and model error counts statistics charts, etc.
  2. Support user management, User list, User details, Administrative operations (Change password, Change mailbox, Block/Unblock user, Set as administrator, Points change, Points setup, Subscription management, Subscription level setup, Release subscription, etc).
  3. Support Gift Code and Redemption Code Management Support management operations, support batch generation and save to file.
  4. Price setting, support model price setting (Counts billing, Token flexible billing, No billing etc.), support synchronize the price setting of upstream Chat Nio site (optional whether to override the price rules of the existing models on this site), unset price model detection (if non-administrator will automatically detect and stop using the model to prevent loss of money). to prevent loss of money) !
  5. Subscription Settings, different from flexible billing, subscription is a kind of fixed price per time billing, platform users can subscribe to a fixed price through the package, support whether to enable subscription (default off), support subscription tiering, support subscription quota settings, support subscription quotas cover the model settings, icon settings, quotas imported from other packages and other functions. !
  6. Customize Model Marketplace, Edit Frontend Model Marketplace Model Name, Description, Tags, Avatar (Built-in Model Image Selection and Custom Model Image Setting), Whether or not to add Model Models and other information!
  7. System Settings, Customized Site Name, Site Logo, Documentation Link, Whether to Suspend Registration, User Initial Points Settings, Customized Purchase Link (Card Address), Contact Information, Footer Information, etc. !
  8. SMTP support, support whether to enable email suffix whitelisting, support customized email suffix whitelisting.
  9. Support model caching, i.e., under the same entry, if it has been requested before, it will return the cached result directly (hit caching will not be billed), to reduce the number of requests. You can customize the maximum number of cached results for a case (default is 1), customize the models that can be cached (default is empty), customize the caching time (default is 1 hour), support one-click settings All models are not cached, Free models are cached, All models are cached, and other operations.
• ⚡ Channel Management System
  1. Chat Nio uses a self-developed channel distribution algorithm (doesn't rely on http context), abstracts Adapter compatibility layer architecture, low coupling, and high scalability.
  2. Supports Multi-channel Management, supports priority allocation, weight load, channel status management, etc. (Priority refers to the allocation of channel priority during the model request process, where higher priority channels are used first. If an error occurs, the system will automatically fall back to a channel of lower priority; weight refers to the weight of a channel within the same priority, with higher weight increasing the probability of being used. For channels at the same priority level, only one can be used at any time, and the higher the weight, the higher the probability of being selected.)
  3. Compatible With Multiple Formats, supports multiple model compatibility layers, see the model support section below for details.
  4. Supports Custom Models, allows for the use of known models through adding models, supports the addition of custom models, supports one-click filling of template models (referring to the template models supported by the current format by default, such as OpenAI format template models like gpt-3.5-turbo-0613), supports one-click clearing of models.
  5. Supports Channel Retries, supports a Retry mechanism for channels, allows for custom retry counts, after surpassing the retry count the system will automatically fall.
  6. Supports Balanced Load Within The Same Channel, a single channel can be configured with multiple keys instead of creating channels in bulk (multiple keys separated by line breaks), assigning requests with equal weight randomly, and the Retry mechanism will also be used in conjunction with multiple keys within the same channel, randomly selecting keys for retries.
  7. Supports Channel Model Mapping, maps models to the models already supported by this channel, format as target model>existing model, adding a prefix ! allows existing models not to be allocated in the channel's hit models, for the specific usage method, please refer to the instructions and tips in the channel settings within the program.
  8. Supports User Grouping, custom selection of user groups that can use this model (such as anonymous users, regular users, basic subscription users, standard subscription users, professional subscription users, etc., setting to group 0 available and setting to all groups available are the same effect).
  9. Built-in Upstream Hiding, automatically hides the upstream URL set within the channel when an error occurs (e.g., channel:https://2/v1/chat/completions), also supports hiding keys (Gemini, you're really good job), to prevent upstream channels from being misused in cases where keys are not set or upstream error information exposes the full key (such as reverse-type channels), and also facilitates troubleshooting in cases where multiple channels serve the same access point simultaneously.
• ✨ Forwarding API Services
  1. Compatible with multiple formats in the OpenAI universal format and supporting multiple model compatible layers, this means you can use a format while supporting multiple AI models.
  2. Replace https://api.openai.com with https://api.chatnio.net (example), enter the API Key from API Settings in the console to use, support for resetting Key.
  3. Supported Formats
  • Chat Completions (/v1/chat/completions)
  • Image Generation (/v1/images)
  • Model List (/v1/models)
  • Dashboard Billing (/v1/billing)
• 🎃 More features waiting for your discovery...

