Work Log

The system now supports AI Credit billing end to end: a unified AI credit account with exact generation pricing, new billing/contract schemas, and an atomically updated balance and append-only ledger for purchases, usage, and refunds. The frontend and backend are aligned on AI credit balances, purchase flows, and order/cart integration, while backend interfaces provide generation credit context for Studio usage. Documentation and tests were added to reflect the billing feature.

1 commit touching 18 files

The system now supports Namefi AI Credits end-to-end with validated credit amounts and pricing, correct ledger listing, and atomic credit fulfillment/cancellation with refund handling. Order processing was updated so failed AI-credit steps cancel only the relevant credit items and preserve partially completed order states when refunds fail. Studio generation and leadgen workflow tests were aligned to ensure correct charging/refunding metadata and idempotent workflow starts.

1 commit touching 31 files

The admin AI Credits experience was refactored from bulk credit awarding and award ledger display to a read-focused workflow. It now shows customer account balance inspection (with refresh) and an immutable, paginated ledger history with search, event types, references, and timestamps, while permission checks were tightened to require only read access.

1 commit touching 9 files

The backend admin AI credits API was refactored to replace award-listing with read-only access to a user’s AI credits account and audited ledger history. History search was expanded to cover ledger fields and related order context, and contract schemas were updated accordingly. Permission surface was adjusted to remove the separate write permission for AI credits, aligning the feature with audited, read-focused grant visibility.

1 commit touching 6 files

Implemented a dedicated AI Credits experience: a new authenticated AI Credits balance/purchase/history page with loading/error states, plus a cart and search/side-nav entry to buy credits. Updated generation flows to verify credit balance and disable generation when unavailable, and ensured credit state is refreshed after relevant generation/order status changes. Orders now render exact AI-credit purchase lines in lists and progress timelines.

1 commit touching 122 files

1 commit touching 6 files

Added a PostgreSQL-backed AI Credits billing service that atomically charges, records ledger entries, and refunds credits when generation/outbound/order actions fail or are canceled. Studio generation workflows now reconcile credit charges on activity failure, and order AI-credit purchases are fulfilled transactionally. Introduced contract and router support for listing the AI credit ledger and cart/order AI-credit items, with updated pricing and validation schemas to unify Studio and Outbound.”

1 commit touching 44 files

1 commit touching 11 files

Introduced Zod schemas and helpers to represent Namefi AI credits in USD cents, validate nonnegative/positive credit values, define AI ledger and billable action types, and model exact AI pricing for supported image and animation generation plus outbound run/outreach reasoning efforts. Added functions to retrieve the correct credit price from a provided pricing configuration based on request selection parameters.

1 commit touching 2 files

2 commits touching 49 files

The admin AI Credits feature was reimplemented as a unified page. It now supports selecting a user to view their authoritative AI credit balance, granting positive credits to one or more recipients with confirmation, and browsing an immutable credit ledger history with search, pagination, and reference details. A manual refresh action was added, and the previous award-ledger UI was removed.

1 commit touching 9 files

6 commits touching 191 files

1 commit touching 11 files

The repository now uses a compatibility-scoped release model: shared/internal distribution artifacts are no longer independently versioned, and release candidate selection is aligned with published runtime compatibility pins. CI and Moon task grouping were updated to use truthful capability tags and to shard/group jobs by setup requirements. Release tooling and metadata checks were tightened to validate compatibility versions and preserve correct dependency graphs across native, WASIX, SDKs, and

15 commits touching 316 files

2 commits touching 16 files

The system now allows more GitHub summary work to start per UTC day and month (100/day, 300/month), raising the hard monthly budget accordingly. The work-unit summary UI no longer shows a processing spinner when a stale summary is still displayed. Cron configuration, core tests, and summary store boundary/concurrency tests were updated to reflect the new limits and expected stale-summary behavior.

