The events Git AI exports to help you measure AI adoption, compare agents, and track the durability of AI code

AI Metrics

The otel-cli exports OpenTelemetry log events that you can query in your observability platform to understand AI tool adoption, compare agent effectiveness, and measure code durability.

Event Schemas

The CLI exports three types of events:

PR Events (`git.pr.merged` / `git.pr.closed`)

Emitted when a PR is merged or closed. One event is created per tool/model combination, plus an aggregate event with tool="all".

Attribute	Type	Description
`event.name`	string	`git.pr.merged` or `git.pr.closed`
`git.pr.link`	string	Full URL to the pull request
`git.pr.repository`	string	Repository URL
`git.pr.tool`	string	AI tool name (e.g., `cursor`, `copilot`) or `all` for aggregate
`git.pr.model`	string	Model name (e.g., `claude-3.5-sonnet`, `gpt-4`) or `all` for aggregate
`git.pr.human_additions`	int64	Lines written entirely by humans
`git.pr.mixed_additions`	int64	Lines written by an AI, then edited by a human
`git.pr.ai_accepted`	int64	AI-generated lines that were accepted without changes
`git.pr.ai_additions`	int64	`ai_accepted` + `mixed_additions`
`git.pr.total_ai_additions`	int64	Total AI lines generated while working on the PR. Will usually be greater than `ai_additions` because engineers revert AI-code
`git.pr.total_ai_deletions`	int64	Total lines deleted by the AI while working on the PR.
`git.pr.time_waiting_for_ai`	int64	Cumulative seconds spent waiting for AI responses
`git.pr.pr_total_added`	int64	Total lines added in PR (from git diff)
`git.pr.pr_total_deleted`	int64	Total lines deleted in PR (from git diff)

Daily Summary Events (`git.daily`)

Emitted during scheduled exports (e.g., nightly cron). Summarizes commits from the last 2 weeks. One event is created per tool/model combination, plus an aggregate event with tool="all".

Attribute	Type	Description
`event.name`	string	`git.daily`
`git.daily.date`	string	UTC date in `YYYY-MM-DD` format
`git.daily.repository`	string	Repository URL
`git.daily.tool`	string	AI tool name (e.g., `cursor`, `copilot`) or `all` for aggregate
`git.daily.model`	string	Model name (e.g., `claude-3.5-sonnet`, `gpt-4`) or `all` for aggregate
`git.daily.total_commits`	int64	Total commits in the date range
`git.daily.commits_with_authorship`	int64	Commits that have authorship data
`git.daily.git_diff_added_lines`	int64	Total lines added (from git diff)
`git.daily.git_diff_deleted_lines`	int64	Total lines deleted (from git diff)
`git.daily.human_additions`	int64	Lines written by humans
`git.daily.mixed_additions`	int64	Lines written by an AI, then edited by a human
`git.daily.ai_additions`	int64	`ai_accepted` + `mixed_additions`
`git.daily.ai_accepted`	int64	AI-generated lines that were accepted without changes
`git.daily.total_ai_added`	int64	Total AI additions
`git.daily.total_ai_deleted`	int64	Total AI deletions
`git.daily.time_waiting_for_ai`	int64	Seconds waiting for AI

Examples

Here are a few examples of how you can use these events to understand your team's AI adoption:

Team Adoption Rate

Question: What percentage of PRs have AI assistance?

filter: git.pr.merged where tool = "all"

total_prs      = count(*)
prs_with_ai    = count(*) where ai_additions > 0
ai_assisted    = prs_with_ai / total_prs

Adoption by Tool

Question: Which AI tools are being used across our org?

filter: git.daily where tool != "all"
group by: tool

total_ai_lines     = sum(ai_accepted)
cumulative_ai      = sum(total_ai_added)
repos_using        = count_distinct(repository)

Weekly Active Repositories

Question: How many repositories had AI-assisted commits this week?

filter: git.daily where tool = "all" and ai_accepted > 0
group by: week

active_repos    = count_distinct(repository)
total_ai_lines  = sum(ai_accepted)

Comparing Agents

Use the tool and model attributes to compare effectiveness across different AI coding assistants.

Acceptance Rate by Agent

Question: Which agent's code is accepted most often?

filter: git.pr.merged where tool != "all"
group by: tool, model

accepted_lines   = sum(ai_accepted)
total_ai_lines   = sum(ai_additions)
acceptance_rate  = ai_accepted / ai_additions

Code Volume by Agent

Question: How much code is each agent contributing?

filter: git.pr.merged where tool != "all"
group by: tool

pr_count         = count(*)
ai_lines_added   = sum(ai_additions)
ai_lines_deleted = sum(total_ai_deletions)
total_pr_lines   = sum(pr_total_added)

Model Comparison Within Tool

Question: For Cursor users, which model performs best?

filter: git.pr.merged where tool = "cursor"
group by: model

pr_count         = count(*)
accepted_lines   = sum(ai_accepted)
total_lines      = sum(ai_additions)
acceptance_rate  = ai_accepted / ai_additions

Under development

AI-Code Halflife - How long does AI-generated code last in production before being removed or rewritten? Broken out by repository, tool, model, and contributor/group.
Accepted after Review - What percentage of AI-generated code is accepted in a Code Review? Inital ai_accepted - # of those lines changed/removed during the PR
o11y events for AI-generated code - How does AI-generated code perform in production? Production errors linked back to LoC,repository, tool, and model that generated the code.

Set up a call if you want the feature-flag for any of the above enabled: https://calendly.com/acunniffe/meeting-with-git-ai-authors

Requesting additional metrics

Need something you don't see here? We probably have it, and just need to start emiting events. Set up a call or open an issue on GitHub.

OTEL Events