github desplega-ai/agent-swarm v1.160.0

latest releases: v2.13.3, v2.13.2, v2.13.1...
one month ago

Install & Run

npm:

npx @desplega.ai/agent-swarm@1.160.0
# or install globally
npm i -g @desplega.ai/agent-swarm@1.160.0

bun:

bunx @desplega.ai/agent-swarm@1.160.0
# or install globally
bun i -g @desplega.ai/agent-swarm@1.160.0

Docker:

# Server
docker pull ghcr.io/desplega-ai/agent-swarm:1.160.0
# Worker
docker pull ghcr.io/desplega-ai/agent-swarm-worker:1.160.0

E2B templates:

# API
agent-swarm-api-1-160-0
agent-swarm-api-latest

# Worker / lead runtime
agent-swarm-worker-1-160-0
agent-swarm-worker-latest

What's Changed

  • docs(specs): sync TLA+ specs with main (heartbeat, workflows) by @desplega-bot in #1738
  • fix(seed): treat source-identical live copy as in sync before pristine check by @desplega-bot in #1734
  • chore(docker): bump codex (2026-09-30) by @desplega-bot in #1735
  • fix(seed-scripts): project failure reasons for every task-failure-audit groupBy by @desplega-bot in #1736
  • fix(store-progress): refuse writes to another agent's task by @desplega-bot in #1737
  • fix(prompts): gate tool-search guidance by harness capability by @desplega-bot in #1727
  • docs(model-catalog): verify task returns overlayProposals, one refresh covers both vendors by @desplega-bot in #1733
  • feat(evals): suite v1 freeze, grader validation, attempt hygiene (Phase 3) by @desplega-bot in #1741
  • feat(evals): fan-out and worker-recovery swarm scenarios with solo baselines (Phase 7) by @desplega-bot in #1745
  • fix(ci): wait for late Dokploy deployment records and retrigger when none appears by @desplega-bot in #1742
  • fix(memory): recall task content instead of worker wrappers by @desplega-bot in #1740
  • fix(ui): resolve dashboard UI audit action items by @desplega-bot in #1743
  • chore(deps): bump agent-fs to 0.15.0 by @tarasyarema in #1748
  • feat(evals): analytics API for picking the best setup (Phase 4) by @desplega-bot in #1744
  • fix(opencode): send X-Source-Task-Id and X-Context-Key on the swarm MCP entry by @desplega-bot in #1749
  • docs(workflows): how to use system-one-decision with provider laya by @desplega-bot in #1750
  • feat(opencode): validate the final assistant message against outputSchema by @desplega-bot in #1753
  • ci(ui-e2e): time out and retry stuck artifact uploads by @tarasyarema in #1754
  • feat(pi): defer non-core swarm tools behind tool_search and pass structuredContent by @desplega-bot in #1729
  • feat(pi): connect installed MCP servers through pi's MCP extension by @desplega-bot in #1730
  • feat(evals): Leaderboard home with Pareto frontier and ranking table (Phase 5) by @desplega-bot in #1755
  • docs: add dashboard design reference by @desplega-bot in #1757
  • feat(evals): Scenario cards, heatmap, reliability and run outcome (Phase 6) by @desplega-bot in #1759
  • feat(workflows): add standard-webhooks verification mode for webhook triggers by @desplega-bot in #1760
  • feat(pi): opt-in codemode behind PI_CODEMODE by @desplega-bot in #1731
  • feat(evals): Phase 8 swarm scenarios — implement-review, capability-routing, human-in-loop by @desplega-bot in #1756
  • feat(evals): publishable benchmark v1.0 page and refusal-gated publish (Phase 10) by @desplega-bot in #1761
  • fix(telemetry): honor the opt-out in dashboard events and correct telemetry docs by @desplega-bot in #1773
  • feat(evals): run codex on the ChatGPT subscription, add gpt-6.1-sol and gpt-5.6, schedule public cells first by @desplega-bot in #1772
  • feat(model): reject a task model that the assignee harness cannot run by @fuvidani in #1764
  • fix(pi): run codemode in the compiled binary and unset deleted reloadable config by @desplega-bot in #1767
  • feat(slack): native working status for the whole life of an ask by @desplega-bot in #1770
  • feat(evals-ui): way back from /benchmark and an always-on Pareto line by @desplega-bot in #1771
  • fix(evals): copy the codex OAuth helpers into the evals image by @desplega-bot in #1778
  • fix(ui): resolve swarm UI audit v2 findings by @desplega-bot in #1774
  • fix(docs-site): fix the docs UI audit findings by @desplega-bot in #1776
  • chore(test): raise the db.ts line cap to 14400 until a frozen-body rule lands by @desplega-bot in #1783
  • fix(credentials): bench a Codex pool login after 2 auth failures in a row by @fuvidani in #1763
  • fix(boot): linear memory_fts backfill and retried worker registration by @desplega-bot in #1791
  • chore(test): freeze db.ts top-level bodies instead of capping its line count by @desplega-bot in #1784
  • fix(credentials): seat-aware Claude OAuth key selection for Fable by @fuvidani in #1765
  • docs(specs): sync TLA+ specs with main (heartbeat) by @desplega-bot in #1798
  • chore(docker): bump claude-code, pi, codex, opencode (2026-10-01) by @desplega-bot in #1792
  • fix(ci): stop the pi version probe from wedging test workers by @desplega-bot in #1795
  • docs: correct Codex plugin installation command by @desplega-bot in #1790
  • feat(evals): add real-diff-agent-fs scenario (a real merged fix, pushed and graded by hidden tests) by @desplega-bot in #1796
  • chore(deps): bump vulnerable docs-site dependencies (Dependabot alerts) by @desplega-bot in #1799
  • chore(deps): bump vulnerable apps/templates-ui dependencies (Dependabot alerts) by @desplega-bot in #1802
  • chore(deps): bump agent-fs image to 0.15.0 by @desplega-bot in #1803
  • release: v1.160.0 + docs daily update 2026-10-01 by @desplega-bot in #1797
  • chore(docker): bump dsh, ACP (2026-10-01) by @desplega-bot in #1804

Full Changelog: v1.159.0...v1.160.0

Don't miss a new agent-swarm release

NewReleases is sending notifications on new releases.