Skip to content

Conversation

@Jasper-Zhang-A
Copy link

Description

This PR adds an AutoDL/Linux CLI deployment guide and links it from the README so that cloud GPU users have an official, end-to-end checklist for running DeepCode’s CLI from source on a fresh Ubuntu instance.

Related Issues

  • No explicit GitHub issue, but this addresses common friction points for new users on cloud GPU platforms (AutoDL-style environments) when:
    • Setting up Python venv + pip from source.
    • Installing Node + MCP servers (Brave Search, filesystem).
    • Wiring the configuration together for the CLI workflow.

Changes Made

  • Added docs/autodl_cli_guide.md:

    • Step-by-step walkthrough for an AutoDL-like Ubuntu 22.04 environment:
      • Cloning HKUDS/DeepCode under /root.
      • Creating and activating a Python virtual environment.
      • Installing dependencies via requirements.txt.
      • Installing Node 20 via the official NodeSource repo.
      • Globally installing MCP servers:
        • @modelcontextprotocol/server-brave-search
        • @modelcontextprotocol/server-filesystem
      • Configuring mcp_agent.secrets.yaml with OpenAI/Anthropic/Brave keys (without committing secrets).
      • Running cli/main_cli.py in U/F/T modes.
    • Included an FAQ / common pitfalls section:
      • Brave path or API key issues.
      • venv activation problems.
      • npm global install permissions.
      • Non-JSON ResourceProcessor outputs and how the stricter JSON contract helps.
  • Updated README.md:

    • Added a link in the CLI section pointing to docs/autodl_cli_guide.md, so:
      • The base config remains platform-agnostic.
      • Linux/AutoDL-specific overrides are documented in one place for users who need them.

Checklist

  • Changes tested locally (guide followed on an AutoDL Ubuntu 22.04 instance)
  • Code reviewed
  • Documentation updated (this PR)
  • Unit tests added (if applicable)

Additional Notes

Test environment for the new guide:

  • OS: Ubuntu 22.04 (cloud GPU / AutoDL-style instance)
  • Python: 3.10.x (virtualenv)
  • Node: 20.x
  • MCP: global Brave Search + filesystem servers installed via npm

Following the guide from a clean instance is sufficient to:

  • Clone the repository,
  • Set up Python and Node environments,
  • Configure MCP servers and secrets,
  • And run the CLI pipeline successfully in both URL and file modes.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

2 participants