• OpenAI
  • Chat Completions (support vision, tools_calling and function_calling)
  • Image Generation
• Azure OpenAI
• Anthropic Claude (support vision)
• Slack Claude (deprecated)
• Sparkdesk (support function_calling)
• Google Gemini (PaLM2)
• New Bing (creative, balanced, precise)
• ChatGLM (turbo, pro, std, lite)
• DashScope Tongyi (plus, turbo)
• Midjourney
  • Mode Toggling (relax, fast, turbo)
  • Support U/V/R Actions
• Tencent Hunyuan
• Baichuan AI
• Moonshot AI
• Groq Cloud AI
• ByteDance Skylark (support function_calling)
• 360 GPT
• LocalAI (Stable Diffusion, RWKV, LLaMa 2, Baichuan 7b, Mixtral, ...) *requires local deployment

Once deployment is successful, the administrator account is root, the default password is chatnio123456

1. ⚡ Install with Docker Compose (Recommended)

   The host mapping address is https://localhost:8000 after successful operation.

   git clone --depth=1 --branch=main --single-branch https://github.com/Deeptrain-Community/chatnio.git
   cd chatnio
   docker-compose up -d # Running the service
   # To use the stable version, replace with `docker-compose -f docker-compose.stable.yaml up -d`.
   # To enable automatic updates with Watchtower, replace with `docker-compose -f docker-compose.watch.yaml up -d`.

   Version Update (With Watchtower auto-updating enabled, there is no need for manual updates):

   docker-compose down 
   docker-compose pull
   docker-compose up -d

   • Mount directory for MySQL database: ~/db
   • Mount directory for Redis database: ~/redis
   • Mount directory for configuration files: ~/config

2. ⚡ Docker installation (when running lightweight, it is often used in external Mysql/RDS service)

   If you need to use the stable version, please use programzmh/chatnio:stable instead of programzmh/chatnio:latest.

   docker run -d --name chatnio \
     --network host \
     -p 8000:8094 \
     -v ~/config:/config \
     -v ~/logs:/logs \
     -v ~/storage:/storage \
     -e MYSQL_HOST=localhost \
     -e MYSQL_PORT=3306 \
     -e MYSQL_DB=chatnio \
     -e MYSQL_USER=root \
     -e MYSQL_PASSWORD=chatnio123456 \
     -e REDIS_HOST=localhost \
     -e REDIS_PORT=6379 \
     -e SECRET=secret \
     -e SERVE_STATIC=true \
     programzmh/chatnio:latest

   • --network host Assigns the container to use the host network, allowing it to access the host's network. You can modify this as needed.
   • -p 8000:8094 Maps host port 8000 to 8094, feel free to modify the port number before the colon.
   • SECRET: JWT Secret Key, generate a random string to modify
   • SERVE_STATIC: Whether to enable static file service (normally there is no need to change this item, see below for answers to frequently asked questions)
   • -v ~/config:/config mount host machine directory of configuration file, -v ~/logs:/logs mount host machine directory of log file,-v ~/storage:/storage mount generated files of additional functions
   • You need to configure MySQL and Redis services, please refer to the information above to modify the environment variables yourself.

