@leanlabsinnov/codegraph 0.1.0

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Ciril Cyriac
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/README.md

@@ -0,0 +1,50 @@
+# @cyrilc/codegraph
+
+Live, queryable knowledge graph for your codebase. Indexes JS/TS into an embedded graph
+DB with vector embeddings, then exposes a local MCP server that Claude Code, Cursor, and
+Windsurf can call to answer questions like "what calls `useAuth`?", "what's the blast
+radius of changing `formatPrice`?", or "find symbols semantically similar to a JWT auth
+helper".
+
+Zero infrastructure. The graph lives in a folder under `~/.codegraph/`. No Docker, no
+external services.
+
+## Install
+
+```bash
+npm i -g @cyrilc/codegraph
+```
+
+Requires Node 20+.
+
+## Quickstart
+
+```bash
+codegraph config llm set byo-openai
+export OPENAI_API_KEY=sk-...
+codegraph index ~/my-project
+codegraph serve
+```
+
+Then point Claude Code / Cursor / Windsurf at `http://127.0.0.1:3748/mcp` with the bearer
+token from `~/.codegraph/config.json`.
+
+See the full [README on GitHub](https://github.com/cyrilc/codegraph) for client setup, all
+10 MCP tools, and troubleshooting.
+
+## Commands
+
+| Command | What it does |
+|---|---|
+| `codegraph index <path>` | Walk the repo, parse JS/TS, embed every symbol, write to the graph |
+| `codegraph index <path> --no-embed` | Skip embeddings (faster, semantic search disabled) |
+| `codegraph status <path>` | Node + edge counts and embedding coverage |
+| `codegraph serve` | Boot the MCP server on `:3748` |
+| `codegraph doctor` | Verify Node version, config, LLM credentials, Kuzu writeable |
+| `codegraph config show` | Print the resolved config |
+| `codegraph config llm set [preset]` | Switch LLM preset (interactive when omitted) |
+| `codegraph config llm test` | Round-trip the configured provider |
+
+## License
+
+MIT

package/dist/bin.js

@@ -0,0 +1,26 @@
+#!/usr/bin/env node
+import {
+  buildProgram,
+  renderError
+} from "./chunk-O4ZO6CP5.js";
+import "./chunk-B2TIVKUB.js";
+import "./chunk-F5QKPRNW.js";
+import "./chunk-XGPZDCQ4.js";
+
+// src/bin.ts
+import { CommanderError } from "commander";
+var program = buildProgram();
+program.parseAsync(process.argv).catch((err) => {
+  if (err instanceof CommanderError) {
+    process.exit(err.exitCode);
+  }
+  if (err instanceof Error && err.name === "ExitPromptError") {
+    process.stderr.write("\nCancelled.\n");
+    process.exit(130);
+  }
+  const verbose = process.env.CODEGRAPH_VERBOSE === "1";
+  process.stderr.write(`${renderError(err, { verbose })}
+`);
+  process.exit(1);
+});
+//# sourceMappingURL=bin.js.map

package/dist/bin.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/bin.ts"],"sourcesContent":["#!/usr/bin/env node\nimport { CommanderError } from \"commander\";\nimport { buildProgram } from \"./program.js\";\nimport { renderError } from \"./ui.js\";\n\nconst program = buildProgram();\n\nprogram.parseAsync(process.argv).catch((err) => {\n  // Commander's own usage errors print themselves; don't double-render them in a box.\n  if (err instanceof CommanderError) {\n    process.exit(err.exitCode);\n  }\n  // Inquirer raises this when the user hits Ctrl-C inside a prompt. Treat as a clean exit.\n  if (err instanceof Error && err.name === \"ExitPromptError\") {\n    process.stderr.write(\"\\nCancelled.\\n\");\n    process.exit(130);\n  }\n  const verbose = process.env.CODEGRAPH_VERBOSE === \"1\";\n  process.stderr.write(`${renderError(err, { verbose })}\\n`);\n  process.exit(1);\n});\n"],"mappings":";;;;;;;;;;AACA,SAAS,sBAAsB;AAI/B,IAAM,UAAU,aAAa;AAE7B,QAAQ,WAAW,QAAQ,IAAI,EAAE,MAAM,CAAC,QAAQ;AAE9C,MAAI,eAAe,gBAAgB;AACjC,YAAQ,KAAK,IAAI,QAAQ;AAAA,EAC3B;AAEA,MAAI,eAAe,SAAS,IAAI,SAAS,mBAAmB;AAC1D,YAAQ,OAAO,MAAM,gBAAgB;AACrC,YAAQ,KAAK,GAAG;AAAA,EAClB;AACA,QAAM,UAAU,QAAQ,IAAI,sBAAsB;AAClD,UAAQ,OAAO,MAAM,GAAG,YAAY,KAAK,EAAE,QAAQ,CAAC,CAAC;AAAA,CAAI;AACzD,UAAQ,KAAK,CAAC;AAChB,CAAC;","names":[]}

