fix: harden graph and MCP behavior and prepare 1.7.8

.claude-plugin/marketplace.json

@@ -9,7 +9,7 @@
       "name": "engraphis-memory",
       "source": "./",
       "description": "Discipline for giving agents durable, scoped, explainable memory across sessions and repos with the Engraphis MCP tools.",
-      "version": "1.7.7"
+      "version": "1.7.8"
     }
   ]
 }

.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "engraphis-memory",
-  "version": "1.7.7",
+  "version": "1.7.8",
   "description": "Give agents durable, scoped, explainable memory across sessions and repos via the Engraphis MCP tools. Use when you learn something worth keeping, need prior context before acting, or ask why/how a fact changed. Covers remember/recall, why/timeline, forget/pin/correct, sessions, and code search.",
   "author": {
     "name": "The Engraphis Authors",

.claude-plugin/skill-assets.sha256

@@ -1,5 +1,5 @@
-d0620f9a4f1c071ed9c1105cae296f9f1d32a673fa7bca6cc1a85ce22b393c57  .claude-plugin/marketplace.json
-c8057476dc560c4dff1ecafbd357b470cbcd7e2c712eb3558f84b4571403753b  .claude-plugin/plugin.json
+df65a383a1ff80572fb6ebd9a807dca6d1ffc0f99f373971262de035b8e3622d  .claude-plugin/marketplace.json
+a03b5c38d836651d2c5c52173d21a5adeacfbbc4b80aac991c2154b6e060e174  .claude-plugin/plugin.json
 4bc8979b9ffeb97190960e551dbf4ddc6f7aeeb7b86894fd2298a59ff0001efa  skills/engraphis-memory/SKILL.md
 055655db84af07561d002f0c69744313d8413c39f3e873f941f0fa0b1e76dc66  skills/engraphis-memory/references/CONVENTIONS.md
 62019760766ff472a76a0f81437898f39e3c1fe2631732b7b7733e50c1ad837f  skills/engraphis-memory/references/SCOPING.md

BENCHMARKS.md

@@ -94,14 +94,14 @@ interpretation and do not count as additional benchmark-quality gains.
 ### Public numeric evidence registry
 
 Every exact public aggregate retained below comes from the checked-in, public-safe
-[`offline-fixtures-v76.json`](docs/benchmark-evidence/offline-fixtures-v76.json) artifact. Its
+[`offline-fixtures-v80.json`](docs/benchmark-evidence/offline-fixtures-v80.json) artifact. Its
 SHA-256 is
-`2fb5ce5b2cbc21541f7ae9cad5d2ef615014c0e86f9881f00621d00986422ba8`, also recorded in the
+`ac63dac1e34c66b658eeb5846599ef42d774a5e212860b2a909941f364c82bc2`, also recorded in the
 adjacent `.sha256` file. The artifact contains no raw questions, answers, prompts, customer data,
 or per-record content fingerprints.
 
 The fixture-suite digest is
-`20131c25c86e3ac5a53285d0631d4ff0c60946879b04ff7a7c38e488f42cda51`. The artifact defines
+`431b525443f5b4875646e7f6470d107f584120bdd6ee7f53d0b6484d003fa0f7`. The artifact defines
 the digest algorithm and records the SHA-256 of every suite and dataset file. Each evidence ID
 also binds its exact command through `sha256(UTF-8 exact command)`:
 
@@ -123,10 +123,10 @@ Historical LoCoMo, graph, handoff, consolidation, and security figures remain pr
 source artifacts but are omitted from the current chart until each has a matching immutable,
 public-safe artifact. The chart labels coding outcomes, external datasets, and operational
 capacity as pending evaluation tracks rather than implying scores. Regenerate it with
-`python scripts/render_benchmark_report.py --report docs/benchmark-evidence/offline-fixtures-v76.json --output docs/images/context-efficiency.svg` after selecting the report to publish.
+`python scripts/render_benchmark_report.py --report docs/benchmark-evidence/offline-fixtures-v80.json --output docs/images/context-efficiency.svg` after selecting the report to publish.
 
 The companion examples are also generated from that artifact with
-`python -m scripts.render_benchmark_examples --report docs/benchmark-evidence/offline-fixtures-v76.json --output docs/images/evidence-backed-agent-examples.svg`.
+`python -m scripts.render_benchmark_examples --report docs/benchmark-evidence/offline-fixtures-v80.json --output docs/images/evidence-backed-agent-examples.svg`.
 The historical-to-executable mapping is in
 [`docs/BENCHMARK_CHANGE_COVERAGE.md`](docs/BENCHMARK_CHANGE_COVERAGE.md).
 

