@lotargo/memory_plugin 1.6.4 → 1.6.6

package/mcp-server/tools/rag_tools.js

@@ -15,247 +15,252 @@ export function registerRagTools(server) {
     {
       description:
         "Selectively preserve a reliable, reusable source in the RAG knowledge base; do not ingest everything encountered. " +
-        "Accepts local file paths, web URLs, or raw Markdown/text content. " +
-        "For type='file' the file is read from disk and indexed with a code-block wrapper. " +
-        "For type='url' the page is fetched and its content is indexed (not just the URL). " +
-        "Processes document through 3-tier hierarchy chunking (Big/Medium/Small), " +
-        "computes dense vectors, and extracts GraphRAG code symbols.",
-      inputSchema: z.object({
-        content: z
-          .string()
-          .describe(
-            "Raw text content, file path, or web URL. For type='file' this can be the file path (reads from disk) or the file content directly"
-          ),
-        type: z
-          .enum(["text", "file", "url"])
-          .nullish()
-          .transform((v) => v || "text")
-          .describe("Input content type: 'text' (raw content), 'file' (reads from disk, wraps in code block), or 'url' (fetches page content)"),
-        title: optStr().describe("Document title"),
+        "Accepts local file paths, web URLs, or raw Markdown/text content. " +
+        "For type='file' the file is read from disk and indexed with a code-block wrapper. " +
+        "For type='url' the page is fetched and its content is indexed (not just the URL). " +
+        "Processes document through 3-tier hierarchy chunking (Big/Medium/Small), " +
+        "computes dense vectors, and extracts GraphRAG code symbols.",
+      inputSchema: z.object({
+        content: z
+          .string()
+          .describe(
+            "Raw text content, file path, or web URL. For type='file' this can be the file path (reads from disk) or the file content directly"
+          ),
+        type: z
+          .enum(["text", "file", "url"])
+          .nullish()
+          .transform((v) => v || "text")
+          .describe("Input content type: 'text' (raw content), 'file' (reads from disk, wraps in code block), or 'url' (fetches page content)"),
+        title: optStr().describe("Document title"),
         path: optStr().describe("Original document file path"),
         scope: z.enum(["project", "global"]).nullish().transform((v) => v || "project").describe("RAG visibility: current Git project (default) or global"),
+        directory: optStr().describe("Optional workspace/project directory path to target"),
+        project: optStr().describe("Alias for directory"),
         generateEmbeddings: defBool(true).describe("Compute dense vector embeddings"),
       }),
     },
-    async ({ content, type, title, path, scope, generateEmbeddings }) => {
+    async ({ content, type, title, path, scope, directory, project, generateEmbeddings }) => {
       const { ingestDocument } = await import("../ingest/pipeline.js");
-      const projectScope = await resolveRagScopeKey(scope);
+      const projectScope = await resolveRagScopeKey(scope, { directory, project });
       const result = await ingestDocument({
-        content,
-        type,
-        title: title || null,
-        path: path || null,
+        content,
+        type,
+        title: title || null,
+        path: path || null,
         generateEmbeddings,
         projectScope,
-      });
-      return {
-        content: [
-          {
-            type: "text",
-            text: JSON.stringify(
-              {
-                status: "success",
-                docId: result.docId,
-                title: result.title,
-                sectionsCount: result.sectionsCount,
-                microChunksCount: result.microChunksCount,
+      });
+      return {
+        content: [
+          {
+            type: "text",
+            text: JSON.stringify(
+              {
+                status: "success",
+                docId: result.docId,
+                title: result.title,
+                sectionsCount: result.sectionsCount,
+                microChunksCount: result.microChunksCount,
                 deduplicated: result.deduplicated,
                 scope: result.projectScope,
-              },
-              null,
-              2
-            ),
-          },
-        ],
-      };
-    }
-  );
-
-  server.registerTool(
-    "query_knowledge_base",
-    {
+              },
+              null,
+              2
+            ),
+          },
+        ],
+      };
+    }
+  );
+
+  server.registerTool(
+    "query_knowledge_base",
+    {
       description:
         "Perform project-isolated hybrid search (RSF/RRF BM25 full-text + dense vector similarity) across the RAG knowledge base. " +
-        "Returns top-ranked candidate document sections with breadcrumbs, GraphRAG defined code symbols, and relevance scores.",
-      inputSchema: z.object({
-        query: z.string().describe("Search query in natural language or symbol name"),
-        limit: defNum(5).describe("Maximum number of sections to return"),
-        instruction: optStr().describe(
-          "Optional task-specific retrieval instruction shaping embedding focus (e.g. 'Retrieve code snippets', 'Find user preferences'). " +
-            "Recommended when using E5/BGE models for domain-specific queries."