package/dist/chunk-B2TIVKUB.js

@@ -0,0 +1,139 @@
+import {
+  namespaceToString
+} from "./chunk-XGPZDCQ4.js";
+
+// ../llm-router/src/router.ts
+import { embedMany, generateText } from "ai";
+async function createLlmRouter(opts) {
+  const env = opts.env ?? process.env;
+  const resolved = {
+    ...opts.config,
+    embeddingNamespace: {
+      provider: opts.config.embeddings.provider,
+      model: opts.config.embeddings.model,
+      dimension: opts.config.embeddings.dimension
+    }
+  };
+  const generationModel = await buildGenerationModel(opts.config, env);
+  const embeddingModel = await buildEmbeddingModel(opts.config.embeddings, env);
+  async function doEmbed(texts) {
+    if (texts.length === 0) return [];
+    const { embeddings } = await embedMany({ model: embeddingModel, values: texts });
+    return embeddings;
+  }
+  async function doGenerate(options) {
+    const { text } = await generateText({
+      model: generationModel,
+      ...options.system !== void 0 ? { system: options.system } : {},
+      messages: options.messages,
+      ...options.temperature !== void 0 ? { temperature: options.temperature } : {},
+      ...options.maxTokens !== void 0 ? { maxTokens: options.maxTokens } : {}
+    });
+    return text;
+  }
+  const router = {
+    config: resolved,
+    embed: doEmbed,
+    generate: doGenerate,
+    async selfTest() {
+      const embedStart = Date.now();
+      const embeds = await doEmbed(["codegraph self-test"]);
+      const embedLatencyMs = Date.now() - embedStart;
+      const genStart = Date.now();
+      const out = await doGenerate({
+        messages: [{ role: "user", content: "Reply with the single word ok." }],
+        maxTokens: 5,
+        temperature: 0
+      });
+      const generateLatencyMs = Date.now() - genStart;
+      return {
+        embedDims: embeds[0]?.length ?? 0,
+        generationOk: typeof out === "string" && out.length > 0,
+        embedLatencyMs,
+        generateLatencyMs
+      };
+    }
+  };
+  return router;
+}
+function namespaceLabel(config) {
+  return namespaceToString({
+    provider: config.embeddings.provider,
+    model: config.embeddings.model,
+    dimension: config.embeddings.dimension
+  });
+}
+async function buildGenerationModel(config, env) {
+  const provider = config.generation.provider;
+  if (provider === "openai") return openaiText(config.generation, env, config.baseUrl);
+  if (provider === "anthropic") return anthropicText(config.generation, env, config.baseUrl);
+  if (provider === "google") return googleText(config.generation, env, config.baseUrl);
+  if (provider === "ollama") return ollamaText(config.generation, env, config.baseUrl);
+  throw new Error(`Unsupported generation provider: ${provider}`);
