simple-dynamsoft-mcp 4.0.0 → 5.0.0

package/.env.example

@@ -0,0 +1,57 @@
+# RAG provider selection (primary + fallback)
+# RAG_PROVIDER: auto | gemini | local | fuse
+#   - auto: gemini if GEMINI_API_KEY is set, otherwise local
+#   - gemini: remote embeddings via Google Gemini
+#   - local: local embeddings via @xenova/transformers
+#   - fuse: legacy fuzzy search only (no embeddings)
+RAG_PROVIDER=auto
+
+# RAG_FALLBACK: none | fuse | local
+#   - none: do not fall back if primary fails
+#   - fuse: fuzzy search fallback (fast, no embeddings)
+#   - local: local embeddings fallback (if primary is gemini)
+RAG_FALLBACK=fuse
+
+# Gemini remote embeddings (required when RAG_PROVIDER=gemini)
+# * GEMINI_API_KEY: your Gemini API key (never commit to git)
+# * GEMINI_EMBED_MODEL: models/embedding-001 | models/gemini-embedding-001 | gemini-embedding-001
+# * GEMINI_API_BASE_URL: override for proxies (default https://generativelanguage.googleapis.com)
+# * GEMINI_EMBED_BATCH_SIZE: batch size for Gemini embedding requests
+# GEMINI_API_KEY=your_key
+# GEMINI_EMBED_MODEL=gemini-embedding-001
+# GEMINI_API_BASE_URL=https://generativelanguage.googleapis.com
+# GEMINI_EMBED_BATCH_SIZE=16
+
+# Local embeddings (used when RAG_PROVIDER=local or fallback=local)
+# * RAG_LOCAL_MODEL: Hugging Face model id (default Xenova/all-MiniLM-L6-v2)
+# * RAG_LOCAL_QUANTIZED: true|false (smaller/faster model download when true)
+# RAG_LOCAL_MODEL=Xenova/all-MiniLM-L6-v2
+# RAG_LOCAL_QUANTIZED=true
+
+# Cache locations
+# * RAG_CACHE_DIR: vector index cache (default data/.rag-cache)
+# * RAG_MODEL_CACHE_DIR: local model cache (default data/.rag-cache/models)
+# RAG_CACHE_DIR=data/.rag-cache
+# RAG_MODEL_CACHE_DIR=data/.rag-cache/models
+
+# Indexing + retrieval tuning
+# * RAG_CHUNK_SIZE: max chars per chunk when embedding doc content
+# * RAG_CHUNK_OVERLAP: overlap between chunks (helps context continuity)
+# * RAG_MAX_CHUNKS_PER_DOC: cap chunks per doc to control index size
+# * RAG_MAX_TEXT_CHARS: max chars per embedding input
+# * RAG_MIN_SCORE: minimum cosine similarity to keep a hit (0 disables filtering). Default 0.2.
+# * RAG_INCLUDE_SCORE: include similarity score in search results (debugging)
+# RAG_CHUNK_SIZE=1200
+# RAG_CHUNK_OVERLAP=200
+# RAG_MAX_CHUNKS_PER_DOC=6
+# RAG_MAX_TEXT_CHARS=4000
+# RAG_MIN_SCORE=0.2
+# RAG_INCLUDE_SCORE=false
+
+# Cache/boot controls
+# * RAG_REBUILD: true to ignore cache and rebuild on startup/search
+# * RAG_PREWARM: true to build the embedding index at startup
+# * RAG_PREWARM_BLOCK: true to block startup until prewarm completes
+# RAG_REBUILD=false
+# RAG_PREWARM=false
+# RAG_PREWARM_BLOCK=false

package/README.md

@@ -19,7 +19,7 @@ https://github.com/user-attachments/assets/cc1c5f4b-1461-4462-897a-75abc20d62a6
 - **Multiple SDKs**: Barcode Reader (Mobile/Python/Web) + Dynamic Web TWAIN + Document Viewer
 - **Multiple API Levels**: High-level (simple) and low-level (advanced) options
 - **Stdio MCP server**: Runs on stdio. Works with any MCP-capable client.
-- **Resource-efficient discovery**: Resources are discovered via tools (fuzzy search + resource links). Only a small pinned set is listed by default; heavy content is fetched on-demand with `resources/read`.
+- **Resource-efficient discovery**: Resources are discovered via tools (semantic RAG search with fuzzy fallback + resource links). Only a small pinned set is listed by default; heavy content is fetched on-demand with `resources/read`.
 - **Latest-major policy**: The server only serves the latest major versions; older major requests are refused with legacy links when available.
 
 ## Available Tools
@@ -28,6 +28,8 @@ https://github.com/user-attachments/assets/cc1c5f4b-1461-4462-897a-75abc20d62a6
 |------|-------------|
 | `get_index` | Compact index of products, editions, versions, samples, and docs |
 | `search` | Unified search across docs and samples; returns resource links |
+| `list_samples` | List available sample IDs and URIs for a scope |
+| `resolve_sample` | Resolve a sample_id (or sample URI) to matching sample URIs |
 | `resolve_version` | Resolve a concrete latest-major version for a product/edition |
 | `get_quickstart` | Opinionated quickstart for a target stack |
 | `generate_project` | Assemble a project structure from a sample (no AI generation) |
@@ -284,10 +286,31 @@ data/
 ## Using Search-Based Discovery (Recommended)
 
 - On session start, let your client call `tools/list` and `resources/list` (pinned only, not exhaustive).
-- For any query, call `search` with keywords; it returns `resource_link` entries.
+- For any query, call `search`; it uses semantic RAG retrieval (with fuzzy fallback) and returns `resource_link` entries.
 - Read only the links you need via `resources/read` to avoid bloating the context window.
 - If unsure what to search, call `get_index` first to see what is available.
 
