This package is called AI GCM and it is responsible for performing an artificial intelligence git commit -am <message> where the <message> shall become the output of a prompt request made to the AI Provider of your choice including Ollama. The summarize package is used in conjunction with this application.

If you have a summaries/ directory with valid summary.*.md files therein, the aigcm will use that for the context of the request.

This is capable of having project-aware capabilities to ask it to create single-sentence summaries of the STDIN with the anticipation that it's going to be come a git commit -m "<message>" where I use AI to populate the <message>.

Easy one-line to auto-commit new changes with meaningful message:

git add --all && git diff --staged | aigcm | git commit -a -F -

Easy one-line to auto-commit new changes to the last commit:

git add --all && git diff --staged | aigcm | git commit --amend -F -

go install github.com/andreimerlescu/aigcm@latest

# do something

# don't run `git add` them yet
#    if you do, then you need to run `git diff --staged | aigcm`

# reuse the output to reduce duplicate calls to the LLM
output=$(git diff | aigcm)
echo -e "${output}"
read -r -p "Commit this? [y|n*] " commit_this
if [ "${commit_this}" =~ /^y|Y/ ]; then
  git commit -m "${output}"
fi 

# or if you trust the AI lol (bruh... terminator... yeshua saves us all!!!)
git diff | aigcm | git commit -a -F -              # diff unstaged only with add
git diff | aigcm | git commit -F -                 # diff unstaged
git diff --staged | aigcm | git commit -a -F -     # staged diff with add
git diff --staged | aigcm | git commit -F -        # staged diff

# lets say that aigcm spits back a stupid output, or lets say you change a file 
# and want to update the commit message...
git diff --staged | aigcm | git commit --amend -F -

[hotfix/stateman-workflow ad197d6] Renamed a GITHUB_ENV variable name to align with artifact upload naming conventions in the Terraform state reporting workflow.
 1 file changed, 1 insertion(+)

As you can see from the above example, the aigcm command is designed to run without passing any arguments into it, however, that doesn't mean that there aren't a ton of configurables that aren't falling back to defaults that leverage Ollama and the qwen3:8b model via the http://127.0.0.1:11434 API and the gollm package.

In order to understand the usage of this binary outside of the default -h that you can pass into it to get:

aigcm -h

Usage of aigcm (powered by figtree v2.0.14):
-api-key [String] AI API Key to use for query (leave empty for ollama)
-caching[=true] [Bool] Enable LLM caching
-max-tokens[=3000] [Int] AI Max Tokens to use for query
-memory[=36963] [Int] AI Memory to use for query
-model[=qwen3:8b] [String] AI Model to use for query
-provider[=ollama] [String] AI Provider to use. (eg. ollama, openai, claude)
-seed[=-1] [Int] AI Seed to use for query
-summaries[=summaries] [String] Path to your summaries
-timeout[=77s] [UnitDuration] AI Timeout on each request allowed
-v[=false] [Bool] Display binary version

Therefore, when understanding how this connects to the actual Go application, you can see the breakdown of Variable to Type with the corresponding data-point where the pattern used demonstrates consumption of the goenv/env and figtree/v2 package.

This package uses the goenv/env package in order to connect ENV variables using env.String(eAiProvider, dAiProvider)

This package uses the figtree package in order to provide a command line interface.

The manner in which these gets registered to the flag package via figtree is through the env.<Type>(k<Property>, d<Property>) strategy.

This makes using <Type> in the figtree like *figs.<Type>(k<Property>) and you know that it'll at least have d<Property of <Type> assigned to it.

One strategy that I use for the aigcm package, instead of typing out:

git diff --staged | aigcm | git commit -a -F -

Is to use a helper script instead, which is why the install_helper.sh has been added. In here, once executed, four new binaries will be installed in the /usr/local/bin under the names of:

• No changes to staging workspace
  • gcai aka ai git commit (read as GC AI)
  • gcsai aka staged ai git commit (read as G CS AI)
  • gcmai aka amended ai git commit (read as G CM AI)
  • gmsai aka amend staged ai git commit (read as G MS AI)
• Includes all changes to staging workspace
  • gacai aka add all & ai git commit (read as GA C AI)
  • gacsai aka add all & staged ai git commit (read as GA CS AI)
  • gacmai aka add all & amended ai git commit (read as GA CM AI)
  • gamsai aka add all & amend staged ai git commit (read as GA MS AI)

So to simplify the matrix of command's naming conventions:

• G is for git
• C is for commit
• GC is for git commit
• GA is for git add
• CS is for commit staged
• CM is for amend commit
• MS is for amend staged
• AI is for artificial intelligence

8 commands that have 8 keys that unlock all 8 capabilities.

When using these helpers (bash shell scripts) on your host (supports any OS that supports bash 5.2+), you'll see:

$ gcsai

Received:  Added aigcm usage examples and installed scripts for automated commit message generation.

Use? [y|n] 

When you provide the y that its looking for, you'll see:

[master 211d58a] Added aigcm usage examples and installed scripts for automated commit message generation.
2 files changed, 114 insertions(+)
create mode 100755 install_helper.sh