-        ),
+        "Returns top-ranked candidate document sections with breadcrumbs, GraphRAG defined code symbols, and relevance scores.",
+      inputSchema: z.object({
+        query: z.string().describe("Search query in natural language or symbol name"),
+        limit: defNum(5).describe("Maximum number of sections to return"),
+        instruction: optStr().describe(
+          "Optional task-specific retrieval instruction shaping embedding focus (e.g. 'Retrieve code snippets', 'Find user preferences'). " +
+            "Recommended when using E5/BGE models for domain-specific queries."
+        ),
         generateEmbeddings: defBool(true).describe("Use vector search alongside BM25"),
         scope: z.enum(["all", "project", "global"]).nullish().transform((v) => v || "all").describe("Search global + current project (default), project only, or global only"),
+        directory: optStr().describe("Optional workspace/project directory path to target"),
+        project: optStr().describe("Alias for directory"),
       }),
     },
-    async ({ query, limit, instruction, generateEmbeddings, scope }) => {
-      const { hybridQuery } = await import("../retrieval/retriever.js");
-      const { getConfig } = await import("../config/config_manager.js");
+    async ({ query, limit, instruction, generateEmbeddings, scope, directory, project }) => {
+      const { hybridQuery } = await import("../retrieval/retriever.js");
+      const { getConfig } = await import("../config/config_manager.js");
       const activeConfig = getConfig();
-      const scopeKeys = await resolveRagScopeKeys(scope);
-
-      const results = await hybridQuery({
-        query,
-        limit,
-        generateEmbeddings,
+      const scopeKeys = await resolveRagScopeKeys(scope, { directory, project });
+
+      const results = await hybridQuery({
+        query,
+        limit,
+        generateEmbeddings,
         instruction: instruction || null,
         scopeKeys,
-      });
-
-      if (!results || results.length === 0) {
-        return {
-          content: [
-            {
-              type: "text",
-              text: `[Active Model: ${activeConfig.embeddingModel}]\nNo matching knowledge found for query.`,
-            },
-          ],
-        };
-      }
-
-      const headerNote = `[Active Model: ${activeConfig.embeddingModel} | Fusion: ${activeConfig.fusionAlgorithm.toUpperCase()}]\n\n`;
-
-      const formatted = results
-        .map((r, i) => {
-          let header = `### [${i + 1}] ${r.doc_title || "Untitled"}`;
-          if (r.heading) header += ` > ${r.heading}`;
-          if (r.breadcrumbs) header += ` (${r.breadcrumbs})`;
-          let body = `Score: ${(r.score || 0).toFixed(4)}\n`;
-          if (r.defined_symbols && r.defined_symbols.length > 0) {
-            body += `Defined Symbols: ${r.defined_symbols.join(", ")}\n`;
-          }
-          body += `\n${r.snippet || r.full_section_content || ""}`;
-          return `${header}\n${body}`;
-        })
-        .join("\n\n---\n\n");
-
-      return { content: [{ type: "text", text: headerNote + formatted }] };
-    }
-  );
-
-  server.registerTool(
-    "batch_query_knowledge_base",
-    {
+      });
+
+      if (!results || results.length === 0) {
+        return {
+          content: [
+            {
+              type: "text",
+              text: `[Active Model: ${activeConfig.embeddingModel}]\nNo matching knowledge found for query.`,
+            },
+          ],
+        };
+      }
+
+      const headerNote = `[Active Model: ${activeConfig.embeddingModel} | Fusion: ${activeConfig.fusionAlgorithm.toUpperCase()}]\n\n`;
+
+      const formatted = results
+        .map((r, i) => {
+          let header = `### [${i + 1}] ${r.doc_title || "Untitled"}`;
+          if (r.heading) header += ` > ${r.heading}`;
+          if (r.breadcrumbs) header += ` (${r.breadcrumbs})`;
+          let body = `Score: ${(r.score || 0).toFixed(4)}\n`;
+          if (r.defined_symbols && r.defined_symbols.length > 0) {
+            body += `Defined Symbols: ${r.defined_symbols.join(", ")}\n`;
+          }
+          body += `\n${r.snippet || r.full_section_content || ""}`;
+          return `${header}\n${body}`;
+        })
+        .join("\n\n---\n\n");
+
+      return { content: [{ type: "text", text: headerNote + formatted }] };
+    }
+  );
+
+  server.registerTool(
+    "batch_query_knowledge_base",
+    {
       description:
         "Execute multiple hybrid search queries in a single batch call. " +
         "Search is isolated to global plus the current project unless another scope is requested. " +
-        "More efficient than separate query_knowledge_base calls: all query embeddings computed in one ONNX pass. " +
-        "Returns one result set per query, in the same order as input.",
-      inputSchema: z.object({
-        queries: z
-          .array(z.string())
-          .describe("Array of search queries to execute in batch"),
-        limit: defNum(5).describe("Maximum number of sections to return per query"),
-        instruction: optStr().describe(
-          "Optional task-specific retrieval instruction shaping embedding focus. Applied to all queries."