   Version Update (With Watchtower auto-updating enabled, there is no need for manual updates, Just run through the steps above again after execution.)：

   docker stop chatnio
   docker rm chatnio
   docker pull programzmh/chatnio:latest

3. ⚒ Custom Build and Install (Highly Customizable)

   After deployment, the default port is 8094, and the access address is https://localhost:8094. Configuration options (~/config/config.yaml) can be overridden using environment variables, such as the MYSQL_HOST environment variable can override the mysql.host configuration item.

   git clone https://github.com/Deeptrain-Community/chatnio.git
   cd chatnio
   
   cd app
   npm install -g pnpm
   pnpm install
   pnpm build
   
   cd ..
   go build -o chatnio
   
   nohup ./chatnio > output.log & # using nohup to run in background

1. Why does my deployed site have access to pages, can sign in, but can't use chat (keeps going in circles)?

   • Chat and other such features communicate through websockets; please ensure that your server supports websockets. (Tip: Transmission through HTTP can be achieved, without the need for websocket support)
   • If you are using reverse proxies such as Nginx, Apache, please make sure you have configured websocket support.
   • If you are using services like port mapping, port forwarding, CDN, API Gateway, etc., ensure that your service supports and has WebSocket enabled.

2. My Midjourney Proxy-formatted channel keeps spinning or showing an error message please provide available notify url

   • If it's going around in circles, make sure your Midjourney Proxy service is up and running and that you've configured the correct upstream address.
   • Midjourney To fill in the channel type, use Midjourney instead of OpenAI (IDK why a lot of people fill in the OpenAI type format and then come back with feedback on why it's an empty response, except for the mj-chat class).
   • After resolving these issues, please check if the backend domain is properly configured in your system settings. If it's not set up, it may prevent the Midjourney Proxy service from functioning correctly.

3. What external dependencies does this project have?

   • MySQL: Store persistent data such as user information, chat records, and administrator details.
   • Redis: Storing user quick authentication information, IP rate limits, subscription quotas, email verification codes, etc.
   • If the environment is not properly set up, it may cause the service to fail. Please ensure that your MySQL and Redis services are running smoothly (for Docker deployment, you'll need to set up external services manually).

4. My server is ARM architecture, does this project support ARM architecture?

   • Yes. The Chat Nio project uses BuildX to build multi-arch images, so you can run it directly with docker-compose or docker without any additional configuration.
   • If you are using a build installation, you can compile directly on the ARM machine without any additional configuration. If you are compiling on an x86 machine, please use GOARCH=arm64 go build -o chatnio for cross-compilation and upload it to the ARM machine for execution.

5. How do I change the default root password?

   • Please click on the avatar in the top right corner or the user box at the bottom of the sidebar to go to the backend management, click on the Modify Root Password under the Operation Bar of the General Settings in the System Settings. Alternatively, you can choose to modify the root user's password in the User Management.

6. What is the backend domain in the System Settings?

   • The backend domain refers to the address of the backend API service, which is usually the address of the site followed by /api, such as https://example.com/api.
   • If set to non-SERVE_STATIC mode, enabling frontend and backend separation deployment, please set the backend domain to the address of your backend API service, such as https://api.example.com.
   • The backend domain is used here for the backend callback address of the Midjourney Proxy service; if you do not need to use the Midjourney Proxy service, please ignore this setting.

7. How do I configure the payment method?

   • The Chat Nio open-source version supports the card issuance model; simply set the Purchase Link in the System Settings to your card issuance address. Card codes can be generated in bulk through the Gift Code Management in User Management.

8. What is the difference between gift codes and redemption codes?

   • A gift code can only be bound once by a single user, and it is not an aff code or a way to distribute benefits; it can be redeemed from the Gift Code option under the avatar dropdown menu.
   • A redemption code can be bound by multiple users and can be used for normal purchases and card issuance; it can be generated in bulk through the Redemption Code Management in User Management and redeemed by entering the code in the Points option (the first item in the menu) under the avatar dropdown menu.
   • For example: If I distribute a welfare item with the type Happy New Year, it is recommended to use gift codes; assuming 100 codes with 66 points each, if they were redemption codes, a quick user could use up all 6600 points, but gift codes ensure each user can use them only once (receiving 66 points).
   • When setting up card issuance, if gift codes are used, the limitation of using once per type may lead to redemption failures when purchasing multiple gift codes, but redemption codes can be used in this scenario.