1 commit touching 6 files

The activity heatmap component now renders a simpler figcaption when the history is partial, instead of showing the full “Less/More” color legend. When history is complete, the figcaption is omitted, preventing unnecessary legend display in partial-history cases.

1 commit touching 1 file

The repository’s CI workflow was updated to remove the production build job, leaving only the test step to run during workflow execution.

1 commit touching 1 file

The system now evaluates and serves GitHub activity-unit summaries using newest-first prioritization, ensuring newer items stay ahead of backlogs. While a replacement summary is pending, the UI marks the unit as refreshing without hiding the previous same-attribution summary. Summary selection and public feed rendering were updated to track summarization state and reuse stale accepted output until a verified current summary is ready.

1 commit touching 12 files

18 commits touching 27 files

1 commit touching 32 files

1 commit touching 7 files

This update bumps Oliphaunt’s embedded Postgres runtimes and SDKs to v0.2.0 (and related Wasix/wasix-ts to new versions), unifying native and WASIX public APIs. It introduces breaking changes around WASIX storage layout/IndexedDB handling and simplifies releases. It also includes a WASIX/Rust WASIX Node-API related refactor, plus performance and bug-fix improvements such as seek/WAL handling and JS exec response streamlining.

1 commit touching 84 files

The GitHub release workflow was updated to correctly handle cases where Release Please produces no pull request and no relevant unreleased shared-contribution changes are detected. In that situation, it now reports the specific condition and terminates successfully, preventing false “no open PR” messaging and avoiding unnecessary failures.

1 commit touching 1 file

The repo refactors CI/release task modeling to better separate static checks, compile/unit/package/qualify responsibilities and to gate release mutations with exact, validated planning. It improves release-pr normalization, affected-target matrix generation, and human-readable task labeling, adds policy/release mutation test scopes, and strengthens release-content-write pacing. Net effect: more deterministic, correctly-scoped CI lanes and a safer, contract-driven release workflow with added self

3 commits touching 118 files

Updated the performance runner configuration to depend on the local workspace paths for the Rust SDK crates instead of registry-pinned versions. Also strengthened policy checks to ensure path dependencies used by the unpublished perf runner do not declare registry versions, requiring the runner to resolve the checked-out workspace source.

1 commit touching 2 files

Oliphaunt was updated to a new v0.2.0 release across runtimes, SDKs, and extensions. The update unifies embedded PostgreSQL public APIs and aligns native and WASIX runtime/SKD packaging, introduces WASIX/IndexedDB layout changes, and shifts wasix-ts host runtimes through Rust Node-API. The release also includes performance and bug fixes such as improved seek behavior and streamlined JS execution handling, alongside simplified release ownership.

1 commit touching 84 files

1 commit touching 3 files

The system now computes work-unit ownership using verified merge-landings and produces deterministic coverage results. A separate, scheduled summary worker claims and settles at most one provider summary request within a bounded runtime, independent of factual projection/backfill. Projection policy digests are upgraded to drive self-activating refreshes, while the public feed remains cacheable and publication waits for the bounded factual and summary workers as configured.

1 commit touching 24 files

1 commit touching 38 files

Increased the GitHub worker execution window and aligned its HTTP and runtime limits, while improving how summary generation is gated to only start when sufficient time remains. The worker also adjusts its internal reserve/projection budgeting and ensures deadline reporting covers the full worker runtime, reducing the chance of partial or overrun summary processing.

1 commit touching 8 files

GitHub work attribution now remains consistent when provider ledger file and aggregate counters drift. Eligibility checks no longer fail solely due to mismatched file-derived additions/deletions versus aggregate values, and tests confirm that such drifted provider-ledger evidence is still projected as owned work and included in crosswalk eligible/projection categories (while still excluding zero-diff or ineligible cases).

1 commit touching 3 files

1 commit touching 10 files

The update adds robust handling for numeric repository date inputs by verifying they are safe, non-negative integers and converting them from Unix seconds to ISO strings. Corresponding tests for push webhook routing were adjusted to ensure the pushed_at value is provided in the expected numeric form, improving consistency and correctness of webhook date processing.

