@nocobase/app-plugin-agents 0.1.0-beta.9 → 1.0.0-beta.15

package/CHANGELOG.md

@@ -1,5 +1,298 @@
 # @nocobase/app-plugin-agents
 
+## 1.0.0-beta.15
+
+### Major Changes
+
+- baf80fe: Remove the old conversation start, fallback and restore HTTP operations and the conversation fallback/restore CLI commands without aliases. Replace start with POST /api/agents/conversations followed by POST /api/agents/conversations/:id/messages, or conversation create followed by conversation message send. These are two non-atomic requests and may leave an empty conversation if sending fails. Fallback and restore have no same-session replacement: explicitly create a new conversation with the desired executor/agent, without automatically copying history or changing an existing binding. Trusted domain start/fallback/restore methods remain available.
+
+  Historical unbound conversations remain read-only, with owner-authorized cancellation of unfinished old runs. Browser callers must use the narrow configuration input: move tool lists, model lists and system-prompt policy into trusted server configuration, and declare business variables through selected-scope schemas. Existing thinking handling remains unchanged. Apply the visibility migration before serving requests; existing ordinary records default to visible and new background sessions are hidden, with known-session authorization preserved.
+
+  Background run still acknowledges admission. Use get to wait for validated, saved and finalized results, or try with a finite observation budget; observation timeout does not cancel execution. Deliver can return null after saving a notice in the caller's transaction without waking a new-bound conversation. Remove proposed-plan quantity hints from consultation cards while retaining historical data and plan mechanisms, and consolidate the framework and application upgrade documentation.
+
+### Minor Changes
+
+- baf80fe: Add extensible UI context snapshots, application-owned tool renderers, and guarded browser form registries to Agents. Preserve exact originating message linkage through public history and recovery, and demonstrate reviewed, draft-only form proposals in Examples without saving records or continuing completed tools.
+- baf80fe: Add a shared public chat-model catalog hook and explicit model selection when creating Online Agent conversations. Update the Examples chat selector to group models by service alongside Agents, with an online-model-or-runner conversation hint and vertical-only option scrolling, preserving Runner status, metadata, and draft content.
+- baf80fe: Use shared AI SDK Chat owners for public chat clients, render ordered UIMessage parts and Runner source segments, and reconcile durable history, tool approval, recovery and independent Stop across application surfaces.
+- baf80fe: Stream native text increments immediately and support callable Runner agents through the shared OnlineAgent realtime service, atomic conversation admission, canonical block history and unified chat controls. Preserve executor bindings, explicit configuration validation and cancellation confirmation, and exclude server business context from public message metadata.
+- baf80fe: Persist native OnlineAgent UI messages directly, expose standard AI SDK response streams, and bind durable tool approval continuations to the owning conversation and execution. Keep private model evidence separate from public parts and retry failed finalization without rerunning effects.
+- baf80fe: Extract Native execution ownership behind the execution-specific OnlineAgentBackend contract and delegate backend selection, session creation, restoration and configuration from OnlineAgentService. Replace the Runner-specific backend interface and forwarding wrapper with one concrete Runner session, require explicit unsupported-capability rejection, preserve accepted response ownership across failed sends, retain deferred HTTP configuration boundaries, pin shared recovery to accepted submissions and drain Runner admissions before shutdown without cancelling or redispatching remote work.
+
+  Use admission-only Runner receipts to avoid treating post-commit conversation projection failures as failed inputs. The server-side `ConversationService.sendRealtime()` return contract is now `{ seq, run }`; ordinary conversation send and HTTP response contracts are unchanged.
+
+  Keep startup repair on the session manager and shared producer probing on OnlineAgentService instead of duplicating them in every backend. Remove the public Service `isActive()` and `notify()` methods, the backend notification entry and the `OnlineAgentNotification` type. Existing transactional conversation notices and Runner event refresh are unchanged. Restore sessions by authorized persisted mode and binding metadata, reject invalid bindings or unavailable backends explicitly, and never use configuration reads to infer backend identity.
+
+  Organize OnlineAgent internals under `server/online/agent/`, alongside the online model gateway and services, by Native execution, Runner execution, persistence, message conversion/validation and shared transport. Consolidate fragmented helpers and separate interfaces and data types from runtime implementations while preserving the public server entry points and execution behavior.
+
+- baf80fe: Require a plain JSON Schema in OnlineAgent background `run()` calls instead of Zod. Pass the schema to the AI SDK for structured generation and validate the generated result before persistence. Reject invalid, remote-reference and oversized schemas before accepting execution.
