Command Line
Run every command from the project root:
bash
weclone-cli [--config-path <config file>] <command> [options]--config-path selects the configuration file. The default is ./settings.jsonc, and the WECLONE_CONFIG_PATH environment variable works too. Add --help to any command to see its options.
Life portrait
| Command | Description |
|---|---|
build-profile | Runs data preparation, memory extraction, organizing and portrait generation on your chats, plus speaking style for a clone. See Build Your Portrait |
organize-memories | Runs domain induction, classification, topic grouping and summaries on their own, for debugging or reruns |
distill-style | Extracts speaking style on its own. See Speaking Style |
build-profile options:
| Option | Description |
|---|---|
--purpose assistant|clone | assistant extracts the portrait and events for personal assistants. clone also extracts speaking style for a clone. Asked interactively when omitted |
--output-dir <dir> | Result directory, which must be new. Defaults to a new timestamped directory under dataset/res_csv/agent/profile_runs |
--max-context-tokens <n> | Input token limit for organizing and portrait generation. Read from Codex by default. Required for the API backend |
Web UI and server
| Command | Description |
|---|---|
server | Starts the web UI and API server |
| Option | Default | Description |
|---|---|---|
--host | 127.0.0.1 | Listen address. See Security |
--port | 5175 | Port |
--run <dir> | Newest result | build-profile result directory to open |
--inference | Off | Loads the fine-tuned model and enables the /v1 inference API for chat bots |
See Review in the Web UI for the web UI and Connect AI Assistants for assistants.
Password and encryption
For encrypted mode. See Security.
| Command | Description |
|---|---|
security-init | Fixes the storage mode from settings.jsonc and sets the password shared by the web UI and file access |
security-change-password | Changes the password using the old one. Existing encrypted data is kept |
server-reset-password | Sets a new password when you forgot the old one. Old encrypted files can no longer be read, and all data must be generated again from the original chats |
decrypt --input <encrypted file> --output <plain file> | Asks for the password and exports an encrypted file to a new plain file |
Fine-tuned clone
| Command | Description |
|---|---|
make-dataset | Turns chat history into a fine-tuning dataset. See Data Preprocessing |
train-sft | Fine-tunes the model. See Model Fine-tuning |
webchat-demo | Chat with the fine-tuned model in the browser. Plaintext mode only. See Model Inference and Testing |
test-model | Tests the model with common chat questions. Needs a server started with --inference. See Model Inference and Testing |
Other
| Command | Description |
|---|---|
version | Shows the WeClone version |
dev | Starts the backend and frontend development servers together, for web UI development |