+}
+async function buildEmbeddingModel(config, env) {
+  if (config.provider === "openai") return openaiEmbedding(config, env);
+  if (config.provider === "google") return googleEmbedding(config, env);
+  if (config.provider === "ollama") return ollamaEmbedding(config, env);
+  if (config.provider === "anthropic") {
+    return openaiEmbedding(
+      { provider: "openai", model: "text-embedding-3-small", dimension: 1536 },
+      env
+    );
+  }
+  throw new Error(`Unsupported embedding provider: ${config.provider}`);
+}
+async function openaiText(config, env, baseUrl) {
+  const { createOpenAI } = await import("@ai-sdk/openai");
+  const openai = createOpenAI({
+    ...env.OPENAI_API_KEY ? { apiKey: env.OPENAI_API_KEY } : {},
+    ...baseUrl ? { baseURL: baseUrl } : {}
+  });
+  return openai(config.model);
+}
+async function openaiEmbedding(config, env) {
+  const { createOpenAI } = await import("@ai-sdk/openai");
+  const openai = createOpenAI(env.OPENAI_API_KEY ? { apiKey: env.OPENAI_API_KEY } : {});
+  return openai.embedding(config.model);
+}
+async function anthropicText(config, env, baseUrl) {
+  const { createAnthropic } = await import("@ai-sdk/anthropic");
+  const anthropic = createAnthropic({
+    ...env.ANTHROPIC_API_KEY ? { apiKey: env.ANTHROPIC_API_KEY } : {},
+    ...baseUrl ? { baseURL: baseUrl } : {}
+  });
+  return anthropic(config.model);
+}
+async function googleText(config, env, baseUrl) {
+  const { createGoogleGenerativeAI } = await import("@ai-sdk/google");
+  const google = createGoogleGenerativeAI({
+    ...env.GOOGLE_GENERATIVE_AI_API_KEY ? { apiKey: env.GOOGLE_GENERATIVE_AI_API_KEY } : {},
+    ...baseUrl ? { baseURL: baseUrl } : {}
+  });
+  return google(config.model);
+}
+async function googleEmbedding(config, env) {
+  const { createGoogleGenerativeAI } = await import("@ai-sdk/google");
+  const google = createGoogleGenerativeAI(
+    env.GOOGLE_GENERATIVE_AI_API_KEY ? { apiKey: env.GOOGLE_GENERATIVE_AI_API_KEY } : {}
+  );
+  return google.textEmbeddingModel(config.model);
+}
+async function ollamaText(config, env, baseUrl) {
+  const { createOllama } = await import("ollama-ai-provider");
+  const base = baseUrl ?? env.OLLAMA_BASE_URL;
+  const ollama = createOllama(base ? { baseURL: `${base.replace(/\/$/, "")}/api` } : {});
+  return ollama(config.model);
+}
+async function ollamaEmbedding(config, env) {
+  const { createOllama } = await import("ollama-ai-provider");
+  const base = env.OLLAMA_BASE_URL;
+  const ollama = createOllama(base ? { baseURL: `${base.replace(/\/$/, "")}/api` } : {});
+  return ollama.embedding(config.model);
+}
+
+export {
+  createLlmRouter,
+  namespaceLabel
+};
+//# sourceMappingURL=chunk-B2TIVKUB.js.map