+- baf80fe: Add the OnlineAgent server service foundation with versioned conversation configuration, complete SDK message persistence, injectable business tools, shared execution control, and first-subscription stream caching. Preserve ordinary conversation and message identifiers without persistent request deduplication. Connect the original chat HTTP resources and first-subscription events to one shared browser submission and stream controller for application and Registry chat views. Preserve uploaded attachments, per-view drafts and unknown-outcome feedback without hidden retries or Runner fallback.
+
+  Recognize no-content streaming responses by HTTP status even when a browser exposes an empty body, so completed chats read saved history without a false disconnect warning.
+
+  Resolve Scope/Provider model changes consistently through the configuration API. Let users explicitly permit a new input after an unknown send result, and preserve Registry drafts and files when the shared client refuses a submission.
+
+  Let application-provided Organize entries own their availability feedback while retaining default Runner availability hints.
+
+  Bind the authoritative execution session ID to each business tool context so applications can associate business proposals with the current chat without trusting model-supplied identifiers.
+
+- baf80fe: Export validated SDK UI message contracts for OnlineAgent integration and align application SDK dependencies with the matching React client.
+- baf80fe: Project saved Runner blocks into SDK UI messages with stable source identities, support optional cross-instance Redis stream recovery and native Stop, and persist independent native business notifications.
+
+### Patch Changes
+
+- baf80fe: Keep an acknowledged realtime recovery from attaching an older stream after a new chat input has started, preserving the active SDK response and its writing state.
+- baf80fe: Avoid empty assistant bubbles while thinking and redundant success notifications in shared Agents chat. Keep chat transcripts within their available width while preserving bounded scrolling for wide code blocks and tables, application-owned message extras, and actionable execution feedback.
+- baf80fe: Rename the OnlineAgent configuration component to `OnlineAgentConfigurator` and expose it as `onlineAgent.configurator`. Update public exports, tool registration examples, and application consumers to use the consistent naming.
+- baf80fe: Reauthorize bounded references from the most recent historical page for follow-up chat input, preserving current context priority and omitting stale parameters or inaccessible records.
+- baf80fe: Persist partial tool approval decisions without executing tools or continuing the model until every approval in the step is answered. Automatically finalize saved terminal Runner conversations after final synchronization when their projection producer has restarted.
+- baf80fe: Describe Runner cancellation history as a request until termination is confirmed, preserving existing notice identifiers and cancellation behavior.
+- baf80fe: Read the chat Agent directory directly from the conversation service without changing the HTTP response or eligibility checks. Remove `chatAgents` from the OnlineAgent service, runtime contract, and constructor dependencies; server callers should use `conversations.chatAgents(userId)` instead.
+- baf80fe: Show shared chat progress and safe outcomes, follow growing replies only near the bottom, and align Native defaults, read state and cancellation confirmation across chat consumers.
+- baf80fe: Describe conversation configuration, native model selection and SDK message input in generated API and CLI help.
+- baf80fe: Save notices for native and Runner OnlineAgent conversations in the caller's transaction without waking an execution, including when input is supplied. Preserve active execution results and propagate persistence errors.
+
+  Reject direct HTTP tool, model-list and system-prompt overrides. Validate business variables against schemas declared by the selected scopes, while preserving trusted server SDK configuration and existing thinking handling.
+
+  Add showInChatList with a non-null true default. Filter ordinary chat lists before database pagination, allow explicit visibility at ordinary conversation creation, and always hide new background sessions. Known-session ownership checks and direct reads remain unchanged.
+
+- baf80fe: Keep conversation agent identifiers non-null with a default sentinel, keep historical conversations readable and read-only while permitting owners to cancel unfinished old runs, and allow application shutdown to complete with a warning if OnlineAgent finalization fails. Persist tool messages using the existing display roles while retaining their canonical SDK roles. Prefix OnlineAgent integration types, narrow server exports, and document the service and configuration API in the plugin Skill.
+- baf80fe: Wait for the admitted OnlineAgent execution in get(), including validated result persistence, finalization and occupancy release. Add try({ wait }) for bounded observation with distinct pending, failed and cancelled results and safe errors. Preserve run() admission semantics and finalization retries without rerunning model or tools.
+
+  Update translation and structured extraction examples to await results without status polling while preserving deadlines, disconnect handling and cancellation of late admission.
+
+- baf80fe: Use `OnlineAgentService` directly as the application service contract and remove redundant service interfaces. Update application dependencies, tests, and integration guidance to use the complete service type while retaining the shared realtime-session and Runner-backend contracts.
+- baf80fe: Integrate single-instance startup correction, transactional UI notifications and background result observation with standard SDK chat. Preserve approval barriers and actual tool occupancy while making interrupted-effect feedback advisory after finalization.
