v2.4.1

Fixed

• HITL prompt loop + context handoff — Clarification responses are now reused across subtasks, repeated clarification prompts are deduplicated/capped, and follow-up tasks inherit prior user decisions instead of re-asking equivalent questions.
• Run view prompt stability — Local clarification prompts now pause/resume live rendering cleanly, reducing TUI corruption and preventing hidden/stuck prompts during interactive runs.
• Silent-failure propagation — Agent failures now surface as explicit task failures (instead of empty-success paths), with clearer errors for missing model configuration.
• Executor compatibility regression — Restored budget_manager compatibility on the runtime executor path to avoid end-of-run AttributeError crashes.

Changed

• Research tool routing defaults — Tool selection now biases web-first for research/business discovery tasks when no local dataset path is provided.
• Export readability + content coverage — Run export now backfills task outputs from tool traces when direct task result payloads are missing, and includes richer run details in markdown/html/latex/docx outputs.
• Duration formatting — CLI run summary and export outputs now prefer HH:MM:SS format.
• Telemetry noise control — Console span export is now opt-in (DEVSPER_OTEL_CONSOLE=1) to keep interactive TUI output clean by default.
• Auto-export workflow — Added config-driven post-run export option ([export] auto_export_on_run) with selectable output format.

Full Changelog: v2.4.0...v2.4.1

v2.4.0-docs

See CHANGELOG.md for details.

Full Changelog: v2.4.0...v2.4.0-docs

v2.4.0

Added

• Distributed runtime extension layer — Added devsper.distributed package with controller and worker-runtime wrappers for clean orchestration composition.
• Agent pool — Added runtime/agent_pool.py with reusable, worker-aware concurrent agent leasing (acquire_agent, release_agent, run_agent, run_parallel).
• Model router (runtime layer) — Added runtime/model_router.py for planning/reasoning/validation model routing with fallback chains.
• Speculative planner (runtime layer) — Added runtime/speculative_planner.py to predict and stage speculative successor tasks.
• Runtime HITL manager — Added runtime/hitl.py for pause/resume and explicit human-input request lifecycle handling.
• Execution graph upgrades — Extended runtime execution graph with worker assignment metadata and distributed lineage tracking.

Changed

• Executor composition — Runtime executor now integrates AgentPool, ModelRouter, speculative planning hooks, and distributed assignment tracking while preserving default behavior.
• Task runner routing — Task execution now supports pool-backed agent execution, scoped model routing, and fallback chain handling.
• Tool execution control — Runtime ToolRunner now supports bounded concurrency, dependency-aware batching, per-call timeout, cancellation, and structured results.
• Event backpressure handling — Runtime event stream now uses bounded queues with controlled overflow behavior.
• Runtime docs — Updated runtime architecture docs in devsper/README.md with distributed flow, worker/controller lifecycle, and HITL/speculative execution details.

Full Changelog: v2.3.0...v2.4.0

v2.3.0

Added

• Autonomous Missions capability layer — New devsper.missions package for long-horizon autonomous execution across research, coding, and experiment workflows.
• Mission planner and DAG generation — Added missions/mission_planner.py to break goals into mission tasks, assign role agents, and define dependency graphs.
• Research mission workflow — Added mission agents (researcher_agent, reviewer_agent, writer_agent, editor_agent) and end-to-end flow: research → critique → improve → finalize.
• Coding mission workflow — Added mission agents (architect_agent, coder_agent, tester_agent, debugger_agent, docs_agent) and end-to-end flow: design → code → test → fix → document.
• Iteration engine — Added missions/iteration_loop.py for quality-threshold-driven iterative execution with critique/improvement cycles.
• Mission memory — Added missions/mission_memory.py to persist research findings, decisions, and experiment artifacts per mission.
• Long-running mission support — Added checkpointing/resume support in missions/mission_runner.py plus HITL task-gating callback support.
• Mission UI — Added dashboard MissionView and TUI mission execution path (mission:research ..., mission:coding ..., mission:experiment ...) to show goal, DAG, and iteration progress.

Changed

• SDK surface — Exported MissionRunner from devsper package root for straightforward mission execution from Python.
• Testing — Added runtime tests for mission DAG planning and checkpoint creation.

