Repository navigation
Version Packages - #40
Merged
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This PR was opened by the Changesets release GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.
Releases
@pwrdrvr/agent-acp@0.16.1
Patch Changes
384493a: Make a low-effort reasoning request actually lower an ACP agent's thinking when that agent offers only graded levels, and choose the level from the session's own option menu rather than whichever session wrote last.
Three fixes, which only work together. Each is guarded by its own test.
Graded thought levels.
reasoningValueForThoughtLevelmapped a low-effort token only onto an off-like value (off/none/disabled/no). An agent that offers graded levels and no off got nothing. That covers Kimi's K3, whose thinking option islow/high/max/on. A low-effort token now falls back to thelowlevel (alsominimal/min/minimum) when there is no off. Off still wins where both exist, and a value that already read as on (say, "On (minimal budget)") still reads as on. Before this,reasoning: "low"changed nothing on any Kimi model: K2.7 Coding offers onlyon, and K3'slownever matched.Per-session option menus. A pooled
AcpAgentClientserves many sessions, but it kept one client-wideruntimeCapabilities, overwritten by everysession/new, option write, and load.applyReasoning(threadId, …)andsetModel(threadId, …)read that shared snapshot. Once sessions can be on different models (0.16.0), and the model decides which thought levels exist, a session could be matched against another session's menu. For example, an enrichment on K3 with a chat thread opened after it on K2.7 got alowrequest matched against[on], and no write was sent. The reverse sentthinking=lowto a K2.7 session, which the agent refuses. Each session now keeps its own menus, taken from itssession/new/session/loadresult and refreshed by its own option writes.setModel,applyReasoning, and a session's runtime-capability notifications read those. A session's menus are built on whatinitializereported, never on the client-wide snapshot. Before, a new session inherited any menu itssession/newomitted from whichever session wrote last, current values included, and could skip a write it believed was already applied. The client-wide snapshot remains for agent-level questions (session load, MCP transport support). An option write's reply now updates the session's runtime state as well as its menus, so options the write reshaped report where the agent actually put them.config_option_updatein the ACP 1.3 shape. ACP 1.3 defines this update as the fullconfigOptionsset. The parser read only a single changed option (update.configOption), so the spec shape was dropped entirely. Kimi emits it after every model change, and it is the only reportsession/set_modelgives (that call answers{}). The full set now replaces the session's menus and current values, includingcurrentModelId, however the update is keyed (sessionUpdate,session_update,kind,type). That means a model the agent changed no longer loses to a stale requested one inthread_settings.Measured against Kimi Code CLI 2.0.0, with two sessions on one client and the prompt refused locally so no model turn ran:
startThread({ model: "kimi-code/k3" }), then a chat thread on the default model, thenstartTurn({ reasoning: "low" })on the first. 0.16.0 sends no thinking write. This change sendsthinking=low, and Kimi's reply confirmsmodel=kimi-code/k3 thinking=low.