+- baf80fe: Repair previous-process native conversation occupancy during server startup while preserving saved results. Keep interrupted tool facts in private model history, remove effect-review admission locks, and allow owners to cancel historical runs while keeping historical conversations read-only. Stop continues to wait for actual native tool termination and finalization; cancellation requests for Runner and historical runs still require terminal confirmation.
+- baf80fe: Show OnlineAgent thinking feedback before the first stream event and place animated waiting feedback directly after the latest chat message in Examples and the Registry block.
+- baf80fe: Remove the unused legacy realtime protocol and process-local replay manager after migrating public chat consumers to SDK UI messages. Keep background structured execution and its private model evidence separate from browser streams, and verify the installed application contract.
+
+  Show writing indicators only for the active SDK response so saved partial replies stay still after cancellation, refresh and subsequent turns.
+
+- baf80fe: Store native realtime OnlineAgent messages once as complete SDK UIMessage originals. Derive public history and streams through shared safe projections and model history through current tool converters, enforce lossless JSON tool results, and preserve approval, recovery and finalization idempotency.
+- baf80fe: Keep committed chat messages, execution occupancy, paginated history and command results consistent across views. Confirm uncertain submissions into canonical history without resending or reconnecting, and guard delayed actions and Undo callbacks for the authenticated provider lifetime.
+
+  Keep published client test fixtures, navigation guards and language providers in the same Router and application contexts as application pages, and exercise translation model selection through its keyboard interaction.
+
+- baf80fe: Unify native realtime and background conversations on complete internal SDK UI messages, deriving safe public UI history and private model replay from the same original. Restore the common `OnlineAgent` interface with `RealtimeOnlineAgent` and `BackgroundOnlineAgent` subinterfaces and concrete session implementations. Background agents expose `run`, `get`, and `try` without `send` or browser-stream dependencies; structured results remain independently validated execution data. Update application consumers to use the background interface. Historical background ModelMessage storage is not supported by the new reader.
+- @nocobase/app-server@3.0.0-beta.0
+  - @nocobase/app-plugin-file@1.0.0-beta.25
+
+## 0.1.0-beta.14
+
+### Patch Changes
+
+- c640d35: Clarify runtime coding tool quota percentages as used capacity.
+- 1952789: Require explicitly declared realtime topics and authorize dynamic resources before subscription and delivery. Isolate the server core behind a broker contract, retain the application adapter and service tokens, and handle asynchronous publication throughout plugins and examples. Add acknowledged browser subscriptions, bounded recovery and heartbeat handling, shared React refresh helpers, original-session revalidation, transport limits, and usage documentation. The included adapter remains in-process; events are transient and clients recover by refetching after acknowledgement.
+- 1952789: Make mail synchronization and office workflow views refresh from committed state notifications, including empty sync results, failures, cancellation and lifecycle transitions. Add optional periodic reconciliation to the shared refresh scheduler, pause hidden views, and retain one-minute reconciliation for transient announcements and out-of-band changes. Reduce existing task, run, session and chat fallback reads to one minute while retaining time-based computer status checks, external indexing status reads, deployment health probes and bounded initialization waits.
+
+  Announce knowledge rechunking start and completion even when no chunks change, and support optional committed status notifications from external indexers. Continue tracking pending index work when some sections have already failed.
+
+  Preserve application providers when `renderWithApp` rerenders its page. Exercise mail refresh through real application services and fix the mail account navigation translation key exposed by those checks.
+
+- 1952789: Add `refreshRealtimeQueries` to refresh active TanStack Query snapshots after pre-event requests settle, including initial loads, prefixes and predicates. Keep inactive snapshots invalid, expose refresh failures, and stop queued follow-ups when their owner is disposed. The application client now declares its TanStack Query peer contract.
+
+  Use the shared catch-up behavior in Projects, Knowledge, Agents, Agent Sessions and the inbox. Refresh additional inbox topics after subscription readiness and preserve refresh promises so concurrent events coalesce correctly. Update application guidance to avoid reusing stale in-flight responses.
+
+- 1952789: Retry transient subscription failures independently with bounded exponential backoff and jitter. Preserve permanent authorization refusals and correlation checks, cancel retries on disposal, and probe healthy foreground connections instead of rebuilding them.
+
+  Retain stale snapshot state after failed reads and retry transient refresh failures without another realtime event. Expose refresh status, last successful synchronization, manual retry and separate subscription error callbacks. Clear previously authorized query data after an HTTP 401 or 403. Reuse the scheduler for Agent Sessions and subject run discovery while preserving slow reads and stream coalescing.
