@iceinvein/code-intelligence-mcp 1.3.8 → 1.3.9
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +2 -2
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -76,9 +76,9 @@ Add to your `opencode.json` (or global config):
|
|
|
76
76
|
|
|
77
77
|
## Standalone Server Mode
|
|
78
78
|
|
|
79
|
-
By default, each MCP client spawns its own server process (stdio transport). If you run multiple clients
|
|
79
|
+
By default, each MCP client spawns its own server process (stdio transport). If you run multiple clients against the same repo, a per-repo leader lock (`flock()`) ensures only one instance performs indexing, file watching, and LLM description generation. The leader loads the LLM (~1.1GB) during indexing and automatically frees it once descriptions are complete. Follower instances never load the LLM — they open the search index read-only and pick up the leader's changes. All instances load their own copy of the embedding model (~531MB) for query-time vector search.
|
|
80
80
|
|
|
81
|
-
**Standalone mode** runs a single long-lived HTTP server that all clients share.
|
|
81
|
+
**Standalone mode** runs a single long-lived HTTP server that all clients share. The main advantage is cross-repo deduplication — in stdio mode, each instance loads its own embedding model regardless of which repo it's on. With 5 instances across 3 repos, that's 5 copies (~2.6GB). Standalone loads the models once and shares them across all repos and clients.
|
|
82
82
|
|
|
83
83
|
### Starting the Server
|
|
84
84
|
|