feat: include OMP session costs

[kevcube]

Summary

• scan ~/.omp/agent/sessions/**/*.jsonl alongside compatible pi session logs
• reuse the existing pi-compatible parser and cache for Codex and Claude attribution
• document OMP support in provider docs, CLI help, and the changelog

Related to #384.

Redacted runtime proof

I ran the branch against an actual OMP session. The first command reads only metadata and redacts the session filename and token counts:

$ jq '<select first OMP openai-codex assistant message and redact usage>' ~/.omp/agent/sessions/<redacted>.jsonl
{"source":"~/.omp/agent/sessions/<redacted>.jsonl","provider":"openai-codex","api":"openai-codex-responses","model":"gpt-5.6-sol","usage":{"input":"<redacted>","cacheRead":"<redacted>","output":"<redacted>","totalTokens":"<redacted>"}}

A fresh local Codex cost scan on the same branch reports that OMP model. Token and cost values are redacted:

$ CodexBarCLI cost --provider codex --format json --refresh --days 1 | jq '<retain source/model and redact usage>'
[
  {
    "provider": "codex",
    "source": "local",
    "daily": [
      {
        "date": "2026-07-17",
        "modelsUsed": ["gpt-5.6-sol"],
        "inputTokens": "<redacted>",
        "cacheReadTokens": "<redacted>",
        "outputTokens": "<redacted>",
        "totalTokens": "<redacted>",
        "totalCost": "<redacted>"
      }
    ]
  }
]

Verification

• swift test --build-system native --filter 'scanner merges omp sessions with pi sessions'
• full sharded Swift suite: 696 selections across 58 groups passed (one group passed on the configured retry)
• make check
• rebuilt CodexBarCLI and ran the redacted runtime proof above against local OMP sessions

The native SwiftPM builder and DEVELOPER_DIR=/Applications/Xcode-beta.app/Contents/Developer were used locally because the current macOS 27 command-line toolchain's default builder could not initialize its property-list parser.

[clawsweeper]

Codex review: needs maintainer review before merge. Reviewed July 18, 2026, 12:26 AM ET / 04:26 UTC.

Summary
The PR scans ~/.omp/agent/sessions/**/*.jsonl through the pi-compatible cost path, updates the cache schema and deduplication behavior, and documents and tests the new OMP source.

Reproducibility: not applicable. this is an additive feature rather than a report of broken behavior. The contributor nevertheless supplied redacted terminal proof from a real OMP session and a post-change Codex cost scan.

Review metrics: 2 noteworthy metrics.

• Patch surface: 7 files affected; 354 added, 41 removed. The implementation spans scanner and cache behavior, targeted tests, CLI help, and both provider-facing documents.
• Focused test coverage: 1 merged-source scanner test added. The new test exercises simultaneous pi and nested OMP session discovery, which is the central changed behavior.

Merge readiness
Overall: 🐚 platinum hermit
Proof: 🦞 diamond lobster
Patch quality: 🐚 platinum hermit
Result: ready for maintainer review.

Overall follows the weaker of proof and patch quality, so missing proof can cap an otherwise strong patch.

Rank-up moves:

• Obtain an explicit maintainer decision on whether OMP becomes a supported pi-compatible local-cost source.

Risk before merge

• [P1] Merging documents and supports a new ~/.omp session-log root, creating an ongoing compatibility expectation for an external JSONL format whose evolution is outside this repository.

Maintainer options:

1. Confirm the OMP support contract (recommended)
   Merge after a maintainer explicitly accepts OMP JSONL as a supported pi-compatible local-cost input and its corresponding maintenance obligation.
2. Keep pi-only session discovery
   Pause or close this PR if external OMP log compatibility should not become a documented CodexBar behavior.

Next step before merge

• [P2] No narrow repair is identified; the remaining action is an explicit maintainer choice about supporting OMP as a documented external session source.

Maintainer decision needed

• Question: Should CodexBar officially support OMP session logs as a pi-compatible local-cost source?
• Rationale: The implementation is narrow and proof-backed, but accepting a new documented third-party session root is a product and maintenance-contract choice that source review cannot resolve.
• Likely owner: steipete — The available PR history ties steipete directly to the most recent scanner correctness work and provides the strongest routing signal.
• Options:
  • Support OMP sessions (recommended): Merge the scanner and documentation change, with OMP JSONL compatibility becoming part of the local-cost support contract.
  • Keep the current source boundary: Decline the OMP root and retain the existing pi-session scope until there is an explicit support plan.

Security
Cleared: The supplied patch changes local session discovery, parsing, cache state, tests, CLI text, and documentation without adding dependencies, privileged execution, secret handling, or CI/release changes.

Review details

Best possible solution:

If maintainers want OMP to be a supported local-cost source, retain the shared pi-compatible parser, entry-ID deduplication, cache invalidation, focused scanner coverage, and explicit provider documentation; otherwise keep the existing pi-only source boundary.

Do we have a high-confidence way to reproduce the issue?

Not applicable: this is an additive feature rather than a report of broken behavior. The contributor nevertheless supplied redacted terminal proof from a real OMP session and a post-change Codex cost scan.

Is this the best way to solve the issue?

Unclear: reusing the existing pi-compatible scanner is the narrowest implementation path, but only after maintainers decide that OMP is an intended supported source rather than an unsupported compatible format.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against 5d1d22630513.

Label changes

Label changes:

• add merge-risk: 🚨 compatibility: The diff adds a documented ~/.omp scan root and therefore an external session-format compatibility commitment.