+
+- 7afd595: Remove Refine from the client runtime
+
+  The application shell no longer mounts Refine. `@nocobase/app-client` drops `app.refine`, `app.refineConfig`, `AppClientRefineConfig`, `AppClientRefineRegistry`, `AppClientRefineSetters`, `AppClientRefineSetterValue` and the provider-context methods `runWithProviderContext` and `runWithProviderContextAsync`, which existed only to attribute Refine settings to the provider that made them. Nothing in the repository read what was registered there: a title set through `app.refine.setOptions` never reached the page, and Refine resources never produced a menu entry.
+
+  Refine also created the TanStack Query cache that pages read through `useQueryClient()`; `AppClientProviders` now creates it in the same place, the same way, so pages and plugins behave as before. `@tanstack/react-query` becomes a peer dependency of `@nocobase/app-client` and joins the identity-sensitive packages that plugins must declare as peers.
+
+  `@nocobase/i18n/client` drops `createRefineI18nProvider`, `CreateRefineI18nProviderOptions` and `RefineI18nProvider`. `@nocobase/app-plugin-repository-example` drops its Client ServiceProvider and the `./client/providers` export, which only registered Refine resources.
+
+  `nocobase build` builds the client with `vite build` instead of `refine build`, and `@nocobase/app-cli` no longer lists `@refinedev/cli` as an optional peer. The Default and Examples templates remove every `@refinedev/*` development dependency. The development and plugin Skills no longer mention Refine, and the upgrade Skill explains how an existing application removes its `this.app.refine` calls and Refine packages.
+
+- 7356599: Vertically center runtime tool controls and quota details in their rows.
+- Updated dependencies [1952789]
+- Updated dependencies [1952789]
+- Updated dependencies [1952789]
+- Updated dependencies [1952789]
+- Updated dependencies [1952789]
+- Updated dependencies [7afd595]
+  - @nocobase/app-cli@1.0.0-beta.25
+  - @nocobase/app-client@4.0.0-beta.0
+  - @nocobase/app-plugin-authentication@2.0.0-beta.11
+  - @nocobase/app-plugin-authorization@1.0.0-beta.29
+  - @nocobase/app-plugin-file@1.0.0-beta.25
+  - @nocobase/app-server@3.0.0-beta.0
+  - @nocobase/i18n@2.0.0-beta.0
+  - @nocobase/authorization@1.0.0-beta.14
+  - @nocobase/db@1.0.0-beta.21
+  - @nocobase/markdown-mermaid@0.1.0-beta.2
+  - @nocobase/service-provider@0.0.2-beta.2
+
+## 0.1.0-beta.13
+
+### Minor Changes
+
+- 32ed5e2: Report coding account quota windows, filter scoped exhausted tools before claims, and restore eligibility at known reset times or after an authorized manual clear. Show quota status and usage in runtime tabs and agent availability, and batch page reads independently of the catalog size.
+
+  Runtime API and CLI: runtime get returns directory totals without directory rows; runtime list omits models, policy, variables, quota details and load, and caps takes at three with takesCount. Use the paginated runtime workspaces command for directory rows. Runtime quota clear takes structured tool/scope/target/window fields and no client revision; only unknown-reset marks may be cleared.
+
+  Classify quota revision races as retryable REVISION_CONFLICT while known-reset clears fail with QUOTA_RESET_KNOWN / FAILED_PRECONDITION.
+
+  Batch related-task visibility and missing blocker reads with the existing issue visibility filter, preserving per-request permissions and hidden blocker counts as task and project counts grow.
+
+### Patch Changes
+
+- Updated dependencies [22717d7]
+- Updated dependencies [32ed5e2]
+  - @nocobase/app-server@2.0.0-beta.9
+  - @nocobase/agent-protocol@0.1.0-beta.10
+  - @nocobase/app-plugin-file@1.0.0-beta.24
+  - @nocobase/app-plugin-authentication@2.0.0-beta.10
+  - @nocobase/app-cli-client@0.1.0-beta.10
+
+## 0.1.0-beta.12
+
+### Minor Changes
+
+- c28f719: The agent queue reads a fixed number of queries however many runs are held or queued: each queued row carries the claim rules' quick explanation, the newest activity of all held runs takes two reads, and the exact reason a queued run waits is read only when its row is unfolded (`RunService.queueWait`, `GET /api/agentBoard/issues/:issueId/agents/:agentId/wait`). The queue refreshes from coalesced announcements with a 45-second fallback poll, and the agents' work marks on issue lists refresh only for announcements about the issues shown, with a 60-second fallback poll.
+
+### Patch Changes
+
+- 14b7aec: Make the run retry back-off and the runners sweep interval configurable: `agents.runs.retryBaseDelayMs` sets the first back-off of a failed run attempt (30 seconds by default, still doubling up to ten minutes) and `agents.runners.sweepIntervalMs` how often the sweep runs (30 seconds by default). The defaults are unchanged; shorter values are meant for tests.