-        ),
+        "More efficient than separate query_knowledge_base calls: all query embeddings computed in one ONNX pass. " +
+        "Returns one result set per query, in the same order as input.",
+      inputSchema: z.object({
+        queries: z
+          .array(z.string())
+          .describe("Array of search queries to execute in batch"),
+        limit: defNum(5).describe("Maximum number of sections to return per query"),
+        instruction: optStr().describe(
+          "Optional task-specific retrieval instruction shaping embedding focus. Applied to all queries."
+        ),
         generateEmbeddings: defBool(true).describe("Use vector search alongside BM25"),
         scope: z.enum(["all", "project", "global"]).nullish().transform((v) => v || "all").describe("Search global + current project (default), project only, or global only"),
+        directory: optStr().describe("Optional workspace/project directory path to target"),
+        project: optStr().describe("Alias for directory"),
       }),
     },
-    async ({ queries, limit, instruction, generateEmbeddings, scope }) => {
-      const { batchHybridQuery } = await import("../retrieval/retriever.js");
-      const { getConfig } = await import("../config/config_manager.js");
+    async ({ queries, limit, instruction, generateEmbeddings, scope, directory, project }) => {
+      const { batchHybridQuery } = await import("../retrieval/retriever.js");
+      const { getConfig } = await import("../config/config_manager.js");
       const activeConfig = getConfig();
-      const scopeKeys = await resolveRagScopeKeys(scope);
-
-      const allResults = await batchHybridQuery(queries, {
-        limit,
-        generateEmbeddings,
+      const scopeKeys = await resolveRagScopeKeys(scope, { directory, project });
+
+      const allResults = await batchHybridQuery(queries, {
+        limit,
+        generateEmbeddings,
         instruction: instruction || null,
         scopeKeys,
-      });
-
-      const formatted = allResults
-        .map((results, qi) => {
-          const header = `## Query ${qi + 1}: "${queries[qi]}"\n`;
-          if (!results || results.length === 0) {
-            return header + "_No results found._";
-          }
-          const items = results
-            .map((r, i) => {
-              let h = `### [${i + 1}] ${r.doc_title || "Untitled"}`;
-              if (r.heading) h += ` > ${r.heading}`;
-              if (r.breadcrumbs) h += ` (${r.breadcrumbs})`;
-              let body = `Score: ${(r.score || 0).toFixed(4)}`;
-              if (r.retrieval_policy && r.retrieval_policy !== "micro_chunk") {
-                body += ` [${r.retrieval_policy}]`;
-              }
-              if (r.defined_symbols && r.defined_symbols.length > 0) {
-                body += `\nDefined Symbols: ${r.defined_symbols.join(", ")}`;
-              }
-              body += `\n\n${r.snippet || r.full_section_content || ""}`;
-              return `${h}\n${body}`;
-            })
-            .join("\n\n---\n\n");
-          return header + items;
-        })
-        .join("\n\n===\n\n");
-
-      const headerNote = `[Active Model: ${activeConfig.embeddingModel} | Fusion: ${activeConfig.fusionAlgorithm.toUpperCase()} | ${queries.length} queries]\n\n`;
-
-      return { content: [{ type: "text", text: headerNote + formatted }] };
-    }
-  );
-
-  server.registerTool(
-    "reindex_knowledge_base",
-    {
-      description:
-        "Re-embed all existing documents in the RAG knowledge base with the active (or specified) embedding model and vector dimension. " +
-        "Use after switching the embedding model or vector dimension so previously stored vectors match the new configuration. " +
-        "Preserves documents, sections, FTS index, graph edges, and fact links.",
-      inputSchema: z.object({
-        model: optStr().describe("Embedding model to use (defaults to active config.embeddingModel)"),
-        dimension: optNum().describe(
-          "Fixed vector dimension (defaults to active config.vectorDimension; auto-detect if unset)"
-        ),
-      }),
-    },
-    async ({ model, dimension }) => {
-      const { reindexEmbeddings } = await import("../ingest/pipeline.js");
-      const result = await reindexEmbeddings({
-        model: model || null,
-        dimension: dimension !== undefined && dimension !== null ? dimension : null,
-      });
-      return {
-        content: [
-          {
-            type: "text",
-            text: JSON.stringify(
-              {
-                status: "success",
-                reindexed: result.reindexed,
-                documentsAffected: result.documentsAffected,
-                model: result.model,
-                dimension: result.dimension || "auto",
-              },
-              null,
-              2
-            ),
-          },