Label justifications:

• P3: This is an optional local-cost-source enhancement with no evidence of an urgent regression in existing provider behavior.
• merge-risk: 🚨 compatibility: The diff adds a documented ~/.omp scan root and therefore an external session-format compatibility commitment.
• rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🦞 diamond lobster and patch quality is 🐚 platinum hermit.
• status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Sufficient (terminal): The PR body contains redacted after-fix terminal evidence from a real OMP session and a fresh local Codex cost scan that shows the discovered OMP model; no further contributor proof is needed.
• proof: sufficient: Contributor real behavior proof is sufficient. The PR body contains redacted after-fix terminal evidence from a real OMP session and a fresh local Codex cost scan that shows the discovered OMP model; no further contributor proof is needed.

Evidence reviewed

What I checked:

• Real behavior proof: The PR body includes redacted terminal output from a real OMP JSONL session and a fresh CodexBarCLI cost --provider codex --format json --refresh --days 1 scan that reports the OMP model. (ab1f76269601)
• Focused implementation and coverage: The supplied diff adds an OMP-root option and merged-source coverage in the pi-compatible scanner, plus cache-version invalidation for the expanded per-entry deduplication state. (Sources/CodexBarCore/PiSessionCostScanner.swift:18, ab1f76269601)
• Compatibility boundary: Both provider documents explicitly add ~/.omp/agent/sessions/**/*.jsonl to supported pi-compatible sessions, making the new third-party log-format support user-visible rather than an internal parser-only change. (docs/codex.md:154, ab1f76269601)
• Review continuity: The previous completed ClawSweeper review at the earlier head reported no actionable findings after the contributor added proof; the current head adds a deduplication follow-up commit authored by steipete. (Sources/CodexBarCore/PiSessionCostScanner.swift:124, ab1f76269601)

Likely related people:

• steipete: The supplied PR history shows steipete authored the current deduplication follow-up in the affected pi-compatible scanner path and merged current main into the branch. (role: recent area contributor; confidence: medium; commits: c188dc9f359c, ab1f76269601; files: Sources/CodexBarCore/PiSessionCostScanner.swift, Sources/CodexBarCore/PiSessionCostCache.swift)

What the crustacean ranks mean

• 🦀 challenger crab: rare, exceptional readiness with strong proof, clean implementation, and convincing validation.
• 🦞 diamond lobster: very strong readiness with only minor maintainer review expected.
• 🐚 platinum hermit: good normal PR, likely mergeable with ordinary maintainer review.
• 🦐 gold shrimp: useful signal, but proof or patch confidence is still limited.
• 🦪 silver shellfish: thin signal; proof, validation, or implementation needs work.
• 🧂 unranked krab: not merge-ready because proof is missing/unusable or there are serious correctness or safety concerns.
• 🌊 off-meta tidepool: rating does not apply to this item.

Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

How this review workflow works

• ClawSweeper keeps one durable marker-backed review comment per issue or PR.
• Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
• A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
• PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
• Maintainers can also comment @clawsweeper review to request a fresh review only.
• Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
• Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
• Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

Review history (3 earlier review cycles)

• reviewed 2026-07-17T19:43:20.284Z sha dd290c2 :: needs real behavior proof before merge. :: none
• reviewed 2026-07-17T20:23:35.926Z sha dd290c2 :: needs maintainer review before merge. :: none
• reviewed 2026-07-17T21:05:39.756Z sha dd290c2 :: needs maintainer review before merge. :: none

[kevcube]

@clawsweeper re-review

[clawsweeper]

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

Re-review progress:

• State: Complete
• Detail: The durable review comment was published and its verdict was queued for target-serialized routing.
• Run: https://github.com/openclaw/clawsweeper/actions/runs/29613544824
• Updated: 2026-07-17T21:08:31.943Z

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: ab1f762696

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Deduplicate unkeyed mirrored session entries

When the same session is present under both scanned roots but an assistant row has no top-level id (a shape this parser accepts elsewhere and stores in unkeyedContributions), this branch applies those unkeyed totals once per file before the per-session seenEntries filter runs. That means mirrored pi/OMP copies with matching sessionID but id-less assistant usage are still double-counted in the daily totals instead of being counted once.

Useful? React with 👍 / 👎.

[steipete]

Merged as 9e9f739b6a96e2a786467adb59f1ce6b35f1d6ac after maintainer review and hardening.

Verification:

• Audited the full diff and contributor context. The original implementation commit is dd290c27f41318957be052bc1a2a5cef2c69ab66; c188dc9f359c036435e3c9d025ba2542538457b9 merged current origin/main; maintainer fix ab1f76269601f2aa15adbf9a5985c6fda98bc5bf added entry-level cross-root deduplication and cache migration.
• Confirmed OMP discovery feeds the existing pi-compatible parser and cache. Incremental offsets remain intact; JSONL reads remain bounded by the existing 16 MiB cap.
• Confirmed matching assistant entry IDs in the same session are counted once across .pi and .omp, while divergent unique turns in copied sessions are retained.
• swift test --filter PiSessionCostScannerTests: 18 passed.
• swift test --filter CostUsageCacheTests: 9 passed.
• make check: passed; SwiftLint reported 0 violations.
• make test: 713/713 test selections passed across 60/60 groups, with zero retries.
• Final autoreview: clean, no actionable findings, correctness confidence 0.90.
• Exact-head CI run 29630500869 for ab1f76269601f2aa15adbf9a5985c6fda98bc5bf: 8 checks passed, 0 failed, 0 pending. Both macOS shards, both Linux builds, lint, aggregate, changes, and GitGuardian were green. Merge state was CLEAN and MERGEABLE.

Thanks @kevcube for bringing OMP session-cost coverage to CodexBar, and for the useful redacted live proof.