Repository navigation
feat(auto-advisor): decision spacing, explicit-call visibility, models rename; bump dev pins - #103
Merged
Merged
Conversation
…y package @opencode/plugin 2.2.0 peers OpenTUI at exactly 0.7.0 and 2.2.5 peers >=0.8.0, so the two move together. 2.2.5 ships OpenTUI 0.8.0 (0.8.1 came after it). Dev pins stay identical across packages so Dependabot keeps one target; peer ranges are unchanged.
Context: the OpenCode adapter closes an event stream on a failed read, so the listener that drops stale consults ended for good. It now re-subscribes after RESUBSCRIBE_MS and stops only on done (plugin unload). The fake stream in the server tests closes on error like the real adapter; AGENTS.md gotcha updated.
Both model chains now share one shape: a single provider/model string or an ordered list. routing.models gains the single-string form. The old key is rejected as unknown.
… steps Context: since 2026-11-01 Zen rate-limits jev-1.13-free, and the 60 s skip sent 86% of decisions to the paid fallback; 28% of calls ran on steps a limit already blocked. routing.minStepsBetweenDecisions (default 10) spaces calls across the session, and blocked steps log a line without calling. The server test harness now writes a base config that decides on every step.
… JSONC config Explicit advisor() calls now write a kind: "explicit" telemetry line with every Advisor model tried, and the review panel names the model that answered the last one. auto-advisor.json accepts comments and trailing commas (Bun.JSONC), and the README shows the full default configuration, checked against DEFAULTS by a test.
Owner
Author
|
/devin review |
|
Starting Devin Review. |
![devin-ai-integration[bot]](https://avatars.githubusercontent.com/in/811515?s=60&v=4){kind=small}
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
…ry explicit consult The timeline crashed on kind: "explicit" telemetry lines (no verdict) and let them take the next step's messages; it now shows each call as its own one-line row. An explicit call that failed before the Advisor chain ran (invalid config, failed history read) wrote no telemetry line; every call past the refusals now logs exactly once, and any Advisor record where no model answered carries advisor.error. Explicit ms measures the Advisor only, like automatic consults.
Owner
Author
|
/devin review |
|
Starting Devin Review. |
Owner
Author
|
/devin review |
|
Starting Devin Review. |
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.
Summary
@opencode/plugin2.2.5 and OpenTUI 0.8.0 in every package. These are dev-only (peer ranges unchanged). The builtdist/matches npm@latestapart from Bun's variable renaming, so no changesets for the other packages.minor):advisor.model→advisor.models, which accepts a string or a list;routing.modelsalso accepts a single string.routing.minStepsBetweenDecisions(default 10). Steps already blocked by the cooldown or budget no longer make a Decision call. This cuts paid-fallback calls while the free decision model is rate-limited.advisor()call writes akind: "explicit"telemetry line, and the review panel showsLast advisor() call: <model> · <seconds> s · <time>.auto-advisor.jsonis JSONC.DEFAULTS.Verification
bun run typecheck,bun run lint,bun run test(all 7 packages; auto-advisor 222 pass) andbun run buildall pass atae18653.ae18653:bun run smokepasses for auto-advisor, output-styles, ram-monitor and adversarial-review. agents-loader and double-tap-timeline have no smoke script; their built entries load through the package entry and export a V2Plugin.define(id+setup), butsetupwas not run against a host.outcome: "error"and the attempt error "the consultation was cancelled".Not verified
Follow-ups
@opencode/pluginto 2.2.2: close it, or@dependabot recreateafter this merges.outcome: "cancelled"instead of"error".