-        ],
-      };
-    }
-  );
-
-  server.registerTool(
-    "manage_knowledge_base",
-    {
+      });
+
+      const formatted = allResults
+        .map((results, qi) => {
+          const header = `### Query [${qi + 1}]: "${queries[qi]}"\n\n`;
+          if (!results || results.length === 0) {
+            return header + "No matching knowledge found for this query.";
+          }
+          const items = results
+            .map((item, j) => {
+              let title = `#### [${j + 1}] ${item.doc_title || "Untitled"}`;
+              if (item.heading) title += ` > ${item.heading}`;
+              if (item.breadcrumbs) title += ` (${item.breadcrumbs})`;
+              let body = `Score: ${(item.score || 0).toFixed(4)}\n`;
+              if (item.defined_symbols && item.defined_symbols.length > 0) {
+                body += `Defined Symbols: ${item.defined_symbols.join(", ")}\n`;
+              }
+              body += `\n${item.snippet || item.full_section_content || ""}`;
+              return `${title}\n${body}`;
+            })
+            .join("\n\n---\n\n");
+          return header + items;
+        })
+        .join("\n\n===\n\n");
+
+      const headerNote = `[Active Model: ${activeConfig.embeddingModel} | Fusion: ${activeConfig.fusionAlgorithm.toUpperCase()} | ${queries.length} queries]\n\n`;
+
+      return { content: [{ type: "text", text: headerNote + formatted }] };
+    }
+  );
+
+  server.registerTool(
+    "reindex_knowledge_base",
+    {
+      description:
+        "Re-embed all existing documents in the RAG knowledge base with the active (or specified) embedding model and vector dimension. " +
+        "Use after switching the embedding model or vector dimension so previously stored vectors match the new configuration. " +
+        "Preserves documents, sections, FTS index, graph edges, and fact links.",
+      inputSchema: z.object({
+        model: optStr().describe("Embedding model to use (defaults to active config.embeddingModel)"),
+        dimension: optNum().describe(
+          "Fixed vector dimension (defaults to active config.vectorDimension; auto-detect if unset)"
+        ),
+      }),
+    },
+    async ({ model, dimension }) => {
+      const { reindexEmbeddings } = await import("../ingest/pipeline.js");
+      const result = await reindexEmbeddings({
+        model: model || null,
+        dimension: dimension !== undefined && dimension !== null ? dimension : null,
+      });
+      return {
+        content: [
+          {
+            type: "text",
+            text: JSON.stringify(
+              {
+                status: "success",
+                reindexed: result.reindexed,
+                documentsAffected: result.documentsAffected,
+                model: result.model,
+                dimension: result.dimension || "auto",
+              },
+              null,
+              2
+            ),
+          },
+        ],
+      };
+    }
+  );
+
+  server.registerTool(
+    "manage_knowledge_base",
+    {
       description:
         "Manage the project-isolated RAG knowledge base: inspect stats, list documents, read full raw document, unlink/delete documents, or export/import complete snapshots.",
-      inputSchema: z.object({
-        action: z.enum(["stats", "list", "read_document", "delete", "export_snapshot", "import_snapshot"]).describe("Management action"),
-        docId: optStr().describe("Document ID, title, or path (required for read_document and delete)"),
+      inputSchema: z.object({
+        action: z.enum(["stats", "list", "read_document", "delete", "export_snapshot", "import_snapshot"]).describe("Management action"),
+        docId: optStr().describe("Document ID, title, or path (required for read_document and delete)"),
         snapshotPath: optStr().describe("File path for snapshot export/import"),
         scope: z.enum(["all", "project", "global"]).nullish().describe("For stats/list/read: global + current project by default. Delete defaults to the current project (or global outside Git); pass all/global explicitly for broader removal"),
+        directory: optStr().describe("Optional workspace/project directory path to target"),
+        project: optStr().describe("Alias for directory"),
       }),
     },
-    async ({ action, docId, snapshotPath, scope }) => {
+    async ({ action, docId, snapshotPath, scope, directory, project }) => {
       const { getDatabase } = await import("../db/database.js");
       const db = await getDatabase();
       const scopeKeys = ["stats", "list", "read_document", "delete"].includes(action)
-        ? await resolveManageRagScopeKeys(action, scope)
+        ? await resolveManageRagScopeKeys(action, scope, { directory, project })
         : null;
       const placeholders = scopeKeys ? scopeKeys.map(() => "?").join(",") : "";
       const visibleDocWhere = scopeKeys