fix: remove mandatory API key check and add LLM gateway support

[ccsert]

Summary

• Fixes Docker install method requires setting either DEEPSEEK_API_KEY, ANTHROPIC_API_KEY, or OPENAI_API_KEY #13: Remove mandatory API key validation in Docker mode. API key check is now a warning instead of a hard error, allowing use with free/local models or custom gateways that don't require standard API keys.
• Fixes How do i use LLM Gateways? #15: Add LLM Gateway configuration support and documentation. Users can now configure OpenAI-compatible gateways (OpenRouter, LiteLLM, Together AI, NVIDIA NIM, Ollama) via OPENAI_BASE_URL environment variable.

Changes

File	Change
entrypoint.sh	API key check → warning; add LLM_API_KEY, OPENROUTER_API_KEY, OPENAI_BASE_URL recognition; update help text
README.md	Add "Using LLM Gateways" section with gateway examples
README_zh.md	Add Chinese LLM Gateway docs; update secrets table
templates/workflow-docker.yaml	Add OPENAI_BASE_URL env var
templates/workflow-source.yaml	Add OPENAI_BASE_URL env var
.gitignore	Add packages/ directory

[Copilot]

Pull request overview

This PR updates the Docker entrypoint and documentation to make LLM API keys optional (warning-only) and to document/configure OpenAI-compatible “LLM gateway” usage via OPENAI_BASE_URL, addressing requests to support free/local models and gateway providers.

Changes:

• Downgrade Docker-mode LLM API key validation from a hard error to a warning; expand help text to mention gateway/key env vars.
• Add “Using LLM Gateways” documentation (EN + ZH) with example OPENAI_BASE_URL values for common gateways.
• Update workflow templates to include commented OPENAI_BASE_URL and ignore packages/ in git.

Reviewed changes

Copilot reviewed 5 out of 6 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
entrypoint.sh	Removes mandatory LLM key check; adds warning + help output for gateway-related env vars.
README.md	Adds LLM gateway usage section and updates secrets table.
README_zh.md	Adds Chinese LLM gateway docs and updates secrets table.
templates/workflow-docker.yaml	Documents optional OPENAI_BASE_URL in the Docker workflow template.
templates/workflow-source.yaml	Documents optional OPENAI_BASE_URL in the source workflow template.
.gitignore	Ignores packages/ directory.

[Copilot]

The warning message can be misleading when no keys are set because the script will still default to MODEL=deepseek/deepseek-chat (see print_config) which typically requires DEEPSEEK_API_KEY. Consider adjusting the warning to explicitly say that if you keep the default model you still need the corresponding provider key, otherwise set MODEL to a local/gateway-backed provider and/or OPENAI_BASE_URL.

Suggested change

	# Warn if no LLM API key is set (not fatal - some providers don't need one)
	if [ -z "$DEEPSEEK_API_KEY" ] && [ -z "$ANTHROPIC_API_KEY" ] && \
	[ -z "$OPENAI_API_KEY" ] && [ -z "$LLM_API_KEY" ] && \
	[ -z "$OPENROUTER_API_KEY" ] && [ -z "$OPENAI_BASE_URL" ]; then
	log_warn "No LLM API key detected (DEEPSEEK_API_KEY, ANTHROPIC_API_KEY, OPENAI_API_KEY, LLM_API_KEY, OPENROUTER_API_KEY)"
	log_warn "If you are using a local model or custom gateway, you can ignore this warning"
	log_warn "Set OPENAI_BASE_URL to use an OpenAI-compatible gateway (LiteLLM, OpenRouter, etc.)"
	# Warn if no LLM API key or OpenAI-compatible base URL is set
	if [ -z "$DEEPSEEK_API_KEY" ] && [ -z "$ANTHROPIC_API_KEY" ] && \
	[ -z "$OPENAI_API_KEY" ] && [ -z "$LLM_API_KEY" ] && \
	[ -z "$OPENROUTER_API_KEY" ] && [ -z "$OPENAI_BASE_URL" ]; then
	log_warn "No LLM API key or OPENAI_BASE_URL detected (DEEPSEEK_API_KEY, ANTHROPIC_API_KEY, OPENAI_API_KEY, LLM_API_KEY, OPENROUTER_API_KEY)"
	log_warn "If you keep the default MODEL=deepseek/deepseek-chat, you still need the corresponding provider key (typically DEEPSEEK_API_KEY)"
	log_warn "Only ignore this warning if you set MODEL to a local/gateway-backed provider and/or set OPENAI_BASE_URL for an OpenAI-compatible gateway (LiteLLM, OpenRouter, etc.)"

[Copilot]

In the gateway example, the snippet uses OPENAI_API_KEY, but the OpenRouter row below says to use OPENROUTER_API_KEY. This is inconsistent and may confuse users (and the default workflow templates only pass OPENAI_API_KEY). Please align the docs to one env var scheme (either document OpenRouter as OPENAI_API_KEY with OPENAI_BASE_URL, or add OPENROUTER_API_KEY consistently to secrets/workflow examples if it’s actually supported).

[Copilot]

The secrets table doesn’t mention OPENROUTER_API_KEY (or other gateway-specific keys), but the new gateway section later instructs users to use OPENROUTER_API_KEY. Either add the referenced secret(s) here or update the later section to use the keys that are already documented.

Suggested change

	| `OPENAI_API_KEY` | OpenAI API Key (optional) |
	| `OPENAI_API_KEY` | OpenAI API Key (optional) |
	| `OPENROUTER_API_KEY` | OpenRouter API Key (optional, for LLM gateway usage) |

[Copilot]

在网关示例中，代码块里使用的是 OPENAI_API_KEY，但下面 OpenRouter 的备注却写的是使用 OPENROUTER_API_KEY。这两处不一致容易让用户配置错（而默认 workflow 模板也只传了 OPENAI_API_KEY）。建议统一文档：要么说明 OpenRouter 也用 OPENAI_API_KEY + OPENAI_BASE_URL，要么在 secrets / workflow 示例中一致地加入 OPENROUTER_API_KEY（如果确实被支持）。

[Copilot]

Secrets 表里没有列出 OPENROUTER_API_KEY（或其他网关相关的 key），但后面的网关章节又要求使用 OPENROUTER_API_KEY。建议要么在这里补充对应的 secret，要么把后面章节改成使用这里已说明的 key。

Suggested change

	| `OPENAI_API_KEY` | OpenAI API Key（可选） |
	| `OPENAI_API_KEY` | OpenAI API Key（可选） |
	| `OPENROUTER_API_KEY` | OpenRouter API Key（使用 LLM 网关时可选） |