Always include SHA in get_file_contents responses

[yonaka15]

Summary

This PR provides a simpler solution to #595 by always including SHA information in the get_file_contents tool responses without requiring any parameter changes.

Unlike #605 which adds an include_sha parameter, this approach transparently enhances all responses with SHA information while preserving the existing MCP server behavior.

Approach

The implementation adds a preliminary Contents API call to retrieve SHA information before fetching the actual content:

1. Call Contents API to get file metadata (including SHA)
2. Call Raw Content API to get the actual content (as before)
3. Include SHA in both the resource URI and success message

This dual-API approach ensures:

• Text files continue to return raw text content
• Binary files continue to return base64-encoded content
• SHA information is always available for create_or_update_file operations

Changes

• Modified pkg/github/repositories.go to add Contents API call before Raw Content API
• Updated tests in pkg/github/repositories_test.go to include Contents API mocks
• No changes to tool schema or parameters - fully backward compatible

Benefits over #605

1. No parameter required - SHA is always included automatically
2. Simpler for users (AI) - No need to remember to set include_sha=true
3. Better default behavior - Addresses the core issue directly
4. Minimal performance impact - The additional API call is lightweight

Testing

Verified with both unit tests and MCP Inspector against real repositories:

• Text files return raw content with SHA ✅
• Binary files return base64 content with SHA ✅
• Directories return JSON with SHA for each item ✅
• Graceful fallback when Contents API fails ✅

Closes #595

[Copilot]

Pull Request Overview

This PR enhances the get_file_contents tool to always include SHA information in responses by adding a preliminary Contents API call before fetching raw content. This provides a simpler, backward-compatible solution that automatically includes SHA data without requiring parameter changes.

Key changes:

• Added Contents API call to retrieve file metadata (including SHA) before fetching raw content
• Updated resource URI format to include SHA instead of branch reference
• Enhanced success messages to include SHA information

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
pkg/github/repositories.go	Added Contents API call to retrieve SHA before raw content fetch and updated URI/message formatting
pkg/github/repositories_test.go	Updated test mocks to include Contents API responses and adjusted expected URIs to use SHA format

Comments suppressed due to low confidence (1)

pkg/github/repositories.go:513

• The variable name 'errContents' is inconsistent with Go naming conventions. It should be 'err' or follow the existing pattern used elsewhere in the function.

				fileContent, _, respContents, errContents := client.Repositories.GetContents(ctx, owner, repo, path, opts)

[LuluBeatson]

Thank you so much @yonaka15 for reworking this based on our feedback. It's great to see this working!

Screenshot

get_file_contents being used before create_or_update_file to get the blob SHA of file to update

My only minor comments (which I will action):

• Changing the resource URI is not necessary. I will revert this.
• Using the contents API to get the file's blob SHA is not optimal. I will look into using a GraphQL query to get just the SHA.

Thank you again, I'm excited to get your work merged :)

[LuluBeatson]

Examples

File from Default

Get repositories.go from github/github-mcp-server

File from Branch

Get repositories.go from the add-pagination-graphql branch of github/github-mcp-server

File from Commit

Get repositories.go from the f47e5542e419e6d0bea8d3b54849ae5613344af0 commit of github/github-mcp-server

File from PR

Get repositories.go from PR 676 of github/github-mcp-server

Dir from Default (unchanged by PR)

get the pkg/github dir from github/github-mcp-server

Image from Default

Get 10mb.png from lulubeatson/test-image

[LuluBeatson]

I will look into using a GraphQL query to get just the SHA.

The GraphQL query was slow so I've stuck with your original method using the contents method of the GitHub API.

GraphQL (~350ms)

You can run the following query in https://docs.github.com/en/graphql/overview/explorer and see the response time in the network panel

query {
  repository(owner: "github", name:"github-mcp-server") {
    object(expression: "main:README.md") {
      ... on Blob {
        oid
      }
    }
  }
}

Contents (~30ms)

https://api.github.com/repos/github/github-mcp-server/contents/README.md

https://api.github.com/repos/{owner}/{repo}/contents/{path}

Docs: https://docs.github.com/en/rest/repos/contents?apiVersion=2022-11-28#get-repository-content

Raw (~30ms)

https://raw.githubusercontent.com/github/github-mcp-server/d15026b0eb2a2e5d3265a2601798ab28017dc719/README.md

https://raw.githubusercontent.com/{owner}/{repo}/{sha}/{path}

[tommaso-moro]

Nice one! The approach to contents api vs raw api vs graphql makes sense to me!