Skip to content

CLI-Only Install

When you don’t need the Web UI. The CLI-only package is installed from a source checkout, not from PyPI.

Terminal window
git clone https://github.com/HKUDS/DeepTutor.git
cd DeepTutor
# Create a venv (macOS / Linux)
# Windows PowerShell:
# py -3.11 -m venv .venv-cli ; .\.venv-cli\Scripts\Activate.ps1
python3 -m venv .venv-cli && source .venv-cli/bin/activate
python -m pip install --upgrade pip
python -m pip install -e ./packaging/deeptutor-cli
deeptutor init --cli
deeptutor chat

deeptutor init --cli shares the same data/user/settings/ layout as the full app but:

  • Skips the backend / frontend port prompts
  • Defaults embeddings to off (choose Yes if you plan to use deeptutor kb … or RAG tools)
  • Still writes the complete runtime layout (system.json, auth.json, integrations.json, model_catalog.json, main.yaml, agents.yaml)
  • Still prompts for the active LLM provider and model
  • You’re running DeepTutor as a tool inside another agent’s harness (Claude Code, Codex, OpenCode, etc.)
  • Headless server / VM with no browser
  • You want a fast install with no Node.js toolchain
Terminal window
deeptutor chat # interactive REPL
deeptutor chat --capability deep_solve --tool rag --kb my-kb
deeptutor run chat "Explain Fourier transform"
deeptutor run deep_solve "Solve x^2 = 4" --tool rag --kb my-kb
deeptutor kb create my-kb --doc textbook.pdf
deeptutor memory show
deeptutor config show

Full surface: DeepTutor CLI.

The local deeptutor-cli install ships no Web assets or server dependencies. Keep the source checkout around — the editable install points to it.

SurfaceCLI-OnlyFull install
deeptutor chat interactive REPL
deeptutor run capabilities
deeptutor kb (Knowledge Bases)(opt in during init)
deeptutor notebook
deeptutor memory
deeptutor partner (Partners)(after installing channel SDKs — see below)
deeptutor config show
deeptutor provider login
deeptutor serve (FastAPI)
deeptutor start (Web UI)
Co-Writer UI
Memory Workbench UI
Interactive Book UI

To add the Web app later, install the PyPI package (Option 1) and run deeptutor init + deeptutor start from the same workspace.

CLI-only installs can still run deeptutor partner, but channel SDKs are not installed by default. The local deeptutor-cli project defines no extras — install the SDK set from the repo’s requirements mirror instead (you already have the source checkout):

Terminal window
python -m pip install -r requirements/partners.txt

This covers every built-in channel SDK (wecom-aibot-sdk, qq-botpy, slack-sdk, lark-oapi, python-telegram-bot, dingtalk-stream, …). The Matrix channel is packaged separately: pip install -r requirements/matrix.txt, plus requirements/matrix-e2e.txt for encrypted rooms (needs the system libolm library). This does not add the Web UI; Partner config continues to live in the current workspace under data/partners/<id>/ and data/user/settings/.

The CLI is agent-native. deeptutor run --format json emits one JSON event per line, and data-inspection commands such as kb list, kb search, session show, and notebook show expose JSON where implemented — useful for piping into another agent’s harness.

Hand the SKILL.md at the project root to any tool-using LLM and it can drive DeepTutor on its own. See Agent handoff for the patterns.

Terminal window
cd DeepTutor
git pull
pip install -e ./packaging/deeptutor-cli --upgrade

deeptutor: command not found after install

Section titled “deeptutor: command not found after install”

The console script wasn’t placed on PATH. Verify the venv is activated, or call it directly:

Terminal window
python -m deeptutor_cli.main chat

ModuleNotFoundError: No module named 'fastapi' when running deeptutor serve

Section titled “ModuleNotFoundError: No module named 'fastapi' when running deeptutor serve”

deeptutor serve is unavailable in the CLI-only install. Use PyPI or From Source instead.

Partner channel commands say an SDK is missing

Section titled “Partner channel commands say an SDK is missing”

Channel SDKs (wecom-aibot-sdk, qq-botpy, slack-sdk, lark-oapi, etc.) ship via the requirements mirror, not the CLI-only package:

Terminal window
python -m pip install -r requirements/partners.txt

More: Troubleshooting.