Problem: agent unresponsive after install (needs init + provider config) —— Trove problem library

What you will run into

After deepcode starts, the interactive session doesn't respond, or agent tasks error; deepcode loop/goal doesn't work. Installed but won't run - because you haven't initialized and haven't configured an LLM.

Why this happens

First use needs a two-step config:

Without an LLM configured the agent chain spins idle, so the interaction naturally doesn't respond.

The first-use "no response" is the chain broken at the config layer: DeepCode's agent chain is "user input -> agent planning -> call LLM -> output", and the LLM is the mandatory executor in the chain.

deepcode init only generates the config directory (the skeleton); deepcode provider writes the "LLM connection and key" (the power source) — the two steps correspond to "skeleton" and "power".

Missing the second step, every step of the chain spins idle at "call LLM", so the interaction naturally doesn't respond.

That also defines the verification order: confirm the skeleton first (~/.deepcode exists) -> confirm the power second (provider config holds a valid key) -> send one real request last to verify the chain.

One more easy trap: the key is an external credential — a typo, an expired key, or an account without quota all look like "no response/error" — so before suspecting the tool, confirm the key itself is valid (via the provider's own test command or a minimal request).

Step by step

1. Initialize — Required

deepcode init

Generates ~/.deepcode.

2. Configure the provider — Required

deepcode provider

Configures the LLM connection and model.

3. Quick verification without a key — Required

deepcode exec 'simple task'

Use the error to locate the issue; or deepcode --help to confirm the command surface.

How to confirm it works

The command surface works (measured output shape):

deepcode --version -> DeepCode 2.1.0
deepcode --help -> full command surface 

After configuring the provider, tasks run (needs an LLM key; not measured in this marketplace).

Other environments

Symptom

After deepcode starts, the interactive session doesn't respond, or agent tasks error; deepcode loop/goal doesn't work.

Cause

First use needs initialization and LLM config: deepcode init (generates ~/.deepcode) + deepcode provider to configure the model key; without an LLM the agent chain spins idle.

Solution

1. deepcode init to initialize.
2. deepcode provider to configure the LLM connection and model.
3. Quick no-key verification: deepcode exec 'simple task' to see the error; or deepcode --help to confirm the command surface.

Measured install / uninstall in this marketplace (2026-08-28)

Install (manifest install.cmd, isolated venv /tmp/l34-dc/venv, Python 3.14.4):

[install] pip install -i https://pypi.tuna.tsinghua.edu.cn/simple deepcode-hku && deepcode --version
Dependencies: aiohttp/cryptography/pillow/mcp/mcp-server-git/openai/anthropic/google-genai etc., 60+ packages
DeepCode 2.1.0
[install] exit=0

Uninstall (manifest uninstall.cmd + verify):

[uninstall] pip uninstall -y deepcode-hku -> Successfully uninstalled deepcode-hku-2.1.0
[uninstall] verify: deepcode removed: OK (fresh shell, verify exit=0)

Evidence details kept locally

• CLI smoke: local repro log (DeepCode 2.1.0; full command surface).
• Co-install measured: local repro log (omc/agt) + coexist-herdr/logfire (2026-08-28).
• Not measured: TUI interaction, MCP client real run, agent tasks (need an LLM key).

Name confusion

• Three names: PyPI package deepcode-hku / CLI command deepcode / project HKUDS/DeepCode - use the package name for install/uninstall, the command name daily.
• Distinguish from DeepSeek (a model) and DeepCode (other same-name code-analysis tools): this project = HKUDS/DeepCode, the open-source Agentic coding framework.