9. Does the project support deployment on Vercel?

   • Chat Nio itself does not directly support Vercel deployment, but you can use a client-server separation mode where you deploy the frontend on Vercel and the backend with Docker or compiled deployment.

10. What is the client-server separation deployment mode?

• Normally, the frontend and backend are in the same service, with the backend accessible at /api. In client-server separation deployment, the frontend and backend are deployed on separate services: the frontend as a static file service and the backend as an API service.
  • For example, the frontend is deployed using Nginx (or Vercel), with a domain like https://www.chatnio.net.
  • The backend is deployed with Docker, with a domain like https://api.chatnio.net.
• You need to manually package the frontend and set the environment variable VITE_BACKEND_ENDPOINT to your backend URL, such as https://api.chatnio.net.
• To configure the backend, set SERVE_STATIC=false to prevent it from serving static files.

11. Flexible Billing and Subscriptions Explained

• Billing on Demand (Points), represented by a cloud icon, is a universal pricing model where 10 points are fixed at 1 yuan (CNY). Custom exchange rates can be set in the billing rules' embedded templates.
• Subscriptions are fixed-price plans with per-item quotas. Users must have sufficient points to subscribe (e.g., if a user wants to subscribe to a plan for 32 yuan, they need at least 320 points). Subscriptions are a combination of Items, each defining covered models, quotas (-1 for unlimited), names, IDs, icons, and more. Managed in the subscription management section, you can enable subscriptions, set prices, edit Items per subscription level, and import Items from other levels.
• Subscriptions support three predefined tiers: Regular Users (0) , Basic Subscription (1) , Standard Subscription (2) , Professional Subscription (3). Subscription levels serve as user groups, which can be configured in channel management, selecting which user groups can access these models.
• Quotas for subscriptions can be managed, including whether to allow API forwarding (default off).

12. Minimum Point Request Detection for User Quota Insufficient

• To prevent misuse, if a request for points falls below the minimum threshold, an error message indicating insufficient points will be returned. Requests above or equal to the minimum are processed normally.
• Minimum request point rules for models:
  • Free models have no limit.
  • Per-use models require at least one request point (e.g., if a model costs 0.1 points per use, the minimum is 0.1 points).
  • Token-based elastic billing models have a minimum of 1K input token price plus 1K output token price (e.g., if 1K input tokens cost 0.05 points and 1K output tokens cost 0.1 points, the minimum is 0.15 points).

13. Notes for setting up DuckDuckGo API
    • First of all, thanks to the author Binjie for the duckduckgo-api project, which provides the connected search function (prompt implementation) for Chat Nio.
    • The DDG API service needs to be set up by yourself. The default site of Binjie author often runs out of quota. Please set up by yourself and set it in the network settings in the system settings.
    • DuckDuckGo cannot be used in China's network environment, please use a proxy or overseas server to build DDG API endpoints.
    • After successful deployment, please test https://your.domain/search?q=hi to simply test whether the setup is successful. If it cannot be accessed, please check whether your DDG API service is running properly or seek help from the original project.
    • After successful deployment, please go to the network settings in the system settings, set your DDG API endpoint address (do not add suffix /search), the maximum number of results is 5 by default (results set to 0 or negative numbers default to 5)
    • You can use connected search normally after enabling it in the chat. If it still cannot be used, there is usually a problem with the model (such as GPT-3.5 sometimes fails to understand).
    • This connected search is implemented through preset, which means it can guarantee the universal function that all models can support. Compatibility cannot guarantee sensitivity. It doesn't rely on model Function Calling. Other models that support connection can choose to turn off this function directly.
14. Why can't my GPT-4-All and other reverse models use the images in uploaded files?
    • Uploaded model images are in Base64 format. If the reverse engine does not support Base64 format, use a URL direct link instead of uploading the file.
15. How to start strict cross-origin domain inspection?
    • Normally, the backend is open to cross-origin requests for all domains. There is no need to enable strict cross-origin detection if there is no specific requirement.
    • If strict cross-origin detection needs to be enabled, it can be configured in the backend environment variable and ALLOW_ORIGINS, such as ALLOW_ORIGINS=chatnio.net,chatnio.app (no protocol prefix is required, and www parsing does not need to be added manually, The backend will automatically identify and allow cross-domain). In this way, strict cross-origin detection will be supported (such as https://www.chatnio.app, https://chatnio.net will be allowed, and other domains will be rejected).
    • Even when strict cross-origin detection is enabled, /v1 interface will still allow cross-origin requests from all domains to ensure normal use of the transfer API.
16. How is the model mapping feature used?
    • Model mappings within a channel are in the format [from]>[to], with line breaks between mappings, from is the model requested, to is the model actually sent upstream and needs to be actually supported by the upstream.
    • e.g. I have a reverse channel, fill in gpt-4-all>gpt-4, then when my user requests a gpt-4-all model from that channel, the backend will model map to gpt-4 to request gpt-4 from that channel, which supports 2 models, gpt-4 and gpt-4-all (essentially both gpt-4 and gpt-4-all). Both are essentially gpt-4).
    • If I don't want my reverse channel to affect the gpt-4 channel group, I can prefix it with !gpt-4-all>gpt-4, and the channel gpt-4 will be ignored, and the channel will only support 1 model, gpt-4-all (but is essentially gpt-4).