+- 56251bc: Reuse issue context, previous summaries, comment deltas and initialization reads for matching parameters within a transaction, reducing repeated work during runner claims.
+- ff3ce5d: Clarify that native agents must use the CLI for management operations while explicitly authorized independent test instances allow browser and HTTP verification, with existing credential and permission protections preserved.
+- f1f3bd7: Read the dashboard, usage and metrics reports with a fixed number of queries whatever the number of issues, runs and usage records: runs, usage and issues are read in one query each with only the columns the figures use, model usage outside runs is summed in the database, and issue facts are read in three queries instead of two per 200 issues. The Studio dashboard now loads each panel from its own route, `GET /api/reports/delivery`, `GET /api/reports/agents` and `GET /api/reports/projects` (CLI `report delivery`, `report agents`, `report projects`), each loading, failing and refreshing on its own; `GET /api/reports/dashboard` and `report dashboard` are removed.
+- 644a773: Reuse claim-local run, capacity, skill and sealed variable reads, batch candidate agent reads, and update routine runner activity timestamps without a separate transaction while preserving dispatch writes and delivery audits.
+- 8b9fff1: Read run status without a transaction when no new input needs delivery, while preserving cancellation, lease ownership checks, and input redelivery.
+- 616561b: Give dispatched agent runs five minutes to start while their runtime stays online. Immediately requeue server-reclaimed pre-start runtime interruptions without consuming failure attempts or losing unhandled input, stopping after ten consecutive interruptions. Count runner-reported worker startup failures against the normal failure budget with backoff, and distinguish terminal runtime failures from agent failures. Count and list only inbox decisions whose in-app messages still exist, filtering before pagination.
+- f2fa59b: Revoke all prior runner registrations and keys when the same owner registers the same name and host again, without a registration lock or new table. Simultaneous registrations may temporarily leave multiple active records, which the next registration revokes. Add a paginated runtime token list that exposes only usable token metadata within the caller's permissions, provide executable revoke/delete guidance, and require an explicit server for runner installation before any files are installed.
+- 7abd72b: Allow administrators to configure global and new-project skill defaults in the attachment section, and show visible attachment targets by name with project settings links. Align new-project defaults CLI access with the existing settings permissions.
+- c2132c9: Return only workspace totals from the runner list API and `runtime list`, read independent runner summaries concurrently, and fetch directory diagnostics when opening a runner's details. Scripts reading `workspaceUsage.workspaces` from the list must use `runtime get` or the single-runner API instead. The CLI table now includes directory counts, unpushed counts and free disk bytes.
+- e94b470: Send selected PNG and JPEG images through the existing conversation attachment flow when organizing requirements. Keep the source requirements and files in memory across navigation, with a manual return to edit or choose another agent after a reading failure.
+- 49f61b9: Show optional settings permissions separately from business actions in CLI identity output, and explain run-token-only permissions. Add available-command typo suggestions, a hint for truncated text, and clean shutdown when a pipe closes early. Describe route schema inputs so generated CLI help and reference tables explain their parameters. The manifest identity and whoami JSON result gain an optional settings list; command admission and authorization are unchanged.
+- e8f52c9: Keep variable and agent capability dialog actions reachable by limiting dialog height and scrolling long form content internally.
+- Updated dependencies [846e75f]
+- Updated dependencies [d34b6eb]
+- Updated dependencies [98b3862]
+- Updated dependencies [525ace4]
+- Updated dependencies [1cdbaf4]
+- Updated dependencies [5fe8773]
+- Updated dependencies [b22d999]
+- Updated dependencies [616561b]
+- Updated dependencies [83e3609]
+- Updated dependencies [00c5970]
+- Updated dependencies [0037dd2]
+- Updated dependencies [55fcee1]
+- Updated dependencies [c0b7537]
+- Updated dependencies [eb4b081]
+- Updated dependencies [14b7aec]
+- Updated dependencies [49f61b9]
+- Updated dependencies [0037dd2]
+  - @nocobase/app-cli-client@0.1.0-beta.8
+  - @nocobase/app-cli@1.0.0-beta.23
+  - @nocobase/db@1.0.0-beta.21
+  - @nocobase/app-client@3.0.0-beta.2
+  - @nocobase/app-server@2.0.0-beta.8
+  - @nocobase/agent-protocol@0.1.0-beta.9
+  - @nocobase/app-plugin-file@1.0.0-beta.24
+  - @nocobase/app-plugin-authentication@2.0.0-beta.10
+  - @nocobase/app-plugin-authorization@1.0.0-beta.28
+  - @nocobase/authorization@1.0.0-beta.14
+  - @nocobase/i18n@1.0.0-beta.7
+  - @nocobase/markdown-mermaid@0.1.0-beta.2
+  - @nocobase/service-provider@0.0.2-beta.2
+
+## 0.1.0-beta.11
+
+### Minor Changes
+
+- 862a542: Support global default skills and transactional defaults for newly created Studio projects. Resolve skills from global through application scopes to working directories in one query, with more specific targets winning. Stop reading and exposing legacy agent skill attachments while preserving agent variables.