+## RAG Configuration
+
+Search providers are selected at runtime via environment variables (safe for public npm packages). Defaults to `auto` -> `gemini` if `GEMINI_API_KEY` is set, otherwise `local`, with `fuse` fallback on failure.
+
+To keep the legacy fuzzy search (no model download), set `RAG_PROVIDER=fuse`.
+
+Key env vars:
+- `RAG_PROVIDER`: `auto` | `gemini` | `local` | `fuse`
+- `RAG_FALLBACK`: `fuse` | `local` | `none`
+- `GEMINI_API_KEY`: required for remote embeddings
+- `GEMINI_EMBED_MODEL`: e.g. `models/embedding-001` or `models/gemini-embedding-001`
+- `RAG_LOCAL_MODEL`: default `Xenova/all-MiniLM-L6-v2`
+- `RAG_CACHE_DIR`: default `data/.rag-cache`
+
+Local embeddings download the model on first run and cache under `data/.rag-cache/models`.
+Advanced tuning:
+- `RAG_CHUNK_SIZE`, `RAG_CHUNK_OVERLAP`, `RAG_MAX_CHUNKS_PER_DOC`, `RAG_MAX_TEXT_CHARS`
+- `RAG_MIN_SCORE`, `RAG_INCLUDE_SCORE`, `RAG_REBUILD`, `RAG_PREWARM`, `RAG_PREWARM_BLOCK`, `RAG_LOCAL_QUANTIZED`, `GEMINI_EMBED_BATCH_SIZE`, `RAG_MODEL_CACHE_DIR`
+
+For local dev, you can also use a `.env` file (see `.env.example`).
+
 ## Version Policy
 
 - This MCP server serves only the latest major versions (DBR v11, DWT v19).

package/code-snippet/dynamsoft-barcode-reader/web/frameworks/blazor/scan-using-foundational-api/hello-world-blazor/BlazorApp/obj/Debug/net8.0/BlazorApp.AssemblyInfo.cs

@@ -13,7 +13,7 @@ using System.Reflection;
 [assembly: System.Reflection.AssemblyCompanyAttribute("BlazorApp")]
 [assembly: System.Reflection.AssemblyConfigurationAttribute("Debug")]
 [assembly: System.Reflection.AssemblyFileVersionAttribute("1.0.0.0")]
-[assembly: System.Reflection.AssemblyInformationalVersionAttribute("1.0.0+aeb6c2f34fb9b7aca3eb10b805b23565a5a85c3c")]
+[assembly: System.Reflection.AssemblyInformationalVersionAttribute("1.0.0+c15266272a4c5826609d9e0aa7b95eefb2073fbf")]
 [assembly: System.Reflection.AssemblyProductAttribute("BlazorApp")]
 [assembly: System.Reflection.AssemblyTitleAttribute("BlazorApp")]
 [assembly: System.Reflection.AssemblyVersionAttribute("1.0.0.0")]

package/code-snippet/dynamsoft-barcode-reader/web/frameworks/blazor/scan-using-rtu-api/obj/Debug/net8.0/BlazorApp.AssemblyInfo.cs

@@ -13,7 +13,7 @@ using System.Reflection;
 [assembly: System.Reflection.AssemblyCompanyAttribute("BlazorApp")]
 [assembly: System.Reflection.AssemblyConfigurationAttribute("Debug")]
 [assembly: System.Reflection.AssemblyFileVersionAttribute("1.0.0.0")]
-[assembly: System.Reflection.AssemblyInformationalVersionAttribute("1.0.0+aeb6c2f34fb9b7aca3eb10b805b23565a5a85c3c")]
+[assembly: System.Reflection.AssemblyInformationalVersionAttribute("1.0.0+c15266272a4c5826609d9e0aa7b95eefb2073fbf")]
 [assembly: System.Reflection.AssemblyProductAttribute("BlazorApp")]
 [assembly: System.Reflection.AssemblyTitleAttribute("BlazorApp")]
 [assembly: System.Reflection.AssemblyVersionAttribute("1.0.0.0")]

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "simple-dynamsoft-mcp",
-  "version": "4.0.0",
+  "version": "5.0.0",
   "description": "MCP server for Dynamsoft SDKs - Barcode Reader (Mobile/Python/Web), Dynamic Web TWAIN, and Document Viewer. Provides documentation, code snippets, and API guidance.",
   "license": "MIT",
   "repository": {
@@ -18,7 +18,8 @@
     "src",
     "data",
     "code-snippet",
-    "README.md"
+    "README.md",
+    ".env.example"
   ],
   "scripts": {
     "start": "node src/index.js",
@@ -41,6 +42,8 @@
   ],
   "dependencies": {
     "@modelcontextprotocol/sdk": "^1.25.2",
+    "@xenova/transformers": "^2.17.2",
+    "dotenv": "^16.4.5",
     "fuse.js": "^7.0.0",
     "zod": "~3.24.0"
   },