List view
- No due date•22/27 issues closed
- No due date•17/28 issues closed
- No due date•5/25 issues closed
Targeted for CodeWhale v0.8.70
Due by September 8, 2026•2/28 issues closedTargeted for CodeWhale v0.8.69
Due by September 1, 2026•3/24 issues closedTargeted for CodeWhale v0.8.68
Due by August 25, 2026•2/28 issues closedTargeted for CodeWhale v0.8.67
Due by August 18, 2026•8/30 issues closedTargeted for CodeWhale v0.8.66
Due by August 11, 2026•5/29 issues closedTargeted for CodeWhale v0.8.65
Due by August 4, 2026•10/27 issues closedTargeted for CodeWhale v0.8.64
Due by July 28, 2026•29/30 issues closedTargeted for CodeWhale v0.8.63
Due by July 21, 2026•8/23 issues closedTargeted for CodeWhale v0.8.62
Due by July 14, 2026•11/62 issues closedPost-0.8.60 polish and integration follow-up: provider polish, UI affordances, docs, slash-command/workbench cleanup, and non-blocking platform improvements.
Due by July 8, 2026•52/52 issues closedCodeWhale v0.8.60 promotes the v0.8.59 headless-worker direction into an always-running agent fleet control plane: a local-first fleet manager, worker host adapters, inbox/ledger state, health checks, restart/interrupt controls, alerts, and verifiable task/eval loops. Inspired by large-scale research-agent fleets, but scoped for CodeWhale: safe local/SSH workers first, explicit auth/secrets boundaries, replayable runtime events, and no hosted control plane by default. ## In scope - Fleet manager loop that keeps many agent workers running, polls health, and records status into a durable inbox/ledger. - Worker contracts for local and SSH-backed hosts: start, status, heartbeat, logs, interrupt, restart, stop. - Scheduler/backpressure for queues, leases, concurrency, retry policy, and stuck-worker recovery. - Alert integrations for Slack/webhooks and PagerDuty-style escalation behind explicit config. - Verifiable task/eval specs so fleet work produces artifacts, scores, and receipts instead of loose transcripts. - TUI/CLI/API visibility: compact fleet health, worker drilldown, artifacts, and takeover actions. - Runtime API/SDK alignment so TUI, CLI, web/editor clients, and WhaleFlow can consume the same worker events. ## Non-goals - No uncontrolled public cloud agent farm by default. - No secrets in prompts, logs, argv, or persisted event payloads. - No silent auto-page/Slack behavior without explicit user config. - No model-weight training loop; this milestone builds orchestration/eval infrastructure, not RL training itself. - No bypass around Rust-owned sandbox, permissions, provider calls, cancellation, or event accounting. ## Definition of done - A local dev can run a small fleet smoke with multiple workers, durable status, restart/interrupt, and artifact receipts. - A configured SSH host can run at least one remote worker and report health/logs back to the manager. - A failing/stuck worker is detected, recorded, and either restarted or escalated according to policy. - Fleet state is visible from CLI/TUI/API and is replayable enough for handoff/debugging. - Docs explain how to run the fleet safely and how it relates to headless sub-agents, WhaleFlow, and the Runtime API.
Due by July 1, 2026•26/26 issues closed- Due by June 25, 2026•56/56 issues closed
Next patch release
No due date•13/13 issues closedCodeWhale v0.8.57 release cycle. Target: ~15 issues.
No due date•4/4 issues closedThe v0.9.0 architectural release promotes CodeWhale from a turn/subagent workbench into a WhaleFlow workflow workbench: typed branch-and-leaf workflows, pod-style background workflow monitoring, shared ARMH/RLM memoization, deterministic replay, external-memory evaluation, and a GEPA-style teacher/student promotion loop that turns validated lessons into a cached-main overlay. Primary tracker: #2981 EPIC: v0.9.0 WhaleFlow branch/leaf workflow mode (re-established after #2667 was deleted) ## In scope - WhaleFlow workflow mode: background workflow runs, /workflows-style monitoring, done/total progress, longest-running item peek, inspect/replay/report surfaces. - Typed Workflow IR as the source of truth: Starlark/YAML/generated plans compile to Rust-owned IR before execution. - Rust async executor: bounded branches, bounded leaves, cancellation, budgets, permissions, LoopUntil, Cond, Expand, BranchTournament, and Pareto reducers. - Branch/leaf semantics: isolated speculative branches, bounded leaves, losing-branch fruit harvesting, typed results. - ARMH/RLM integration: exact-context shared memo across branches with visible hit/miss/cost telemetry. - External-memory evaluation: decide whether Aleph-style memory belongs in core, optional plugins, or explicit workflow nodes, with visible state and clear/export controls. - TraceStore and deterministic replay: replay from recorded leaf/control outputs, not live model calls, unless explicitly allowed. - Teacher harness: TeacherReview proposes reusable lessons; StudentReplay and PromotionGate validate before promotion. - Cached-main overlay: promoted notes, workflows, tests, branch heuristics, model/cache policies, and prompt patches warm future runs without mutating Git main. - Janitor: stale invalidation, memo cleanup, candidate demotion, trace compaction, capacity enforcement. - Model-provider abstraction: workflow roles map to capabilities and configured providers; no workflow logic hardcodes Arcee, DeepSeek, Claude, tool calls, JSON mode, or large context. ## Non-goals - No model-weight RL in v0.9.0. - No arbitrary JS/Python as workflow source of truth. - No script-level async/await. Starlark is a pure graph builder; Rust executes IR. - No hidden external-memory dependency for normal CodeWhale operation. - No uncontrolled self-modifying agent. Teacher output is inspectable, replayed, and reversible. - No public performance claims until evals are reproducible. ## Definition of done - workflows/rlm_cache_change.star runs with mock provider in CI and can dogfood CodeWhale RLM/ARMH/provider changes. - Branch/leaf engine, control flow, TraceStore, replay, ARMH shared memo, TeacherReview, StudentReplay, PromotionGate, overlay, and janitor have focused tests. - Workflow mode can run, inspect, and replay a workflow from CLI and TUI. - ARMH savings, provider costs, and any external-memory use are visible in workflow telemetry. - All behavior is behind config/feature flags until stable. ## Release gate - Parity gates green on the v0.9.0 integration branch. - CHANGELOG [0.9.0] frames the release as WhaleFlow branch/leaf workflows and validated cached-main learning. - Docs explain the Claude-workflow-inspired UX while preserving CodeWhale's typed IR/Rust executor safety model.
No due date•33/62 issues closed