+- aae3508: Clear finished runs' events and prompt snapshots hourly after a configurable retention period, defaulting to 14 days. Keep run history and delivery records, and show a retention notice when an expired run has no process data.
+
+### Patch Changes
+
+- 42aaf81: Append authorized project manager follow-ups to compatible active runs, deduplicate source comments, preserve comment authorization when dispatching multiple agents, and guard run completion with Collection optimistic revisions.
+- 8e4cf26: Recover runners and agent-session bridges after a protocol upgrade. Heartbeat answers now carry the server's runner protocol (`protocolVersion`), and a runner that speaks an older protocol while no newer version is offered logs what it speaks, what the application needs and how to update, instead of looking up to date. A bridge refused for its protocol now receives the protocols the application speaks in the error's `metadata`; `acp start` then updates the CLI as `update` does and exits non-zero so its launchd, systemd or Startup service starts it again on the new version, and when no update helps it still exits non-zero after a pause rather than stopping with success. A computer whose bridge was refused as too old is shown on My agents with the commands to update it.
+- facfe5c: Clear failures when a retried agent run is dispatched or succeeds, and show queued retry explanations separately from current failures while preserving historical errors.
+- c1d2601: Claim polls no longer lock or write anything for queued runs they cannot take: comment consent, identity and the agent's concurrency limit are decided from reads, and a transaction opens only for the run about to be claimed. Consent decisions use guarded updates on the request and on each run instead of locking the whole comment run chain, which had exhausted the database pool when several runners polled at once.
+- c1d2601: Require owner consent before another person's comment can run on the owner's personal runner. Team runners can take the queued work directly. Keep consent within the comment's run chain, show decisions in Studio's existing Needs you card and inbox, and let the original commenter run pending or rejected work as themselves.
+- 92afa27: Make My follow-ups fast for someone who owns many open issues. `GET /api/followUps` reads what every row's progress needs in a fixed number of queries instead of a few per issue, and a row no longer works out why queued work has not been claimed (`waiting`, `managerWait`): the issue's own progress (`GET /api/issueProgress/{issueId}`), shown beside the list, still does. A new `GET /api/followUps/issues?issueIds=…` answers the rows of the named issues, and the page uses it to refresh only the rows an announced change names, a burst at a time, reading the whole overview again only when the list itself changed, on reconnect, for a change that names no issue, and on its fallback poll. The follow star on an issue reads only what the viewer follows from a new `GET /api/followUps/following` (`follow-up following` on the command line) instead of the whole overview, and an agent session's issue progress reads its issues in the same fixed number of queries. The projects plugin's comment queries gain `newestByEach`, the agents plugin's runs gain `openOnEach`, and its run request list takes `subjectIds` and `statuses`.
+- 943a9a8: Fix MySQL installation of the agents and releases plugins by removing the unused directory/status index and limiting archive artifact keys to 172 characters while preserving full-value uniqueness. Add a Collection builder inspection method and forward migrations to converge existing installations without truncating data; oversized legacy artifact keys stop the upgrade before any column or index change. Rebuild the existing artifact-key unique index within the migration transaction on SQL Server so indexed legacy columns can be shortened while retaining uniqueness, nullable behavior and the actual index name under custom naming configurations.
+
+  This fix edits the released migrations `202610020003_ag_create_runs` and `202610020001_rel_create_tables` because their oversized indexes prevent fresh MySQL installations from reaching a later repair migration. Existing installations should expect checksum warnings on the next `nocobase db apply`; installations using `onChecksumMismatch: error` stop instead. After reviewing these exact changes, use `nocobase db repair --dry-run` and then `nocobase db repair` to acknowledge the new checksums, followed by `nocobase db apply` to perform the forward schema changes. Repairing checksums alone does not upgrade the schema. The forward corrections are irreversible because restoring the old indexes would make MySQL installations invalid again. See the plugins' `docs/mysql-installation-upgrade.md` for the upgrade and failed-installation recovery procedure.
+
+- 4626949: Start issue runs from the earliest linked open pull request in the same repository, allow agents to manage branches and pull request associations without rewriting pushed history, and preserve shared local branches and stashes when removing run worktrees. Only published runner-created temporary branches are cleaned up. A run that ends on another branch is preserved by that branch, without a saved head or push failure on the starting target.
