{"full_name":"AgentSwarms-fyi/agentswarms","owner":"AgentSwarms-fyi","name":"agentswarms","shard":"1","url":"https://github.com/AgentSwarms-fyi/agentswarms","description":"Unified Agentic AI and Data Platform","homepage":"https://agentswarms.fyi","topics":["agentic-ai","agentic-bi","agentic-workflow","agents","ai-workflows","artificial-intelligence","business-analytics","langchain","langgraph","rag","rag-chatbot","rag-pipeline","ai-analyst","business-inteligence","data-catalog"],"tag_pages":["agentic-ai","agentic-workflow","agents","ai-workflows","artificial-intelligence","data-catalog","langchain","langgraph","rag"],"language":"TypeScript","license":"NOASSERTION","avatar":"https://avatars.githubusercontent.com/u/307769312?v=4","created_at":"2026-07-21T21:35:03Z","pushed_at":"2026-11-15T09:45:29Z","metrics":{"stars":302,"forks":66,"watchers":4,"issues":9,"prs":101,"releases":10,"issues_open":7,"issues_closed":2,"prs_open":21,"prs_closed":66,"prs_merged":14,"discussions":6},"growth":{"day":{"stars":1,"forks":0,"base_day":"2026-11-15"}},"history":{"dates":["2026-11-13","2026-11-14","2026-11-15","2026-11-16"],"stars":[300,301,301,302],"forks":[66,66,66,66],"watchers":[4,4,4,4],"issues_open":[7,7,7,7],"prs_open":[22,22,24,21]},"boards":{},"same_owner":[],"issues":[{"number":23,"title":"Export traces and metrics to OpenTelemetry / Prometheus","comments":0,"created_at":"2026-07-28T20:17:04Z"},{"number":22,"title":"Document generation: plan large decks in batches instead of one 16k-token JSON","comments":0,"created_at":"2026-07-28T20:17:03Z"},{"number":21,"title":"Scheduled notebook jobs (run a published notebook on a cron)","comments":0,"created_at":"2026-07-28T20:17:01Z"},{"number":20,"title":"Canvas parallel execution: commit in level order, and stop bailing out on condition levels","comments":0,"created_at":"2026-07-28T20:16:59Z"},{"number":19,"title":"Make the embedding vector width configurable (currently pinned to 1536)","comments":0,"created_at":"2026-07-28T20:16:57Z"},{"number":18,"title":"Automatic re-ingestion: keep Knowledge Base sources fresh on a schedule","comments":0,"created_at":"2026-07-28T20:16:55Z"},{"number":17,"title":"Knowledge Base connectors: Google Drive, Confluence, Notion, SharePoint, Slack","comments":0,"created_at":"2026-07-28T20:16:52Z"}],"discussions":[{"number":57,"title":"AgentSwarms 1.2.2","comments":0,"category":"Announcements","created_at":"2026-08-29T08:34:16Z"},{"number":50,"title":"AgentSwarms 1.2.1","comments":0,"category":"Announcements","created_at":"2026-08-20T15:41:27Z"},{"number":44,"title":"AgentSwarms 1.1.0","comments":0,"category":"Announcements","created_at":"2026-08-08T17:24:01Z"},{"number":43,"title":"AgentSwarms 1.0.0","comments":0,"category":"Announcements","created_at":"2026-08-06T12:59:24Z"},{"number":37,"title":"General Discuss","comments":2,"category":"General","created_at":"2026-08-03T15:34:57Z"},{"number":35,"title":"Welcome to agentswarms Discussions!","comments":0,"category":"Announcements","created_at":"2026-08-02T09:48:51Z"}],"latest_release":{"name":"AgentSwarms 1.7.0","tag":"v1.7.0","published_at":"2026-11-14T01:13:59Z","body":"AgentSwarms 1.7.0 is the first release published as images, and the first that deploys\nto AWS without assembling it by hand. Every image is built for amd64 and arm64,\nsmoke-tested, scanned and signed. An install picks a release with `AGENTSWARMS_VERSION`,\nand rolls back the same way. The schema now installs on a plain PostgreSQL such as\nAmazon RDS. `deploy/aws` deploys the whole platform with Terraform: as VMs behind a load\nbalancer, on Amazon ECS, or on EKS from the new Helm chart. Users' code runs on Fargate\nthere, or