Install & accounts

Install, configure accounts and run DanyAPI.

Setup script

One-command install: paste irm https://raw.githubusercontent.com/FANATFANATA/DanyAPI/main/docs/install.ps1 | iex into PowerShell, or curl -fsSL https://raw.githubusercontent.com/FANATFANATA/DanyAPI/main/docs/install.sh | bash on Linux/macOS - it clones the repo and runs docs/setup.py. The script installs dependencies, creates .env from .env.example if missing, asks for the provider credentials and server settings, live-checks the tokens against the provider, writes everything back to .env and creates a DanyAPI launcher shortcut on the desktop. The server auto-updates itself from the latest GitHub release at every launch (DANYAPI_AUTO_UPDATE=0 disables it). Press Enter to keep a current value, type !clear to erase one.

bash

docs\setup.bat      # Windows
                    docs/setup.sh       # Linux/macOS
                    # or directly:
                    python docs/setup.py

Requires Python 3.13+. Passwords are masked during input and written to .env which is gitignored and never committed.

Install

Requires Python 3.13+.

bash

pip install -r requirements.txt

For development (tests + linting) install the dev extras:

bash

pip install -r requirements-dev.txt

Account setup

DeepSeek

Set a pool of tokens (from different accounts), comma-separated, or a single token via DEEPSEEK_TOKEN:

bash

export DEEPSEEK_TOKENS="token1,token2,token3"
                    # or
                    export DEEPSEEK_TOKEN="token1"

Each account can generate one message at a time, so a pool of N tokens gives up to N parallel generations. Grab a token in the browser: DevTools → Application → Local Storage → https://chat.deepseek.com → userToken.

Or a single email + password account (login happens at startup):

bash

export DEEPSEEK_EMAIL="you@example.com"
                    export DEEPSEEK_PASSWORD="secret"

Qwen

Same model, different token location (single token via QWEN_TOKEN):

bash

export QWEN_TOKENS="token1,token2,token3"
                    # or
                    export QWEN_TOKEN="token1"

Grab a token in the browser: DevTools → Application → Local Storage → https://chat.qwen.ai → token.

For Qwen accounts used by DanyAPI, disable the built-in Tools switch (code interpreter, image generation, and other Qwen built-ins) in the chat.qwen.ai web interface. When it is enabled, Qwen's built-in tools fire their own response phases on every request, which DanyAPI cannot parse, and the server-side conversation history grows fast (hitting the input token limit quickly). With the switch off, DanyAPI's emulated tool calling (tools / tool_choice) and plain chat work as intended.

Or a single email + password account (login happens at startup):

bash

export QWEN_EMAIL="you@example.com"
                    export QWEN_PASSWORD="secret"

Both providers are optional. Run at least one of them (or both) - requests are routed to the right provider by the model name (deepseek-* / qwen*).

At startup every token is validated against its provider; invalid or expired tokens are skipped with a warning (the server refuses to start when no valid credential remains). The Qwen model list for /v1/models is fetched from the first Qwen account at startup (text-chat models only); if the fetch fails, a built-in default list is used. New requests without a session_id are distributed across healthy accounts round-robin.

Run

The .env file (gitignored, created from .env.example) is loaded automatically at startup:

bash

cp .env.example .env   # fill in tokens
                    python -m danyapi
                    # or
                    uvicorn danyapi.api.openai:app --host 0.0.0.0 --port 8000

Or with the helper scripts:

bash

run.bat   # Windows
                    ./run.sh  # Linux/macOS

Logging

By default logs go to the console. To persist them to a file, set DANYAPI_LOG_FILE in .env (or as an environment variable):

bash

DANYAPI_LOG_FILE=/var/log/danyapi.log   # or danyapi.log for the working dir
                    DANYAPI_LOG_LEVEL=INFO                  # DEBUG / INFO / WARNING / ERROR
                    DANYAPI_LOG_MAX_BYTES=10485760          # rotate at 10 MB per file
                    DANYAPI_LOG_BACKUP_COUNT=3              # keep 3 rotated files

The file is rotated by size (DANYAPI_LOG_MAX_BYTES, default 10 MB) keeping DANYAPI_LOG_BACKUP_COUNT backups (default 3). When started via python -m danyapi, uvicorn's own startup/access logs are routed through the same root logger and also land in the file.