Full Changelog: v2.2.0...v2.3.0

v2.2.0

Added

• Supermemory ranking backend — [memory] backend = "supermemory" (default) uses the Rust supermemory-core ranker for hybrid lexical (+ optional embedding) memory retrieval, deduplication, and recency-aware context formatting.
• Platform memory backends — [memory] backend can be platform or hybrid with platform_api_url and platform_org_slug; config resolver honors DEVSPER_PLATFORM_API_URL and DEVSPER_PLATFORM_ORG when set.
• Remote platform workflows — load_workflow("org/pkg@N") fetches workflow specs from the Devsper Platform API when DEVSPER_PLATFORM_API_URL, DEVSPER_PLATFORM_ORG, and (when required) DEVSPER_PLATFORM_TOKEN are configured.
• Local worker pool — devsper pool subcommands to run the local worker pool manager (foreground spawner) alongside improved swarm execution when using pool-backed paths.
• Devsper Cloud CLI — devsper cloud login, logout, run, status, and logs for JWT authentication (OS keychain) and runs against the platform API.

Changed

• Telemetry — OpenTelemetry configuration and OTLP export wiring expanded (otel_endpoint, otel_headers, and related options).
• Documentation — README refresh; removed outdated standalone docs where superseded by the current docs/ tree.

Full Changelog: v2.1.8...v2.2.0

v2.1.8

Changed

• Memory management — Runtime lifecycle and retention tuned for lower steady-state memory use (agent/worker paths, caches, and long-run cleanup).

Full Changelog: v2.1.7...v2.1.8

v2.1.7

Added

• Multi-format run export — New devsper export-runs command exports run history bundles to Markdown, reStructuredText, LaTeX, BibTeX, HTML, JSON, and DOCX (when python-docx is available), with per-run markdown artifacts under runs/.
• Dual PDF pipelines — Export supports both LaTeX (pdflatex) and HTML (wkhtmltopdf/pandoc) PDF generation, plus manifest reporting for generated outputs and pipeline errors.
• Branding-aware documents — Export templates now consume branding tokens and logo assets from homepage for consistent "docproc/devsper" presentation.
• Export docs and tests — Added docs/export-runs.md and automated tests covering format generation and event-path resolution.

Changed

• Distributed run stability — Controller dispatch/claim handling is more robust (slot-aware routing, shorter unclaimed-task requeue path, safer task-completion fallback parsing) to reduce stuck runs.
• Snapshot restore safety — Leader restore now avoids rolling back to stale/equivalent snapshots, reducing repeated re-dispatch loops.
• HITL/MCQ UX — Clarification prompts now use interactive numeric selection with a custom answer option and explicit "answer received" feedback.
• Run summary accuracy — Duration/cost aggregation now uses event-log grounded values for more reliable final summaries.
• CLI/TUI branding and noise — Theme/style namespace now prefers devsper.* aliases, and TUI logging suppression reduces controller log flooding during active runs.

Full Changelog: v2.1.6...v2.1.7

v2.1.6

Added

• Docs site overhaul — Completely redesigned landing page at hivemind.rithul.dev with animated hero, feature grid, code tabs, and registry callout. Restructured sidebar navigation with 7 categories: Getting Started, Core Concepts, Plugins, CLI Reference, Registry, and Self-hosting.
• Docs hosting on EC2 — Docusaurus static site now served via Caddy alongside the registry at hivemind.rithul.dev, with automatic TLS via Let's Encrypt.
• Unified CI/CD pipeline — Replaced separate registry-api.yml and registry-web.yml workflows with a single registry-deploy.yml (test, build, push, deploy) using OIDC + EC2 Instance Connect. New docs-deploy.yml workflow for automatic docs deployment on push.
• Registry deployment hardening — Improved Dockerfiles with multi-stage builds, production-ready Caddy configuration with security headers, and comprehensive deployment documentation.

Changed

• Docs site uses Geist + JetBrains Mono fonts, amber color system, dark-only mode
• Simplified docs.yml workflow to versioning-only (removed GitHub Pages deployment)

Full Changelog: https://github.com/devsper-com/runtime/commits/v2.1.6