Bump sagents from 0.7.0 to 0.11.0

[dependabot]

Bumps sagents from 0.7.0 to 0.11.0.

Release notes

Sourced from sagents's releases.

v0.11.0

What's Changed

• Keep an open ask_user question alive when the agent sleeps by @​brainlid in sagents-ai/sagents#159

Full Changelog: sagents-ai/sagents@v0.10.2...v0.11.0

v0.10.2

What's Changed

• Let a sub-agent run without publishing its inner work by @​lostbean in sagents-ai/sagents#158
• prep for v0.10.2 by @​brainlid in sagents-ai/sagents#160

Full Changelog: sagents-ai/sagents@v0.10.1...v0.10.2

v0.10.1

What's Changed

• Fix ExDoc reference warnings in AgentServer docstrings by @​brainlid in sagents-ai/sagents#154
• Add :otel_attributes for application context on agent spans by @​brainlid in sagents-ai/sagents#155

Full Changelog: sagents-ai/sagents@v0.10.0...v0.10.1

v0.10.0

What's Changed

• Bump actions/checkout from 6.0.2 to 7.0.0 by @​dependabot[bot] in sagents-ai/sagents#132
• Bump actions/cache from 5.0.5 to 6.1.0 by @​dependabot[bot] in sagents-ai/sagents#138
• Bump actions/cache/restore from 5.0.5 to 6.1.0 by @​dependabot[bot] in sagents-ai/sagents#137
• Bump actions/cache/save from 5.0.5 to 6.1.0 by @​dependabot[bot] in sagents-ai/sagents#136
• Add transform_display_message middleware hook for shaping display messages by @​brainlid in sagents-ai/sagents#139
• Preserve whitespace-only content when creating a file by @​brainlid in sagents-ai/sagents#147
• Bump erlef/setup-beam from 1.24.0 to 1.24.1 by @​dependabot[bot] in sagents-ai/sagents#140
• Return a named error when resuming a tool-raised subagent interrupt by @​lostbean in sagents-ai/sagents#144
• Propagate a sub-agent :halt to the parent as a halt by @​brainlid in sagents-ai/sagents#150
• Stop sub-agents inheriting AskUserQuestion middleware by @​brainlid in sagents-ai/sagents#151
• Extract the matched tool result on an until_tool termination by @​lostbean in sagents-ai/sagents#143
• Carry the pause cause from the mode to the paused status event by @​lostbean in sagents-ai/sagents#149
• Add a pending-message queue for mid-run user and tool messages by @​brainlid in sagents-ai/sagents#152
• prep for v0.10.0 by @​brainlid in sagents-ai/sagents#153

Full Changelog: sagents-ai/sagents@v0.9.0...v0.10.0

v0.9.0

What's Changed

• Add Horde participation-based membership with partition support by @​brainlid in sagents-ai/sagents#134
• prep for v0.9.0 release by @​brainlid in sagents-ai/sagents#135

Full Changelog: sagents-ai/sagents@v0.8.0...v0.9.0

... (truncated)

Changelog

Sourced from sagents's changelog.

v0.11.0

An open ask_user question no longer disappears when the agent goes to sleep.

The interrupt was always durable — it is persisted with the state, and a fresh boot rebuilds it and comes up :interrupted. The loss was in the host UI, which treated "the process went away" as a status change and cleared the prompt. This release separates the two facts hosts were conflating: agent_status is what the conversation is waiting on, agent_alive? is whether a process is backing it right now. Answering an interrupt now goes through Sagents.Session.resume/4, which wakes a sleeping agent and hands it the answer at boot.

The same is true of any other restorable interrupt, including a :halt panel, whose Dismiss button goes through the matching Sagents.Session.dismiss/3. If your app renders a halt panel, read step 6 of the migration guide, since the button has to be routed through the new path to work on a dormant conversation.

No compile-time breakage, and upgrading without touching host code is safe: everything behaves as v0.10.1 did, bug included. The fix is opt-in, because the affected modules were generated into your app. See https://github.com/sagents-ai/sagents/blob/main/MIGRATION_PROMPT_v0.10.x_TO_v0.11.0.md — it is written to be handed to a coding agent.

Full write-up in #159.

Added

• Sagents.Session.resume/4 — answer an interrupt, waking the agent if needed. #159
• Sagents.Session.dismiss/3 — acknowledge a terminal :halt, waking the agent if needed. The mirror of resume/4 for the interrupt type that is dismissed rather than answered. A halt is restorable, so its panel now survives a nap, and AgentServer.dismiss_interrupt/1 alone cannot clear one on a dormant conversation. An interrupt that needs a real response is passed through as an error rather than woken. The generated Coordinator gains dismiss_agent_session/2 and the generated AgentLiveHelpers gains handle_halt_dismissal/1. #159
• :pending_resume, a start option applied during boot before the initial status broadcast, so a woken agent announces :running rather than an :interrupted snapshot it is about to leave. #159
• Sagents.AgentUtils.shutdown_session_changes/2 and the agent_alive? subscriber-state key. #159
• Sagents.State.interrupt_restorable?/2 is now public — the authoritative predicate for whether an interrupt survives a cold boot. #159
• {:agent_shutdown, _} payloads carry :interrupt_restorable. #159

... (truncated)

Commits

• 7d2b400 prep for v0.11.0 (#161)
• b110042 Keep an open ask_user question alive when the agent sleeps (#159)
• ed4db87 prep for v0.10.2 (#160)
• 2113e33 Let a sub-agent run without publishing its inner work (#158)
• d2099d5 Add :otel_attributes for application context on agent spans (#155)
• dd86d3d Fix ExDoc reference warnings in AgentServer docstrings (#154)
• a018cc7 prep for v0.10.0 (#153)
• a7b1d66 support ability for a tool to queue a user message (#152)
• 504133f Carry the pause cause from the mode to the paused status event (#149)
• 2a93497 Extract the matched tool result on an until_tool termination (#143)
• Additional commits viewable in compare view

You can trigger a rebase of this PR by commenting @dependabot rebase.

---

Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

• @dependabot rebase will rebase this PR
• @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
• @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
• @dependabot ignore this major version will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this minor version will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this dependency will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)

Note
Automatic rebases have been disabled on this pull request as it has been open for over 30 days.

[nelsonkopliku]

This requires changes to the backend. Specifically what AgentServer.subscribe(agent_id) returns.

With the new version returns {:ok, pid(), ref()} instead of only :ok

[Copilot]

Pull request overview

Updates Sagents from 0.7.0 to 0.11.0 and adapts agent subscriptions to its direct-delivery API.

Changes:

• Bumps the Sagents dependency and lockfile.
• Updates subscription contracts and mocks.
• Adds real-stack subscription coverage.

Reviewed changes

Copilot reviewed 5 out of 6 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
mix.exs	Updates the Sagents requirement.
mix.lock	Locks Sagents 0.11.0.
lib/trento/ai/agent.ex	Handles direct subscription results.
lib/trento/ai/agent/server.ex	Updates the subscription callback contract.
test/trento/ai/agent_test.exs	Updates mocks and adds integration coverage.
test/trento_web/channels/ai_assistant_channel_test.exs	Updates subscription mocks.

Suppressed comments (1)

test/trento/ai/agent_test.exs:405

• Remove the duplicated space in this comment.

  # Points the LLM at a closed local port so the run starts  and then fails fast

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[dependabot]

A newer version of sagents exists, but since this PR has been edited by someone other than Dependabot I haven't updated it. You'll get a PR for the updated version as normal once this PR is merged.