1 commit touching 2 files

1 commit touching 23 files

1 commit touching 107 files

1 commit touching 29 files

1 commit touching 58 files

1 commit touching 64 files

The repository adds an index and new canonicalization/aliasing logic to detect exact duplicate authored commits within completed fingerprints, choose a deterministic canonical root, and set canonical aliases (with evidence) across chained aliases. It also marks affected summary attempts as indeterminate after applying canonical aliases, and updates audit/feed/store logic and tests to ensure UI and historical/current projections remain consistent.

1 commit touching 13 files

The GitHub activity backfill script and workflow were updated to use a typed environment object and add stronger validation for required configuration. It now enforces that DATABASE_URL is set, CRON_SECRET meets minimum length requirements, and required GitHub tokens are present per target account. Unit tests were added to confirm validation behavior, including failure when CRON_SECRET is missing or too short/blank.

1 commit touching 3 files

GitHub push observation now properly preserves provider error details so another configured token can retry private comparisons after over-cap scenarios. Exact comparisons that exceed durable commit limits can fall back to a bounded evidence gap instead of returning a generic incomplete status. Tests verify correct retry behavior across 404/409 and preservation of retryable GraphQL rate-limit metadata.

1 commit touching 2 files

GitHub activity backfill and auditing are now more resilient and informative: inconclusive projection retries use bounded progressive delays, database connections are closed before strict projection audits, and projection errors are standardized into structured codes/names. Durable push handling now correctly bypasses reachable history when exact webhook evidence is complete and rejects oversized durable fallback requests. Tests were updated to validate the new retry behavior, structured error,5

1 commit touching 6 files

GitHub push observation processing now validates compare results against durable push evidence. If the compare history becomes incomplete or inconsistent (e.g., surplus commits or force-push changes), it automatically falls back to fetching the exact expected commit set. Tests confirm recovery of the correct head and full durable members when compare responses are misleading or fail, while rejecting surplus data when durable evidence is not complete.

1 commit touching 2 files

The worker and backfill pipeline now track commit canonicalization attempts/results alongside aliases, enabling progress detection when canonicalization advances without queue claims. PR membership evidence is strengthened by explicitly validating stored membership counts and head-sha matches, invalidating derived aliases and forcing membership refresh when inconsistent. Retry logic for inconclusive audit reads is bounded, and canonicalization is only applied when pending, improving correctness.

1 commit touching 8 files

Updated GitHub polling logic to compute the next poll timestamp from response cache-control when a 304 response omits the poll interval, and to avoid falling back to cached freshness when the poll interval is malformed or missing. Tests were adjusted to verify the new cache-based behavior and stricter rejection of invalid provider poll intervals.

1 commit touching 2 files

The repository adds an auditable backfill workflow that completes GitHub activity discovery within a bounded time window, then runs a stored-projection pipeline audit to verify results. Runs succeed only when discovery bounds finish and scoped audits confirm stored projection correctness; otherwise they fail as incomplete or completed-with-gaps. The system improves idempotency, deduplication, pagination stability, worker scope isolation, and terminal gap handling for provider failures.

1 commit touching 28 files

JavaScript, React Native, and WASIX TypeScript SDKs now use a single shared @oliphaunt/js-core workspace package for protocol and query helpers. Packaging was updated to bundle only the intended compiled js-core artifacts, and the shared API surface across platforms was aligned. Tests and runtime code were adjusted to import shared helpers, removing duplicated generated mirrors and protocol/query implementations.

1 commit touching 79 files

Integration coverage improves by reworking the external PostgreSQL raw client into a persistent session so connection-local state is preserved across multiple protocol requests. Server-side tests also handle client concurrency deterministically with timeouts and queued-client expectations, and native-extension server test lifecycles properly track both server shutdown and client session closure.

3 commits touching 8 files