Repository navigation
Conversation
![cubic-dev-ai[bot]](https://avatars.githubusercontent.com/in/1082092?s=60&v=4){kind=small}
There was a problem hiding this comment.
3 issues found across 5 files
Prompt for AI agents (unresolved issues)
Check if these issues are valid — if so, understand the root cause of each and fix them. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. If appropriate, use sub-agents to investigate and fix each issue separately.
<file name="site/index.html">
<violation number="1" location="site/index.html:107">
P2: This applies the legacy `logprobs`/`top_logprobs` requirement to every backend. Native `backend: "systemone"` does not need it, so scope the requirement to the legacy `backend: "local"` chat path.</violation>
<violation number="2" location="site/index.html:107">
P1: This FAQ answer now says v3 is not supported, which contradicts the README and the local setup instructions on this same page. Restore the original answer: v3 works with `backend: "systemone"` on LM Studio 0.4.26+ or a compatible System One server.</violation>
</file>
<file name="site/robots.txt">
<violation number="1" location="site/robots.txt:4">
P2: This `robots.txt` sits under the project path, so crawlers will never fetch it. Only a `robots.txt` at the host root (`overbit.github.io/robots.txt`) is honored, so the `Sitemap:` line has no effect. Submit `sitemap.xml` directly in Search Console or remove the claimed discovery benefit from the PR description.</violation>
</file>
Reply to a comment to ask cubic a question or push back. It learns from your replies.
Click Fix with cubic on a comment, or reply @cubic-dev-ai fix this, and cubic writes the fix.
View guided diff | Turn on auto-fix | Re-trigger cubic
| <section class="section wrap reference" aria-labelledby="reference-title"><div class="reference-copy"><h2 id="reference-title">Conservative settings.<br>Explicit limits.</h2><p>The plugin supplies a deterministic checkpoint string through OpenCode V2’s compaction hook. Retained text is not model-rewritten, but the original message structure is flattened.</p><p>This is an experimental port, not an arbitrary replacement-message API.</p><a class="text-link" href="https://github.com/overbit/fast-jev-compaction-opencode#configuration">Explore all configuration options <span aria-hidden="true">↗</span></a></div><dl class="defaults"><div><dt>Recent messages pinned</dt><dd>6 <code>preserveRecentMessages</code></dd></div><div><dt>Keep threshold</dt><dd>0.5 <code>keepThreshold</code></dd></div><div><dt>Minimum reduction</dt><dd>25% <code>minReductionRatio</code></dd></div><div><dt>Truncated result head</dt><dd>300 chars <code>truncateHeadChars</code></dd></div></dl></section> | ||
|
|
||
| <section class="wrap troubleshooting" aria-labelledby="troubleshooting-title"><h2 id="troubleshooting-title">When something doesn’t compact.</h2><details><summary>No request reaches the local model</summary><p>A classification request needs at least one completed tool call outside the pinned recent-message window. Short sessions may correctly send no request. Check the diagnostic log for candidate counts before changing your configuration.</p></details><details><summary>The local classifier falls back</summary><p>Check that the System One endpoint is reachable, the requested model is a decision model, and its loaded context is large enough. The legacy chat backend additionally requires <code>logprobs</code> and <code>top_logprobs</code>. Every compaction writes an outcome reason to <code>~/.local/share/opencode/log/fast-jev-compaction.log</code>.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#troubleshooting">See diagnostic reasons and smoke-test configuration</a></details><details><summary>Can I use the smaller 0.8B v3 model?</summary><p>Yes. Use <code>backend: "systemone"</code> with LM Studio 0.4.26+ or another compatible System One server and the identifier of a loaded decision model. The legacy <code>backend: "local"</code> chat path is still intended for v1 models.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#local-models">Read the model compatibility notes</a></details></section> | ||
| <section class="wrap troubleshooting" aria-labelledby="troubleshooting-title"><h2 id="troubleshooting-title">OpenCode /compact troubleshooting.</h2><details><summary>No request reaches the local model</summary><p>A classification request needs at least one completed tool call outside the pinned recent-message window. Short sessions may correctly send no request. Check the diagnostic log for candidate counts before changing your configuration.</p></details><details><summary>The local classifier falls back</summary><p>Check that the endpoint is reachable, the loaded context is large enough, and the model exposes <code>logprobs</code> and <code>top_logprobs</code>. Every compaction writes an outcome reason to <code>~/.local/share/opencode/log/fast-jev-compaction.log</code>.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#troubleshooting">See diagnostic reasons and smoke-test configuration</a></details><details><summary>Can I use the smaller 0.8B v3 model?</summary><p>Not through this plugin’s current local adapter. The v3 model needs its dedicated scoring runtime / System One-compatible API, which is not directly wired into this plugin yet.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#local-models">Read the model compatibility notes</a></details></section> |
There was a problem hiding this comment.