• Frontend: React + Radix UI + Tailwind CSS + Shadcn + Tremor + Redux
• Backend: Golang + Gin + Redis + MySQL
• Application Technology: PWA + WebSocket

APIs are divided into Gateway APIs and exclusive features of Chat Nio.

The transfer API is a general format for OpenAI, supporting various formats. For details, see the OpenAI API documentation and SDKs.

The following SDKs are unique features of Chat Nio API SDKs

• JavaScript SDK
• Python SDK
• Golang SDK
• Java SDK (Thanks to @hujiayucc)
• PHP SDK (Thanks to @hujiayucc)

Frontend projects here refer to projects that focus on user chat interfaces, backend projects refer to projects that focus on API transfer and management, and one-stop projects refer to projects that include user chat interfaces and API transfer and management

• Next Chat @yidadaa (Front-end Oriented Projects)
• Lobe Chat @arvinxx (Front-end Oriented Projects)
• Chat Box @bin-huang (Front-end Oriented Projects)
• OpenAI Forward @kenyony (Back-end Oriented Projects)
• One API @justsong (Back-end Oriented Projects)
• New API @calon (Back-end Oriented Projects)
• FastGPT @labring (Knowledge Base)
• Quivr @quivrhq (Knowledge Base)
• Bingo @weaigc (Knowledge Base)
• Midjourney Proxy @novicezk (Model Library)

Apache License 2.0

@Sh1n3zZ @4EvEr

• LightXi Provide Font CDN support
• BootCDN & Static File Provide Resources CDN support

Chat Nio leans towards a one-stop service, integrating user chat interface and API intermediary and management projects.

• Compared to projects like NextChat which are front-end and lightweight deployment-oriented, Chat Nio's advantages include more convenient cloud synchronization, account management, richer sharing functionalities, as well as a billing management system.
• Compared to projects like OneAPI which are backend and lightweight deployment-oriented, Chat Nio's advantages include a richer user interface, a more comprehensive channel management system, richer user management, and the introduction of a subscription management system aimed at the user interface.

The advantage of a one-stop service is that users can perform all operations on one site without the need to frequently switch sites, making it more convenient. This includes more convenient viewing of one's points, the consumption of points by messages, and subscription quotas. While using the chat interface, it also opens up intermediary APIs and Chat Nio's unique functional APIs.

At the same time, we strive to achieve Chat Nio > Next Chat + One API, realizing richer functionalities and more detailed experiences.

Additionally, since the developers are still in school, the development progress of Chat Nio may be affected. If we consider this issue to be non-essential, we will postpone dealing with it or choose to close it directly, not accepting any form of urging. We highly welcome PR contributions, and we are very thankful for your contribution.