CHANGELOG.md

@@ -5,6 +5,20 @@ All notable changes to Engraphis are documented here. Format loosely follows
 
 ## [Unreleased]
 
+## [1.7.8] - 2026-09-27
+
+- Improved graph rendering and overlay scheduling, preserved saved Compact and custom
+  slider preferences, and corrected orbit radii, focus validation, and worker force limits.
+- Hardened Windows MCP startup by preloading configured embedding and reranking
+  dependencies before background warmup, while retaining exact-backend requirements,
+  source-integrity validation, and an explicit preload opt-out.
+- Added an experimental, explicitly authorized Jev decision adapter with fail-closed
+  response validation; it does not write memories or participate in grounded recall.
+- Corrected API capacity and evidence projections and refreshed immutable offline
+  evidence and charts against the release source. Offline fixtures do not establish
+  live hosted-service or full-product qualification.
+- Updated the optional Codex SDK to 0.155.1 and pinned CodeQL actions to 4.38.2.
+
 ## [1.7.7] - 2026-09-23
 
 - Cloud Sync shows local encryption-key and dependency readiness before enabling Sync now,

README.md

@@ -84,9 +84,9 @@ neither is an end-to-end question-answer score. Coding outcomes, external datase
 operational capacity remain separate pending evaluation tracks until their artifacts are selected.
 
 These values are evidence IDs `offline-chunking` and `offline-performance` in
-[`offline-fixtures-v76.json`](https://github.com/Coding-Dev-Tools/engraphis/blob/main/docs/benchmark-evidence/offline-fixtures-v76.json),
+[`offline-fixtures-v80.json`](https://github.com/Coding-Dev-Tools/engraphis/blob/main/docs/benchmark-evidence/offline-fixtures-v80.json),
 SHA-256
-`2fb5ce5b2cbc21541f7ae9cad5d2ef615014c0e86f9881f00621d00986422ba8`.
+`ac63dac1e34c66b658eeb5846599ef42d774a5e212860b2a909941f364c82bc2`.
 [`BENCHMARKS.md`](https://github.com/Coding-Dev-Tools/engraphis/blob/main/BENCHMARKS.md#public-numeric-evidence-registry)
 records the matching suite digest, exact commands, and per-command config digests. The offline
 fixture registry intentionally excludes external, model-dependent, consolidation, productivity,
@@ -798,6 +798,7 @@ file. It never searches the working directory for `.env`, and explicit process v
 | `ENGRAPHIS_HTTP_INDEX_ROOT` | First `ENGRAPHIS_INDEX_ROOTS` entry, or current directory | Single root for dashboard and REST `POST /api/code/index`; submitted paths resolve beneath it. An explicit root (or fallback entry) must be absolute; an explicit HTTP root is included in the engine-approved set. MCP and CLI indexing continue to use `ENGRAPHIS_INDEX_ROOTS`. |
 | `ENGRAPHIS_DB_KEY` | Not set | Encrypt the database at rest (SQLCipher). Or use `ENGRAPHIS_DB_KEY_FILE` |
 | `ENGRAPHIS_EMBED_MODEL` | `sentence-transformers/all-MiniLM-L6-v2` | sentence-transformers model |
+| `ENGRAPHIS_MCP_PRELOAD_EMBEDDER` | `auto` | Standalone MCP launchers import optional semantic dependencies on the launcher thread on Windows before serving requests. Set `0` to disable or `1` to enable on any platform; model loading and backend fallback policy remain unchanged. |
 | `ENGRAPHIS_EMBED_REVISION` | Not set | Optional immutable lowercase 40-hex Hugging Face commit for the embedding model. Loaded Hub commits or local artifact manifests identify persistent vector spaces; unresolved mutable identities keep vector recall fail-closed. |
 | `ENGRAPHIS_RERANK_MODEL` | Not set | Optional sentence-transformers cross-encoder reranker |
 | `ENGRAPHIS_RERANK_REVISION` | Not set | Optional immutable lowercase 40-hex Hugging Face commit for the reranker |

docs/MCP_TOOLS.md

@@ -49,6 +49,20 @@ Hosts which already own chat history should use `POST /api/adaptive-context`, no
 The gateway works in general MCP clients without native deferred tool search; clients that
 explicitly support OpenAI's deferred `tool_search` can apply it as an optional host optimization.
 
+### Standalone semantic startup
+
+On Windows, the standalone stdio and HTTP launchers import optional semantic dependencies
+on the launcher thread before serving MCP requests. This prevents the first semantic tool
+call from stalling during a native dependency import in a worker thread. Models are still
+loaded lazily, and exact-backend validation retains its normal failure policy. Dependency
+import diagnostics go to stderr so stdout stays reserved for the stdio protocol.
+
+`ENGRAPHIS_MCP_PRELOAD_EMBEDDER=auto` enables this startup step on Windows when a semantic
+embedding or reranking model is configured. Configured sources are validated before
+optional imports, including the immutable revision policy. Set `0` to disable it or `1` to
+enable it on other platforms. An unavailable optional dependency is left to the configured
+backend fallback policy.
+
 ## Classic direct-tool inventory
 
 The following inventory applies to the Classic compatibility server. Start with