+- c1d2601: Preserve owner-initiated comment delegation without extra personal runner consent and retain every unfinished target and input when the commenter chooses Run as me.
+- bdb0581: Reduce the server load of open issue and My follow-ups pages. Issue change announcements are gathered for a moment and name the issues that changed, so a page refreshes only what it shows of them; Studio announces new runs on an issue, so issue pages, What it waits for, owner handoffs and My follow-ups fall back to a slow poll instead of polling every few seconds; a preview label the server failed to set no longer makes the issue page poll every three seconds. Polled reads check issue visibility without loading the whole issue page (`IssueQueries.visible`), What it waits for reads only the project manager's newest comments (`CommentQueries.newestBy`), My follow-ups shares lookups across its issues, and pull request marks check visibility in one read. A runner heartbeat no longer announces a runner change, waking every waiting claim and refreshing every runtimes page, unless its disk sample moved by a gigabyte, crossed the free-space threshold or changed size.
+- f36da62: Keep a runner claiming work while it collects working directories. Collection no longer measures every directory's size by reading every file, which on runners with many directories took tens of minutes per round and held all claims meanwhile; only pruning the shared pnpm store still holds claims. Routine collection now runs an hour after the previous round ended (`WORKSPACE_REPORT_INTERVAL_MS`), still at once when the disk runs low, and the runner logs each round's start and duration. Workspace diagnostics show a directory's size only when a runner reported one.
+- 99be3d0: Use short private per-run temporary directories on Linux and macOS so Unix sockets can start, reserve TMP and TEMP alongside TMPDIR, and reclaim temporary files through the existing run lifecycle and garbage collection.
+- bdb0581: A public realtime topic can be defined with `authenticated: true` (`defineTopic` and `registerTopic` of `@nocobase/app-server/realtime`): subscribing without a signed-in principal is then refused with `AUTHENTICATION_REQUIRED`, as for a user topic, and publishing skips connections without one. The projects plugin's `pm:changes`, the agents plugin's `agents:runners` and Studio's `studio:issues:runs`, whose announcements name records by id, now reach signed-in connections only. The authentication plugin opens the realtime connection again when the signed-in person changes, so a page signed in without a reload receives these topics and user topics.
+- Updated dependencies [8e4cf26]
+- Updated dependencies [69e212c]
+- Updated dependencies [943a9a8]
+- Updated dependencies [943a9a8]
+- Updated dependencies [e23633e]
+- Updated dependencies [1cd749c]
+- Updated dependencies [f36da62]
+- Updated dependencies [bdb0581]
+- Updated dependencies [9b946f4]
+  - @nocobase/agent-protocol@0.1.0-beta.8
+  - @nocobase/app-cli-client@0.1.0-beta.7
+  - @nocobase/app-server@2.0.0-beta.7
+  - @nocobase/app-cli@1.0.0-beta.22
+  - @nocobase/db@1.0.0-beta.20
+  - @nocobase/app-plugin-authentication@2.0.0-beta.9
+  - @nocobase/app-plugin-file@1.0.0-beta.23
+  - @nocobase/app-plugin-authorization@1.0.0-beta.27
+
+## 0.1.0-beta.10
+
+### Minor Changes
+
+- ddfe1b1: Discover agents from what runners report instead of having people create them. Each coding tool and normalized model a runner reports becomes one agent (`source: 'runner'`, a deterministic id, named from a model catalog such as "Claude · Opus 5.5"), pending until someone who manages agents writes what it is good at and enables it (`PATCH /api/agents/{agentId}` with `description`, `effort`, `enabled`); an application keeps agents of its own with `agents.ensureSystem`, and sets every discovered agent's instructions and actions with `agents.setDefaults`. A model a run reports unavailable is marked on that runner until the runner reports its models again, and an agent no runner reported for 14 days is shown without a runner and left out of `roster`. Online agents, consultations, the online executor, agent presets, creating, archiving and deleting agents, the chat model choice and the default chat model settings are removed; conversations go to an enabled agent and say why one cannot answer (`othersPersonal` added, `AGENT_NOT_ENABLED`). Model services remain for the application's own model calls through `models.gateway` (renamed from `online.gateway`). Claims carry `workspace.key` and `workspace.readOnly` from the subject's assembly (`SubjectAssembly.workspaceKey`, `readOnly`). The migration `202610240020_ag_runner_agents` adds the new columns, removes every existing agent (rows kept and marked deleted so history still names them), cancels their queued runs, withdraws their pending run requests and clears default chat agents.
