feat(web): wire agent/chat through HTTP in backend mode

Add HTTP wrapper functions (executeQuery, textSearch) to the ingestion
worker and a new initializeBackendAgent method that creates the agent
with HTTP-backed tool closures instead of local KuzuDB. useAppState
detects backend mode and routes agent initialization accordingly.

README.md

@@ -32,7 +32,7 @@ https://github.com/user-attachments/assets/172685ba-8e54-4ea7-9ad1-e31a3398da72
 | ----------------- | -------------------------------------------------------------- | ------------------------------------------------------------ |
 | **What**    | Index repos locally, connect AI agents via MCP                 | Visual graph explorer + AI chat in browser                   |
 | **For**     | Daily development with Cursor, Claude Code, Windsurf, OpenCode | Quick exploration, demos, one-off analysis                   |
-| **Scale**   | Full repos, any size                                           | Limited by browser memory (~5k files)                        |
+| **Scale**   | Full repos, any size                                           | Limited by browser memory (~5k files), or unlimited via backend mode |
 | **Install** | `npm install -g gitnexus`                                    | No install —[gitnexus.vercel.app](https://gitnexus.vercel.app) |
 | **Storage** | KuzuDB native (fast, persistent)                               | KuzuDB WASM (in-memory, per session)                         |
 | **Parsing** | Tree-sitter native bindings                                    | Tree-sitter WASM                                             |
@@ -235,7 +235,7 @@ npm run dev
 
 The web UI uses the same indexing pipeline as the CLI but runs entirely in WebAssembly (Tree-sitter WASM, KuzuDB WASM, in-browser embeddings). It's great for quick exploration but limited by browser memory for larger repos.
 
-**Local Backend Mode:** Run `gitnexus serve` and open the web UI locally — it auto-detects the server and shows all your indexed repos. No need to re-upload or re-index.
+**Local Backend Mode:** Run `gitnexus serve` and open the web UI locally — it auto-detects the server and shows all your indexed repos, with full AI chat support. No need to re-upload or re-index. The agent's tools (Cypher queries, search, code navigation) route through the backend HTTP API automatically.
 
 ---
 

gitnexus-web/src/hooks/useAppState.tsx

@@ -11,7 +11,7 @@ import type { LLMSettings, ProviderConfig, AgentStreamChunk, ChatMessage, ToolCa
 import { loadSettings, getActiveProviderConfig, saveSettings } from '../core/llm/settings-service';
 import type { AgentMessage } from '../core/llm/agent';
 import { DEFAULT_VISIBLE_EDGES, type EdgeType } from '../lib/constants';
-import { runCypherQuery } from '../services/backend';
+import { runCypherQuery, getBackendUrl } from '../services/backend';
 
 export type ViewMode = 'onboarding' | 'loading' | 'exploring';
 export type RightPanelTab = 'code' | 'chat';
@@ -571,25 +571,39 @@ export const AppStateProvider = ({ children }: { children: ReactNode }) => {
   }, []);
 
   const initializeAgent = useCallback(async (overrideProjectName?: string): Promise<void> => {
-    const api = apiRef.current;
-    if (!api) {
-      setAgentError('Worker not initialized');
-      return;
-    }
-
     const config = getActiveProviderConfig();
     if (!config) {
       setAgentError('Please configure an LLM provider in settings');
       return;
     }
 
+    const api = apiRef.current;
+    if (!api) {
+      setAgentError('Worker not initialized');
+      return;
+    }
+
     setIsAgentInitializing(true);
     setAgentError(null);
 
     try {
-      // Use override if provided (for fresh loads), fallback to state (for re-init)
       const effectiveProjectName = overrideProjectName || projectName || 'project';
-      const result = await api.initializeAgent(config, effectiveProjectName);
+      let result: { success: boolean; error?: string };
+
+      if (isBackendMode && backendRepo) {
+        // Backend mode: pass HTTP config + file contents to worker
+        const entries = Array.from(fileContents.entries());
+        result = await api.initializeBackendAgent(
+          config,
+          getBackendUrl(),
+          backendRepo,
+          entries,
+          effectiveProjectName,
+        );
+      } else {
+        result = await api.initializeAgent(config, effectiveProjectName);
+      }
+
       if (result.success) {
         setIsAgentReady(true);
         setAgentError(null);
@@ -607,7 +621,7 @@ export const AppStateProvider = ({ children }: { children: ReactNode }) => {
     } finally {
       setIsAgentInitializing(false);
     }
-  }, [projectName]);
+  }, [projectName, isBackendMode, backendRepo, fileContents]);
 
   const sendChatMessage = useCallback(async (message: string): Promise<void> => {
     const api = apiRef.current;

gitnexus-web/src/workers/ingestion.worker.ts

@@ -16,7 +16,7 @@ import { SystemMessage } from '@langchain/core/messages';
 import { enrichClustersBatch, ClusterMemberInfo, ClusterEnrichment } from '../core/ingestion/cluster-enricher';
 import { CommunityNode } from '../core/ingestion/community-processor';
 import { PipelineResult } from '../types/pipeline';
-import { buildCodebaseContext } from '../core/llm/context-builder';
+import { buildCodebaseContext, type CodebaseContext } from '../core/llm/context-builder';
 import { 
   buildBM25Index, 
   searchBM25, 
@@ -54,6 +54,65 @@ let enrichmentCancelled = false;
 // Chat cancellation flag
 let chatCancelled = false;
 