P1: This FAQ answer now says v3 is not supported, which contradicts the README and the local setup instructions on this same page. Restore the original answer: v3 works with backend: "systemone" on LM Studio 0.4.26+ or a compatible System One server.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At site/index.html, line 107:
<comment>This FAQ answer now says v3 is not supported, which contradicts the README and the local setup instructions on this same page. Restore the original answer: v3 works with `backend: "systemone"` on LM Studio 0.4.26+ or a compatible System One server.</comment>
<file context>
@@ -82,7 +104,7 @@ <h1 id="hero-title">Keep the context.<br>Not the clutter.</h1>
<section class="section wrap reference" aria-labelledby="reference-title"><div class="reference-copy"><h2 id="reference-title">Conservative settings.<br>Explicit limits.</h2><p>The plugin supplies a deterministic checkpoint string through OpenCode V2’s compaction hook. Retained text is not model-rewritten, but the original message structure is flattened.</p><p>This is an experimental port, not an arbitrary replacement-message API.</p><a class="text-link" href="https://github.com/overbit/fast-jev-compaction-opencode#configuration">Explore all configuration options <span aria-hidden="true">↗</span></a></div><dl class="defaults"><div><dt>Recent messages pinned</dt><dd>6 <code>preserveRecentMessages</code></dd></div><div><dt>Keep threshold</dt><dd>0.5 <code>keepThreshold</code></dd></div><div><dt>Minimum reduction</dt><dd>25% <code>minReductionRatio</code></dd></div><div><dt>Truncated result head</dt><dd>300 chars <code>truncateHeadChars</code></dd></div></dl></section>
- <section class="wrap troubleshooting" aria-labelledby="troubleshooting-title"><h2 id="troubleshooting-title">When something doesn’t compact.</h2><details><summary>No request reaches the local model</summary><p>A classification request needs at least one completed tool call outside the pinned recent-message window. Short sessions may correctly send no request. Check the diagnostic log for candidate counts before changing your configuration.</p></details><details><summary>The local classifier falls back</summary><p>Check that the System One endpoint is reachable, the requested model is a decision model, and its loaded context is large enough. The legacy chat backend additionally requires <code>logprobs</code> and <code>top_logprobs</code>. Every compaction writes an outcome reason to <code>~/.local/share/opencode/log/fast-jev-compaction.log</code>.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#troubleshooting">See diagnostic reasons and smoke-test configuration</a></details><details><summary>Can I use the smaller 0.8B v3 model?</summary><p>Yes. Use <code>backend: "systemone"</code> with LM Studio 0.4.26+ or another compatible System One server and the identifier of a loaded decision model. The legacy <code>backend: "local"</code> chat path is still intended for v1 models.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#local-models">Read the model compatibility notes</a></details></section>
+ <section class="wrap troubleshooting" aria-labelledby="troubleshooting-title"><h2 id="troubleshooting-title">OpenCode /compact troubleshooting.</h2><details><summary>No request reaches the local model</summary><p>A classification request needs at least one completed tool call outside the pinned recent-message window. Short sessions may correctly send no request. Check the diagnostic log for candidate counts before changing your configuration.</p></details><details><summary>The local classifier falls back</summary><p>Check that the endpoint is reachable, the loaded context is large enough, and the model exposes <code>logprobs</code> and <code>top_logprobs</code>. Every compaction writes an outcome reason to <code>~/.local/share/opencode/log/fast-jev-compaction.log</code>.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#troubleshooting">See diagnostic reasons and smoke-test configuration</a></details><details><summary>Can I use the smaller 0.8B v3 model?</summary><p>Not through this plugin’s current local adapter. The v3 model needs its dedicated scoring runtime / System One-compatible API, which is not directly wired into this plugin yet.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#local-models">Read the model compatibility notes</a></details></section>
<section class="closing wrap"><h2>Keep what matters.<br>Inspect how it works.</h2><a class="button primary" href="https://github.com/overbit/fast-jev-compaction-opencode">Explore the source<svg viewBox="0 0 20 20" width="18" height="18" aria-hidden="true"><path d="M5 15 15 5M5 5h10v10" fill="none" stroke="currentColor" stroke-width="1.5"/></svg></a></section>
</main>
</file context>
Suggested change
| <section class="wrap troubleshooting" aria-labelledby="troubleshooting-title"><h2 id="troubleshooting-title">OpenCode /compact troubleshooting.</h2><details><summary>No request reaches the local model</summary><p>A classification request needs at least one completed tool call outside the pinned recent-message window. Short sessions may correctly send no request. Check the diagnostic log for candidate counts before changing your configuration.</p></details><details><summary>The local classifier falls back</summary><p>Check that the endpoint is reachable, the loaded context is large enough, and the model exposes <code>logprobs</code> and <code>top_logprobs</code>. Every compaction writes an outcome reason to <code>~/.local/share/opencode/log/fast-jev-compaction.log</code>.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#troubleshooting">See diagnostic reasons and smoke-test configuration</a></details><details><summary>Can I use the smaller 0.8B v3 model?</summary><p>Not through this plugin’s current local adapter. The v3 model needs its dedicated scoring runtime / System One-compatible API, which is not directly wired into this plugin yet.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#local-models">Read the model compatibility notes</a></details></section> | |