+- 6fc0af2: Require the issue owner to start or continue Studio project manager work, and ignore agent wakeups from comments on unstarted issues. Run the manager and its delegated steps as the current owner while preserving trigger audit information. Withdraw all queued agent work and request cancellation of held runs on owner transfers, without automatically restarting them. Correct legacy queued identities before claim and expose waiting reasons.
+
+  Expose the existing transactional `Projects.activity` recorder for application issue events and deliver owner transfers to all work kinds before ordinary triggers, including executor clearing. Restore Studio wording for models, efforts, the start dialog and historical sub-issue activity.
+
+  Preserve delegated execution identity throughout repeated retries, including handled inputs, and show the initiator of the current retry instead of copied retry history.
+
+- abc3fcb: Keep enabled agents prominent, group new discoveries by tool and catalog model, and persist shared ignore and restore preferences with revision protection. Choose a provider without changing existing execution identities and show personal runner names. Discover only authenticated Pi providers and prevent model overrides from bypassing that filter. In Studio, authorize runs through their running identity and existing command scope instead of intersecting with legacy agent action lists, including reports, while retaining object permissions, credential scopes, read-only modes and human approval requirements.
+
+  Show recognizable provider and runner labels when selecting a model and cap atomic bulk discovery operations at 256 selected groups with explicit counts and limits.
+
+### Patch Changes
+
+- 10e261b: Use a private linked Git worktree per run and save local continuation after stopping its tools. Fast-forward the resume reference, retain divergent heads as separately listed saved references, and preserve the directory on cleanup failure. Keep Codex state outside disposable worktrees. Other tools start fresh when their code directory changes; legacy sessions and clones are not migrated. Distinguish private local branches from task push targets and retain older runner compatibility.
+- ee24e95: Clean idle build artifacts before pushed workspaces when disk space is low, pause new work until heartbeats detect recovery, and report disk-full failures for reassignment without counting a project-manager step failure. Show current disk capacity and workspace size, type and push diagnostics.
+
+  Retry interrupted removal of verified pushed attempt directories from GC or the next run, stop their remaining processes before deleting remnants, and retain cleanup failure reasons without blocking later tasks. Preserve unpublished or uncommitted work for recovery.
+
+  Verify the actual checkout branch and HEAD before retrying removal, report disk exhaustion even when local preservation metadata or events cannot be written, and continue safe pressure cleanup when diagnostic cache writes fail.
+
+- 4748117: Allow the project manager to use an idle fallback after a configurable queue delay, exclude runners that cannot honor its workspace, and show selection and waiting reasons. Persist queue episodes through retries and restarts, withdraw unclaimed manager delegations after a configurable timeout, reliably report their cause in the dispatch thread, and publish identity-aware immediate capacity in each manager roster.
+- fad531f: Initialize the project manager's name and description once in the installation's default language and allow administrators to edit them, including clearing the description. Restarts and unrelated settings saves preserve the identity. Later changes to the installation language do not rename an initialized manager.
+
+  An upgrade seed preserves valid explicit settings, custom legacy agent values, empty descriptions, and unknown settings. Legacy agent values identical to the former built-in English defaults cannot be distinguished from manual writes and are initialized in the upgrade language when no valid explicit setting exists. A damaged `studioSettings.projectManager.value` stops initialization with a repair instruction instead of overwriting it.
+
+  Resolve historical run inputs, notifications, delegation cards, plans, and consultation identity labels from stable agent IDs without rewriting historical text or execution snapshots. Legacy limit notices and plans whose identity cannot be recovered use neutral labels, and unresolved consultations mark the originally requested subject. Refresh identity caches when the manager is renamed.
+
+  Mount application query effects inside the projects plugin's cache using its optional `queryEffects` composition point. Keep completed consultation labels current after renames and distinguish loading, inaccessible agents, and historical request snapshots.
+
+- a2cf036: Interrupt agent runs for runner updates after stopping tools and preserving local continuation references. Requeue upgrade interruptions without consuming failure attempts or notifying project managers, retain task inputs, and report draining capacity and reasons in runner and queue views.
+- Updated dependencies [157a914]
+- Updated dependencies [8698e80]
+- Updated dependencies [10e261b]
+- Updated dependencies [9e99702]
+- Updated dependencies [ddfe1b1]
+- Updated dependencies [ee24e95]
+- Updated dependencies [4748117]
+- Updated dependencies [a2cf036]
+- Updated dependencies [6758f9f]
+  - @nocobase/app-cli-client@0.1.0-beta.6
+  - @nocobase/agent-protocol@0.1.0-beta.7
+  - @nocobase/app-server@2.0.0-beta.6
+  - @nocobase/app-plugin-authentication@2.0.0-beta.8
+  - @nocobase/app-plugin-file@1.0.0-beta.23
+
 ## 0.1.0-beta.9
 
 ### Patch Changes