+// ============================================================
+// HTTP helpers for backend mode
+// ============================================================
+
+const httpFetchWithTimeout = async (
+  url: string,
+  init: RequestInit = {},
+  timeoutMs: number = 30_000,
+): Promise<Response> => {
+  const controller = new AbortController();
+  const timer = setTimeout(() => controller.abort(), timeoutMs);
+  try {
+    return await fetch(url, { ...init, signal: controller.signal });
+  } finally {
+    clearTimeout(timer);
+  }
+};
+
+const createHttpExecuteQuery = (backendUrl: string, repo: string) => {
+  return async (cypher: string): Promise<any[]> => {
+    const response = await httpFetchWithTimeout(`${backendUrl}/api/query`, {
+      method: 'POST',
+      headers: { 'Content-Type': 'application/json' },
+      body: JSON.stringify({ cypher, repo }),
+    });
+    if (!response.ok) {
+      const body = await response.json().catch(() => ({}));
+      throw new Error(body.error || `Backend query failed: ${response.status}`);
+    }
+    const body = await response.json();
+    return (body.result ?? body) as any[];
+  };
+};
+
+const createHttpTextSearch = (executeQuery: (cypher: string) => Promise<any[]>) => {
+  return async (query: string, k: number = 15): Promise<any[]> => {
+    const sanitized = query.replace(/'/g, "\\'");
+    const cypher = `
+      MATCH (n)
+      WHERE n.name IS NOT NULL AND toLower(n.name) CONTAINS toLower('${sanitized}')
+      RETURN n.id AS id, label(n) AS type, n.name AS name, n.filePath AS filePath, n.content AS content
+      LIMIT ${k}
+    `;
+    try {
+      const results = await executeQuery(cypher);
+      return results.map((r: any, i: number) => ({
+        id: r.id,
+        name: r.name,
+        type: r.type,
+        filePath: r.filePath,
+        content: r.content ?? '',
+        score: 1 - (i * 0.05),
+      }));
+    } catch {
+      return [];
+    }
+  };
+};
+
 /**
  * Worker API exposed via Comlink
  * 
@@ -540,6 +599,67 @@ const workerApi = {
     }
   },
 
+  /**
+   * Initialize the Graph RAG agent in backend mode (HTTP-backed tools).
+   * Uses HTTP wrappers instead of local KuzuDB for all tool queries.
+   * @param config - Provider configuration for the LLM
+   * @param backendUrl - Base URL of the gitnexus serve backend
+   * @param repoName - Repository name on the backend
+   * @param fileContentsEntries - File contents as [path, content][] (Comlink can't transfer Maps)
+   * @param projectName - Display name for the project
+   */
+  async initializeBackendAgent(
+    config: ProviderConfig,
+    backendUrl: string,
+    repoName: string,
+    fileContentsEntries: [string, string][],
+    projectName?: string,
+  ): Promise<{ success: boolean; error?: string }> {
+    try {
+      // Rebuild Map from serializable entries (Comlink can't transfer Maps)
+      const contents = new Map<string, string>(fileContentsEntries);
+      storedFileContents = contents;
+
+      // Create HTTP-based tool wrappers
+      const executeQuery = createHttpExecuteQuery(backendUrl, repoName);
+      const textSearch = createHttpTextSearch(executeQuery);
+
+      // Build codebase context (uses Cypher queries — works via HTTP)
+      let codebaseContext: CodebaseContext | undefined;
+      try {
+        codebaseContext = await buildCodebaseContext(executeQuery, projectName || repoName);
+      } catch {
+        // Non-fatal — agent works without context
+      }
+
+      // Create agent with HTTP-backed tools
+      currentAgent = createGraphRAGAgent(
+        config,
+        executeQuery,          // Cypher via HTTP
+        textSearch,            // semanticSearch → Cypher text search
+        textSearch,            // semanticSearchWithContext → same
+        textSearch,            // hybridSearch → same
+        () => true,            // isEmbeddingReady → always true (routes to textSearch)
+        () => true,            // isBM25Ready → always true (routes to textSearch)
+        contents,              // fileContents Map
+        codebaseContext,
+      );
+
+      currentProviderConfig = config;
+
+      if (import.meta.env.DEV) {
+        console.log('🤖 Backend agent initialized with provider:', config.provider);
+      }
+
+      return { success: true };
+    } catch (err: any) {
+      if (import.meta.env.DEV) {
+        console.error('❌ Backend agent initialization failed:', err);
+      }
+      return { success: false, error: err.message || 'Failed to initialize backend agent' };
+    }
+  },
+
   /**
    * Check if the agent is initialized
    */

gitnexus/README.md

@@ -179,7 +179,7 @@ Installed automatically by both `gitnexus analyze` (per-repo) and `gitnexus setu
 
 GitNexus also has a browser-based UI at [gitnexus.vercel.app](https://gitnexus.vercel.app) — 100% client-side, your code never leaves the browser.
 
-**Local Backend Mode:** Run `gitnexus serve` and open the web UI locally — it auto-detects the server and shows all your indexed repos. No need to re-upload or re-index.
+**Local Backend Mode:** Run `gitnexus serve` and open the web UI locally — it auto-detects the server and shows all your indexed repos, with full AI chat support. No need to re-upload or re-index. The agent's tools (Cypher queries, search, code navigation) route through the backend HTTP API automatically.
 
 ## License