feat(pi): auto-detect oh-my-pi (omp) sessions - #1338

no API key found — this repo is configured to use deepseek/deepseek-v4-pro, which needs DEEPSEEK_API_KEY, but the runner has no key for it.

To fix: add the key as a GitHub Actions secret (referenced from your workflow's env: block) or as a Pullfrog secret in the console — or switch this repo to a different model (free models need no key).

Open repo secrets → · Configure model → · Setup docs → · Ask in Discord →

  ^{｜ Rerun failed job ➔ ｜ View workflow run ｜ via Pullfrog ｜ Using DeepSeek Pro (free via Pullfrog for OSS) ｜ 𝕏}

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

_{Comment @coderabbitai help to get the list of available commands.}

This PR was auto-closed. Only contributors approved with lgtm can open PRs. Open an issue first.

Maintainers review auto-closed issues and reopen worthwhile ones. Issues that do not meet the quality bar in CONTRIBUTING.md may not be reopened or receive a reply.

If a maintainer replies lgtmi, your future issues will stay open. If a maintainer replies lgtm, your future issues and PRs will stay open.

See CONTRIBUTING.md.

Bump, would love to see omp support!

@pullfrog review it
@codesmith review it
@cubic review it
@coderabbitai review it

@pullfrog resolve conflicts

@pullfrog review it
@codesmith review it
@cubic review it
@coderabbitai review it

@ryoppippi I have started the AI code review. It will take a few minutes to complete.

On it, I'm reviewing the PR changes now.

Resolved the merge conflicts between feat/pi-omp-support and origin/main and verified the resolution with cargo test -p ccusage pi (57 tests passed).

Files resolved:

• docs/guide/pi/index.md — kept the omp auto-detection row/paragraph and integrated the named stores section. Updated the named-store example to a non-default path (myfork) so it does not conflict with the now-default ~/.omp/agent/sessions scan.
• rust/crates/ccusage/src/adapter/pi/paths.rs — combined the new default pi + omp directory detection with the existing named-store helpers and tests, keeping everything pub(crate).

However, the push failed:

remote: Permission to sgfeniex/ccusage.git denied to github-actions[bot].
fatal: unable to access 'https://github.com/sgfeniex/ccusage.git/': The requested URL returned error: 403

maintainerCanModify is enabled, but the runner token still cannot push to the fork. A maintainer with write access to sgfeniex/ccusage (or the author) will need to pull the resolved commit and push it, or the conflicts can be re-resolved locally on the PR branch.

  ^{｜ ⚠️ this action is pinned to a commit SHA, which freezes the cleanup step — switch to @v0 or keep the SHA fresh with Dependabot ｜ View workflow run ｜ via Pullfrog ｜ Using Kimi K2 (free via Pullfrog for OSS) ｜ 𝕏}

@ryoppippi: I’ll review the changes.

@ryoppippi

@pullfrog review it
@codesmith review it
@cubic review it
@coderabbitai review it

I reviewed the diff on the PR head (c0241a9). Here's my assessment.

Verdict: looks good, safe to merge. The change is minimal, correct, and well-tested, and it follows the repo's cross-cutting docs flow (getting-started, pi guide, and the adapter README are all updated in the same commit).

What I checked:

• default_session_dirs in rust/crates/ccusage/src/adapter/pi/paths.rs:42 returns .pi before .omp in a deterministic order, filtering each by is_dir(). The override precedence in paths() is right: --pi-path wins, then PI_AGENT_DIR, then auto-detect, so the "overrides scan only the given paths" claim in the description holds.
• The dedup claim is accurate. load_entries_inner (loader.rs:27) keeps a single seen set across all returned paths and dedups by parser::entry_id (content-based: project + session_id + timestamp + tokens + cost), so genuinely distinct pi/omp sessions are both counted while overlapping content collapses.
• The four tests cover every dir-presence combination and assert exact vector equality (including order). They line up with the Fixture API (create_dir_all/root() both return tempdir.join(...)), so the equality assertions are valid.

Two minor, non-blocking notes:

1. Symlinked default dirs get read twice. If a user has ~/.omp/agent/sessions symlinked to ~/.pi/agent/sessions (plausible for a fork that reuses config), is_dir() follows symlinks so both paths are returned and every file is parsed twice. The entry-level dedup makes this correct (no double counting), just wasteful I/O. The codex adapter dedupes at the file/path level (codex/paths.rs) instead; not worth adopting that here unless it shows up in practice. This also mirrors existing_path_list, which doesn't canonicalize either, so it's consistent with current behavior.

2. Override drops omp auto-detect. Setting PI_AGENT_DIR (e.g. a pi user customizing only their pi path) silently stops scanning ~/.omp. That's the documented, intentional tradeoff, so fine, just flagging it as the one behavioral sharp edge for anyone reading later.

I didn't run cargo test locally (bare sandbox, full crate build is slow); CI is the source of truth and the tests are straightforward. No code changes needed from me.

@ryoppippi
I have no connection to this repository!!!!
Please don't mention!!!!!!

sorry! i'll take care of it

@pullfrog plz fix conflicts

I resolved the merge conflicts in PR #1338 and verified the changes, but I cannot push the resolution to the fork branch.

What I did:

• Merged origin/main into feat/pi-omp-support.
• Resolved conflicts in docs/guide/pi/index.md and rust/crates/ccusage/src/adapter/pi/paths.rs, keeping both the PR's oh-my-pi auto-detection and the main branch's named-store support.
• Ran cargo test -p ccusage pi -- --nocapture — all 57 pi-related tests passed.
• Committed the resolution locally.

Blocker:
pullfrog_push_branch failed with a 403 permission error when trying to push to the fork sgfeniex/ccusage:

remote: Permission to sgfeniex/ccusage.git denied to github-actions[bot].
fatal: unable to access 'https://github.com/sgfeniex/ccusage.git/': The requested URL returned error: 403

The resolved merge commit is on the local pr-1338 branch, but because the environment is ephemeral it will be lost unless the branch can be pushed. The most likely fix is ensuring the fork grants github-actions[bot] (or the Pullfrog app) write access to the PR branch.

  ^{｜ ⚠️ this action is pinned to a commit SHA, which freezes the cleanup step — switch to @v0 or keep the SHA fresh with Dependabot ｜ View workflow run ｜ via Pullfrog ｜ Using Kimi K2 (free via Pullfrog for OSS) (DeepSeek Pro not used — the program covers this model; add its provider key to run your pick) ｜ 𝕏}

Thanks for putting this together, and sorry for the slow reply.

Since this PR was opened, #1397 landed named pi-format stores, which covers this use case in a more general way. omp is the documented example for it:

{
	"$schema": "https://ccusage.com/config-schema.json",
	"pi": {
		"stores": [
			{ "name": "omp", "path": "~/.omp/agent/sessions" }
		]
	}
}

That reports omp alongside the default pi agent in ccusage daily / monthly / weekly / session, under its own omp agent name with [omp]-prefixed model labels, so the two sources stay separately attributable.

Auto-detecting ~/.omp/agent/sessions as a default pi path would also conflict with that feature. resolve_named_pi_store_paths seeds its owner map from the default pi paths and rejects overlaps, so anyone using the config above would hit a hard error on startup:

Invalid ccusage config: pi.stores name 'omp' paths overlap default pi store

I verified this by porting the change onto current main (the pi adapter has since moved to rust/adapters/pi/): the documented config loads fine on main, and fails with the error above once this change is applied.

The code itself is good — main just grew a different solution in the meantime. Closing for that reason. Thanks again for the contribution, and for following up on #1193.