| <section class="wrap troubleshooting" aria-labelledby="troubleshooting-title"><h2 id="troubleshooting-title">OpenCode /compact troubleshooting.</h2><details><summary>No request reaches the local model</summary><p>A classification request needs at least one completed tool call outside the pinned recent-message window. Short sessions may correctly send no request. Check the diagnostic log for candidate counts before changing your configuration.</p></details><details><summary>The local classifier falls back</summary><p>Check that the System One endpoint is reachable, the requested model is a decision model, and its loaded context is large enough. The legacy chat backend additionally requires <code>logprobs</code> and <code>top_logprobs</code>. Every compaction writes an outcome reason to <code>~/.local/share/opencode/log/fast-jev-compaction.log</code>.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#troubleshooting">See diagnostic reasons and smoke-test configuration</a></details><details><summary>Can I use the smaller 0.8B v3 model?</summary><p>Yes. Use <code>backend: "systemone"</code> with LM Studio 0.4.26+ or another compatible System One server and the identifier of a loaded decision model. The legacy <code>backend: "local"</code> chat path is still intended for v1 models.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#local-models">Read the model compatibility notes</a></details></section> |
| <section class="section wrap reference" aria-labelledby="reference-title"><div class="reference-copy"><h2 id="reference-title">Conservative settings.<br>Explicit limits.</h2><p>The plugin supplies a deterministic checkpoint string through OpenCode V2’s compaction hook. Retained text is not model-rewritten, but the original message structure is flattened.</p><p>This is an experimental port, not an arbitrary replacement-message API.</p><a class="text-link" href="https://github.com/overbit/fast-jev-compaction-opencode#configuration">Explore all configuration options <span aria-hidden="true">↗</span></a></div><dl class="defaults"><div><dt>Recent messages pinned</dt><dd>6 <code>preserveRecentMessages</code></dd></div><div><dt>Keep threshold</dt><dd>0.5 <code>keepThreshold</code></dd></div><div><dt>Minimum reduction</dt><dd>25% <code>minReductionRatio</code></dd></div><div><dt>Truncated result head</dt><dd>300 chars <code>truncateHeadChars</code></dd></div></dl></section> | ||
|
|
||
| <section class="wrap troubleshooting" aria-labelledby="troubleshooting-title"><h2 id="troubleshooting-title">When something doesn’t compact.</h2><details><summary>No request reaches the local model</summary><p>A classification request needs at least one completed tool call outside the pinned recent-message window. Short sessions may correctly send no request. Check the diagnostic log for candidate counts before changing your configuration.</p></details><details><summary>The local classifier falls back</summary><p>Check that the System One endpoint is reachable, the requested model is a decision model, and its loaded context is large enough. The legacy chat backend additionally requires <code>logprobs</code> and <code>top_logprobs</code>. Every compaction writes an outcome reason to <code>~/.local/share/opencode/log/fast-jev-compaction.log</code>.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#troubleshooting">See diagnostic reasons and smoke-test configuration</a></details><details><summary>Can I use the smaller 0.8B v3 model?</summary><p>Yes. Use <code>backend: "systemone"</code> with LM Studio 0.4.26+ or another compatible System One server and the identifier of a loaded decision model. The legacy <code>backend: "local"</code> chat path is still intended for v1 models.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#local-models">Read the model compatibility notes</a></details></section> | ||
| <section class="wrap troubleshooting" aria-labelledby="troubleshooting-title"><h2 id="troubleshooting-title">OpenCode /compact troubleshooting.</h2><details><summary>No request reaches the local model</summary><p>A classification request needs at least one completed tool call outside the pinned recent-message window. Short sessions may correctly send no request. Check the diagnostic log for candidate counts before changing your configuration.</p></details><details><summary>The local classifier falls back</summary><p>Check that the endpoint is reachable, the loaded context is large enough, and the model exposes <code>logprobs</code> and <code>top_logprobs</code>. Every compaction writes an outcome reason to <code>~/.local/share/opencode/log/fast-jev-compaction.log</code>.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#troubleshooting">See diagnostic reasons and smoke-test configuration</a></details><details><summary>Can I use the smaller 0.8B v3 model?</summary><p>Not through this plugin’s current local adapter. The v3 model needs its dedicated scoring runtime / System One-compatible API, which is not directly wired into this plugin yet.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#local-models">Read the model compatibility notes</a></details></section> |
There was a problem hiding this comment.
