Add release/summarize-changelog action for ref-range changelog compilation with template + LangChain multi-provider LLM hooks

[Copilot]

This introduces a new release action that compiles changelog content from commits between two refs. It supports conventional commit grouping, customizable Markdown rendering, and provider-agnostic LLM integration.

• New action: actions/release/summarize-changelog

  • Adds a composite action to build release notes from base-ref..head-ref using git log --no-merges --pretty=format:%s.
  • Exposes structured outputs:
    • changelog (fully rendered Markdown)
    • changes (compiled sections/lists)
    • commit-count
    • llm-prompt (prompt payload for any downstream LLM step)

• Conventional commit support

  • Adds conventional-commits input (true by default).
  • When enabled, commit subjects are grouped into sections (Features, Bug fixes, Documentation, etc.) based on commit type.
  • Non-matching messages fall back to Other changes.

• Template-driven Markdown output

  • Adds markdown-template input with a default template.
  • Supports placeholders:
    • {{base_ref}}
    • {{head_ref}}
    • {{commit_count}}
    • {{summary}}
    • {{changes}}
  • Adds optional llm-summary input for injecting externally generated summaries.

• LLM prompt-based summary generation

  • Adds npm-package-based summary generation using LangChain.
  • Adds optional inputs:
    • llm-provider (default: openai)
    • llm-model
    • llm-api-key
    • llm-base-url (default: https://api.openai.com/v1, used for openai)
  • When llm-summary is empty and llm-model is provided, the action generates the summary from llm-prompt using LangChain.
  • Supports multiple providers: openai, anthropic, google-genai.
  • Keeps optional llm-summary-command support as a fallback integration path.

• Docs/catalog update

  • Adds action README with usage, inputs, and outputs.
  • Adds the new action to the root README.md Release actions list.

- id: changelog
  uses: hoverkraft-tech/ci-github-publish/actions/release/summarize-changelog@main
  with:
    base-ref: v1.2.0
    head-ref: HEAD
    conventional-commits: "true"
    llm-summary: ""
    llm-provider: "openai"
    llm-model: "gpt-4o-mini"
    llm-api-key: ${{ secrets.OPENAI_API_KEY }}
    llm-base-url: "https://api.openai.com/v1"
    llm-summary-command: ""
    markdown-template: |
      ## Release notes
      Range: `{{base_ref}}..{{head_ref}}`

      {{summary}}

      {{changes}}

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

[neilime]

@copilot I want this action to use llm prompt to generate summary

[Copilot]

@copilot I want this action to use llm prompt to generate summary

Implemented in 9ebc381: the action can now generate the summary from llm-prompt via a new llm-summary-command input (prompt passed through stdin, summary read from stdout), while still supporting manual llm-summary override. No UI changes in this update.

[neilime]

@copilot we must use a npm package to help using LLM

[Copilot]

@copilot we must use a npm package to help using LLM

Implemented in bdfa639: summary generation now uses the npm openai package from llm-prompt when llm-model is set, with llm-api-key and OpenAI-compatible llm-base-url inputs. No UI changes in this update.

[neilime]

@copilot is it possible to use https://docs.langchain.com/oss/javascript/langchain/overview to handle LLM requests?

[Copilot]

@copilot is it possible to use https://docs.langchain.com/oss/javascript/langchain/overview to handle LLM requests?

Yes, it’s possible. We can use LangChain JS (for example @langchain/openai with ChatOpenAI) to handle the llm-prompt request flow while keeping the current inputs (llm-model, llm-api-key, llm-base-url). If you want, I can switch the current implementation to LangChain in a follow-up commit. No UI changes in this reply.

[neilime]

@copilot is it possible to use https://docs.langchain.com/oss/javascript/langchain/overview to handle LLM requests?

Yes, it’s possible. We can use LangChain JS (for example @langchain/openai with ChatOpenAI) to handle the llm-prompt request flow while keeping the current inputs (llm-model, llm-api-key, llm-base-url). If you want, I can switch the current implementation to LangChain in a follow-up commit. No UI changes in this reply.

@copilot use langchain, we shoud be able to use other providiver thanopenai too

[Copilot]

@copilot is it possible to use https://docs.langchain.com/oss/javascript/langchain/overview to handle LLM requests?

Yes, it’s possible. We can use LangChain JS (for example `@langchain/op...

Implemented in dd1ff6a: LLM summary generation now uses LangChain, with llm-provider support for openai, anthropic, and google-genai (not only OpenAI). It still uses llm-prompt and keeps llm-summary / llm-summary-command fallback behavior. No UI changes in this update.