on a node pool of its own. Behind a load balancer, restarts and rolling\ndeploys no longer cut replies off, and the scheduler runs once across the fleet. The\nlakehouse gains Iceberg v3, time travel and Iceberg targets on the Spark engine, and the\ncatalog reads Hive Metastore and AWS Glue. On security: credentials, traces and\napprovals are sealed under per-user and customer keys, sandboxes are isolated from one\nanother, and a dynamic scan of the running app found nothing exploitable. Read Upgrading\nbefore you move: several migrations must be applied in the order given there.\n\nEverything 1.7.0 changes, adds and fixes is in [CHANGELOG.md](https://github.com/AgentSwarms-fyi/agentswarms/blob/v1.7.0/CHANGELOG.md): the whole of it is longer than a GitHub release may be.\n\n### Upgrading\n\n- **Apply `20261114000000` before choosing the ECS sandbox backend** (R477). It lets\n  the runtime settings name `ecs` and adds the three columns and the function a Fargate\n  sandbox fetches its environment through. It only adds things: either order with the\n  build is safe, and the Docker and Kubernetes backends never touch them.\n- **The images have new names** (R470): `ghcr.io/agentswarms-fyi/<image>:local` for a\n  build from the checkout, where they were `agentswarms/<image>:latest`. Re-running\n  `scripts/setup.sh` (or `docker compose up -d --build`) builds them under the new\n  names. The installer clears `NOTEBOOK_RUNTIME_IMAGE` when `.env` still holds the old\n  default, `agentswarms/notebook-runtime:latest`; do it by hand otherwise, or the app\n  keeps starting kernels from the old name. The old images stay on disk until\n  `docker image prune`. `setup-k8s.sh` deploys the new names; a cluster that loaded\n  the old ones needs the new ones loaded the same way.\n- **Update the notebook gateway image with the app** (R466). Sandboxes the new app\n  starts refuse a gateway that sends no token, so an old gateway in front of a new app\n  cannot open notebook kernels. The new gateway works with sandboxes an older app\n  started, and `docker compose up -d --build` updates both. Running MCP servers and\n  scorers keep working until they restart; the ones started after the update check the\n  token.\n- **Apply `20261111000000` first, then deploy the build** (R456). The catalog page now\n  reads the new `metadata` column by name, so the new build fails on a database that does\n  not have it yet. The migration only adds things, so the running build is unaffected by\n  it. A bucket with Hive-partitioned folders is re-grouped on its next crawl: the\n  per-folder assets are replaced by one table, and anything written on those old assets\n  by hand goes with them.\n- **Deploy the build first, then apply `20261110000000`** (R454). The migration drops\n  the two functions the browser used to seed sample datasets with. In that order nothing\n  breaks. The other way round, a new instance can't seed its samples until the build\n  arrives, and existing samples are unaffected either way. Run\n  `npx supabase db push --remote` from the repository folder, because the link to your\n  project is saved there. To confirm it\n  landed without `psql`, see the [schema audit](docs/security/README.md#the-schema-audit).\n  Before you upgrade, or after, see `docs/security/assessments/2026-11-06-schema.md` for\n  how to check whether a running install was affected.\n- **`bash scripts/setup-selfhosted.sh` now needs `--yes` when nobody is at a\n  terminal.** The Supabase CLI asks before it applies","is_prerelease":false},"readme":true,"readme_zh":false,"news":[],"stale":false,"snapshot_at":"2026-11-16T09:23:14Z","history_start":"2026-09-06"}