P2: This applies the legacy logprobs/top_logprobs requirement to every backend. Native backend: "systemone" does not need it, so scope the requirement to the legacy backend: "local" chat path.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At site/index.html, line 107:
<comment>This applies the legacy `logprobs`/`top_logprobs` requirement to every backend. Native `backend: "systemone"` does not need it, so scope the requirement to the legacy `backend: "local"` chat path.</comment>
<file context>
@@ -82,7 +104,7 @@ <h1 id="hero-title">Keep the context.<br>Not the clutter.</h1>
<section class="section wrap reference" aria-labelledby="reference-title"><div class="reference-copy"><h2 id="reference-title">Conservative settings.<br>Explicit limits.</h2><p>The plugin supplies a deterministic checkpoint string through OpenCode V2’s compaction hook. Retained text is not model-rewritten, but the original message structure is flattened.</p><p>This is an experimental port, not an arbitrary replacement-message API.</p><a class="text-link" href="https://github.com/overbit/fast-jev-compaction-opencode#configuration">Explore all configuration options <span aria-hidden="true">↗</span></a></div><dl class="defaults"><div><dt>Recent messages pinned</dt><dd>6 <code>preserveRecentMessages</code></dd></div><div><dt>Keep threshold</dt><dd>0.5 <code>keepThreshold</code></dd></div><div><dt>Minimum reduction</dt><dd>25% <code>minReductionRatio</code></dd></div><div><dt>Truncated result head</dt><dd>300 chars <code>truncateHeadChars</code></dd></div></dl></section>
- <section class="wrap troubleshooting" aria-labelledby="troubleshooting-title"><h2 id="troubleshooting-title">When something doesn’t compact.</h2><details><summary>No request reaches the local model</summary><p>A classification request needs at least one completed tool call outside the pinned recent-message window. Short sessions may correctly send no request. Check the diagnostic log for candidate counts before changing your configuration.</p></details><details><summary>The local classifier falls back</summary><p>Check that the System One endpoint is reachable, the requested model is a decision model, and its loaded context is large enough. The legacy chat backend additionally requires <code>logprobs</code> and <code>top_logprobs</code>. Every compaction writes an outcome reason to <code>~/.local/share/opencode/log/fast-jev-compaction.log</code>.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#troubleshooting">See diagnostic reasons and smoke-test configuration</a></details><details><summary>Can I use the smaller 0.8B v3 model?</summary><p>Yes. Use <code>backend: "systemone"</code> with LM Studio 0.4.26+ or another compatible System One server and the identifier of a loaded decision model. The legacy <code>backend: "local"</code> chat path is still intended for v1 models.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#local-models">Read the model compatibility notes</a></details></section>
+ <section class="wrap troubleshooting" aria-labelledby="troubleshooting-title"><h2 id="troubleshooting-title">OpenCode /compact troubleshooting.</h2><details><summary>No request reaches the local model</summary><p>A classification request needs at least one completed tool call outside the pinned recent-message window. Short sessions may correctly send no request. Check the diagnostic log for candidate counts before changing your configuration.</p></details><details><summary>The local classifier falls back</summary><p>Check that the endpoint is reachable, the loaded context is large enough, and the model exposes <code>logprobs</code> and <code>top_logprobs</code>. Every compaction writes an outcome reason to <code>~/.local/share/opencode/log/fast-jev-compaction.log</code>.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#troubleshooting">See diagnostic reasons and smoke-test configuration</a></details><details><summary>Can I use the smaller 0.8B v3 model?</summary><p>Not through this plugin’s current local adapter. The v3 model needs its dedicated scoring runtime / System One-compatible API, which is not directly wired into this plugin yet.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#local-models">Read the model compatibility notes</a></details></section>
