fix: harden graph and MCP behavior and prepare 1.7.8 - #233
Conversation
Publish renderer-safe evidence mass and relation weight without changing scene identity, keep historical ghosts at zero mass, use visual weight for zero-strength history edges, and expose GRAPH_CAPACITY consistently across graph APIs. Remove duplicate graph-layer normalization and bind the public benchmark registry to the updated source tree with immutable v74 evidence.
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: c662167d4c
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 28ec6febf3
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 28ec6febf3
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: b3f76ba38c
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Graph rendering could lose labels on slow frames, reuse a cleared canvas, discard authored orbit lanes, and change a reloaded Compact view back to Galaxy. The fixes preserve custom physics settings, reject synthetic-cluster focus before state changes, and keep upper slider ranges responsive in Every-node and oversized views. Standalone overlays now refresh getter snapshots, wake when a paused engine is replaced, clear immediately when disabled, and ignore setters after destruction. The mobile Explore label and existing evidence-projection/capacity-error repairs are retained.
Standalone MCP launchers now preload optional semantic dependencies on the Windows launcher thread, preventing first-tool stalls during worker-thread imports. Embedding-only and reranker-only configurations are covered. Immutable-source validation precedes optional imports; diagnostics stay off the stdio protocol, and exact-backend/fallback policy remains enforced. ENGRAPHIS_MCP_PRELOAD_EMBEDDER controls the platform default. The HTTP launcher imports its required startup checks directly.
The experimental Jev adapter requires an injected configured client, an explicit model and per-call remote authorization. Missing, fallback, malformed or uncertain results defer. It remains outside core writes and grounded recall; the local client protocol still requires independent live SDK qualification.
This change also prepares version 1.7.8 for the reviewed code, including package and plugin metadata, the commercial manifest, campaign-adapter identity, plugin checksums, and release notes. Public fixtures were rerun into immutable v80 against the combined 580-file source manifest, with both charts and visible checksums regenerated. Historical evidence remains intact; the compatibility export command refuses existing output paths.
Validation:
These deterministic regression results do not establish full-product release qualification or live-provider readiness. Publication still requires the exact-artifact signed qualification enforced by the release workflow.
The latest Every-node theme-repaint review was checked using the actual renderer functions and frame queue: dark-to-light and light-to-dark transitions each repaint once within two frames with flow disabled, the graph frozen, and reduced motion enabled.
setThemeColorsschedulesdraw, whose unconditionalscheduleLabels()already handles these cases; all six stopped-motion transitions pass on b3f76ba. The review is resolved as non-actionable.