Update command: `docker run --rm -v /var/run/docker.sock:/var/run/docker.sock containrrr/watchtower -cR`
The first `3000` in `-p 3000:3000` is the port of the host, which can be modified as needed.
Data will be saved in the `/home/ubuntu/data/one-api` directory on the host. Ensure that the directory exists and has write permissions, or change it to a suitable directory.
Nginx reference configuration:
```
server{
server_name openai.justsong.cn; # Modify your domain name accordingly
location / {
client_max_body_size 64m;
proxy_http_version 1.1;
proxy_pass http://localhost:3000; # Modify your port accordingly
3. Access [http://localhost:3000/](http://localhost:3000/) and log in. The initial account username is `root` and password is `123456`.
For more detailed deployment tutorials, please refer to [this page](https://iamazing.cn/page/how-to-deploy-a-website).
### Multi-machine Deployment
1. Set the same `SESSION_SECRET` for all servers.
2. Set `SQL_DSN` and use MySQL instead of SQLite. All servers should connect to the same database.
3. Set the `NODE_TYPE` for all non-master nodes to `slave`.
4. Set `SYNC_FREQUENCY` for servers to periodically sync configurations from the database.
5. Non-master nodes can optionally set `FRONTEND_BASE_URL` to redirect page requests to the master server.
6. Install Redis separately on non-master nodes, and configure `REDIS_CONN_STRING` so that the database can be accessed with zero latency when the cache has not expired.
7. If the main server also has high latency accessing the database, Redis must be enabled and `SYNC_FREQUENCY` must be set to periodically sync configurations from the database.
Please refer to the [environment variables](#environment-variables) section for details on using environment variables.
### Deployment on Control Panels (e.g., Baota)
Refer to [#175](https://github.com/songquanpeng/one-api/issues/175) for detailed instructions.
If you encounter a blank page after deployment, refer to [#97](https://github.com/songquanpeng/one-api/issues/97) for possible solutions.
3. Create a new project. In Service -> Add Service, select Marketplace, and choose MySQL. Note down the connection parameters (username, password, address, and port).
4. Copy the connection parameters and run ```create database `one-api` ``` to create the database.
5. Then, in Service -> Add Service, select Git (authorization is required for the first use) and choose your forked repository.
6. Automatic deployment will start, but please cancel it for now. Go to the Variable tab, add a `PORT` with a value of `3000`, and then add a `SQL_DSN` with a value of `<username>:<password>@tcp(<addr>:<port>)/one-api`. Save the changes. Please note that if `SQL_DSN` is not set, data will not be persisted, and the data will be lost after redeployment.
7. Select Redeploy.
8. In the Domains tab, select a suitable domain name prefix, such as "my-one-api". The final domain name will be "my-one-api.zeabur.app". You can also CNAME your own domain name.
9. Wait for the deployment to complete, and click on the generated domain name to access One API.
</div>
</details>
## Configuration
The system is ready to use out of the box.
You can configure it by setting environment variables or command line parameters.
After the system starts, log in as the `root` user to further configure the system.
## Usage
Add your API Key on the `Channels` page, and then add an access token on the `Tokens` page.
You can then use your access token to access One API. The usage is consistent with the [OpenAI API](https://platform.openai.com/docs/api-reference/introduction).
In places where the OpenAI API is used, remember to set the API Base to your One API deployment address, for example: `https://openai.justsong.cn`. The API Key should be the token generated in One API.
Note that the specific API Base format depends on the client you are using.
```mermaid
graph LR
A(User)
A --->|Request| B(One API)
B -->|Relay Request| C(OpenAI)
B -->|Relay Request| D(Azure)
B -->|Relay Request| E(Other downstream channels)
```
To specify which channel to use for the current request, you can add the channel ID after the token, for example: `Authorization: Bearer ONE_API_KEY-CHANNEL_ID`.
Note that the token needs to be created by an administrator to specify the channel ID.
If the channel ID is not provided, load balancing will be used to distribute the requests to multiple channels.
### Environment Variables
1.`REDIS_CONN_STRING`: When set, Redis will be used as the storage for request rate limiting instead of memory.
6.`SYNC_FREQUENCY`: When set, the system will periodically sync configurations from the database, with the unit in seconds. If not set, no sync will happen.
10.`POLLING_INTERVAL`: The time interval (in seconds) between requests when updating channel balances and testing channel availability. Default is no interval.
1. What is quota? How is it calculated? Does One API have quota calculation issues?
+ Quota = Group multiplier * Model multiplier * (number of prompt tokens + number of completion tokens * completion multiplier)
+ The completion multiplier is fixed at 1.33 for GPT3.5 and 2 for GPT4, consistent with the official definition.
+ If it is not a stream mode, the official API will return the total number of tokens consumed. However, please note that the consumption multipliers for prompts and completions are different.
2. Why does it prompt "insufficient quota" even though my account balance is sufficient?
+ Please check if your token quota is sufficient. It is separate from the account balance.
+ The token quota is used to set the maximum usage and can be freely set by the user.
3. It says "No available channels" when trying to use a channel. What should I do?
+ Please check the user and channel group settings.
+ Also check the channel model settings.
4. Channel testing reports an error: "invalid character '<' looking for beginning of value"
+ This error occurs when the returned value is not valid JSON but an HTML page.
+ Most likely, the IP of your deployment site or the node of the proxy has been blocked by CloudFlare.
5. ChatGPT Next Web reports an error: "Failed to fetch"
+ Do not set `BASE_URL` during deployment.
+ Double-check that your interface address and API Key are correct.
This project is an open-source project. Please use it in compliance with OpenAI's [Terms of Use](https://openai.com/policies/terms-of-use) and **applicable laws and regulations**. It must not be used for illegal purposes.
According to the MIT license, users should bear the risk and responsibility of using this project, and the developer of this open-source project is not responsible for this.