<section class="closing wrap"><h2>Keep what matters.<br>Inspect how it works.</h2><a class="button primary" href="https://github.com/overbit/fast-jev-compaction-opencode">Explore the source<svg viewBox="0 0 20 20" width="18" height="18" aria-hidden="true"><path d="M5 15 15 5M5 5h10v10" fill="none" stroke="currentColor" stroke-width="1.5"/></svg></a></section>
</main>
</file context>
Suggested change
| <section class="wrap troubleshooting" aria-labelledby="troubleshooting-title"><h2 id="troubleshooting-title">OpenCode /compact troubleshooting.</h2><details><summary>No request reaches the local model</summary><p>A classification request needs at least one completed tool call outside the pinned recent-message window. Short sessions may correctly send no request. Check the diagnostic log for candidate counts before changing your configuration.</p></details><details><summary>The local classifier falls back</summary><p>Check that the endpoint is reachable, the loaded context is large enough, and the model exposes <code>logprobs</code> and <code>top_logprobs</code>. Every compaction writes an outcome reason to <code>~/.local/share/opencode/log/fast-jev-compaction.log</code>.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#troubleshooting">See diagnostic reasons and smoke-test configuration</a></details><details><summary>Can I use the smaller 0.8B v3 model?</summary><p>Not through this plugin’s current local adapter. The v3 model needs its dedicated scoring runtime / System One-compatible API, which is not directly wired into this plugin yet.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#local-models">Read the model compatibility notes</a></details></section> | |
| <section class="wrap troubleshooting" aria-labelledby="troubleshooting-title"><h2 id="troubleshooting-title">OpenCode /compact troubleshooting.</h2><details><summary>No request reaches the local model</summary><p>A classification request needs at least one completed tool call outside the pinned recent-message window. Short sessions may correctly send no request. Check the diagnostic log for candidate counts before changing your configuration.</p></details><details><summary>The local classifier falls back</summary><p>Check that the System One endpoint is reachable, the requested model is a decision model, and its loaded context is large enough. The legacy chat backend additionally requires <code>logprobs</code> and <code>top_logprobs</code>. Every compaction writes an outcome reason to <code>~/.local/share/opencode/log/fast-jev-compaction.log</code>.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#troubleshooting">See diagnostic reasons and smoke-test confidence</a></details><details><summary>Can I use the smaller 0.8B v3 model?</summary><p>Yes. Use <code>backend: "systemone"</code> with LM Studio 0.4.26+ or another compatible System One server and the identifier of a loaded decision model. The legacy <code>backend: "local"</code> chat path is still intended for v1 models.</p><a href="https://github.com/overbit/fast-jev-compaction-opencode#local-models">Read the model compatibility notes</a></details></section> |
| User-agent: * | ||
| Allow: / | ||
|
|
||
| Sitemap: https://overbit.github.io/fast-jev-compaction-opencode/sitemap.xml |
There was a problem hiding this comment.
P2: This robots.txt sits under the project path, so crawlers will never fetch it. Only a robots.txt at the host root (overbit.github.io/robots.txt) is honored, so the Sitemap: line has no effect. Submit sitemap.xml directly in Search Console or remove the claimed discovery benefit from the PR description.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At site/robots.txt, line 4:
<comment>This `robots.txt` sits under the project path, so crawlers will never fetch it. Only a `robots.txt` at the host root (`overbit.github.io/robots.txt`) is honored, so the `Sitemap:` line has no effect. Submit `sitemap.xml` directly in Search Console or remove the claimed discovery benefit from the PR description.</comment>
<file context>
@@ -0,0 +1,4 @@
+User-agent: *
+Allow: /
+
+Sitemap: https://overbit.github.io/fast-jev-compaction-opencode/sitemap.xml
</file context>
This branch has not been deployed
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
/compact,/summarize, context pruning, and context-window search intentSoftwareSourceCodestructured metadata to the GitHub Pages siterobots.txtandsitemap.xmlfor crawler discoveryVerification
/compactcommandmain