package/dist/chunk-B2TIVKUB.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../llm-router/src/router.ts"],"sourcesContent":["import type { EmbeddingModel, LanguageModel } from \"ai\";\nimport { embedMany, generateText } from \"ai\";\nimport type { LlmConfig, LlmEmbeddingConfig, LlmProviderConfig } from \"@codegraph/shared\";\nimport { namespaceToString } from \"@codegraph/shared\";\nimport type { GenerateOptions, LlmRouter, ResolvedLlmConfig } from \"./types.js\";\n\nexport interface CreateRouterOptions {\n  config: LlmConfig;\n  env?: NodeJS.ProcessEnv;\n}\n\nexport async function createLlmRouter(opts: CreateRouterOptions): Promise<LlmRouter> {\n  const env = opts.env ?? process.env;\n  const resolved: ResolvedLlmConfig = {\n    ...opts.config,\n    embeddingNamespace: {\n      provider: opts.config.embeddings.provider,\n      model: opts.config.embeddings.model,\n      dimension: opts.config.embeddings.dimension,\n    },\n  };\n\n  const generationModel = await buildGenerationModel(opts.config, env);\n  const embeddingModel = await buildEmbeddingModel(opts.config.embeddings, env);\n\n  async function doEmbed(texts: string[]): Promise<number[][]> {\n    if (texts.length === 0) return [];\n    const { embeddings } = await embedMany({ model: embeddingModel, values: texts });\n    return embeddings;\n  }\n\n  async function doGenerate(options: GenerateOptions): Promise<string> {\n    const { text } = await generateText({\n      model: generationModel,\n      ...(options.system !== undefined ? { system: options.system } : {}),\n      messages: options.messages,\n      ...(options.temperature !== undefined ? { temperature: options.temperature } : {}),\n      ...(options.maxTokens !== undefined ? { maxTokens: options.maxTokens } : {}),\n    });\n    return text;\n  }\n\n  const router: LlmRouter = {\n    config: resolved,\n    embed: doEmbed,\n    generate: doGenerate,\n    async selfTest() {\n      const embedStart = Date.now();\n      const embeds = await doEmbed([\"codegraph self-test\"]);\n      const embedLatencyMs = Date.now() - embedStart;\n      const genStart = Date.now();\n      const out = await doGenerate({\n        messages: [{ role: \"user\", content: \"Reply with the single word ok.\" }],\n        maxTokens: 5,\n        temperature: 0,\n      });\n      const generateLatencyMs = Date.now() - genStart;\n      return {\n        embedDims: embeds[0]?.length ?? 0,\n        generationOk: typeof out === \"string\" && out.length > 0,\n        embedLatencyMs,\n        generateLatencyMs,\n      };\n    },\n  };\n  return router;\n}\n\n/**\n * Used by the CLI's `config llm` printer; surfaces the resolved namespace\n * tag without forcing a provider import upstream.\n */\nexport function namespaceLabel(config: LlmConfig): string {\n  return namespaceToString({\n    provider: config.embeddings.provider,\n    model: config.embeddings.model,\n    dimension: config.embeddings.dimension,\n  });\n}\n\nasync function buildGenerationModel(\n  config: LlmConfig,\n  env: NodeJS.ProcessEnv,\n): Promise<LanguageModel> {\n  const provider = config.generation.provider;\n  if (provider === \"openai\") return openaiText(config.generation, env, config.baseUrl);\n  if (provider === \"anthropic\") return anthropicText(config.generation, env, config.baseUrl);\n  if (provider === \"google\") return googleText(config.generation, env, config.baseUrl);\n  if (provider === \"ollama\") return ollamaText(config.generation, env, config.baseUrl);\n  throw new Error(`Unsupported generation provider: ${provider}`);\n}\n\nasync function buildEmbeddingModel(\n  config: LlmEmbeddingConfig,\n  env: NodeJS.ProcessEnv,\n): Promise<EmbeddingModel<string>> {\n  if (config.provider === \"openai\") return openaiEmbedding(config, env);\n  if (config.provider === \"google\") return googleEmbedding(config, env);\n  if (config.provider === \"ollama\") return ollamaEmbedding(config, env);\n  if (config.provider === \"anthropic\") {\n    // Anthropic has no embedding endpoint; fall back to OpenAI so the\n    // pipeline keeps working when generation is anthropic.\n    return openaiEmbedding(\n      { provider: \"openai\", model: \"text-embedding-3-small\", dimension: 1536 },\n      env,\n    );\n  }\n  throw new Error(`Unsupported embedding provider: ${config.provider}`);\n}\n\nasync function openaiText(\n  config: LlmProviderConfig,\n  env: NodeJS.ProcessEnv,\n  baseUrl?: string,\n): Promise<LanguageModel> {\n  const { createOpenAI } = await import(\"@ai-sdk/openai\");\n  const openai = createOpenAI({\n    ...(env.OPENAI_API_KEY ? { apiKey: env.OPENAI_API_KEY } : {}),\n    ...(baseUrl ? { baseURL: baseUrl } : {}),\n  });\n  return openai(config.model);\n}\n\nasync function openaiEmbedding(\n  config: LlmEmbeddingConfig,\n  env: NodeJS.ProcessEnv,\n): Promise<EmbeddingModel<string>> {\n  const { createOpenAI } = await import(\"@ai-sdk/openai\");\n  const openai = createOpenAI(env.OPENAI_API_KEY ? { apiKey: env.OPENAI_API_KEY } : {});\n  return openai.embedding(config.model);\n}\n\nasync function anthropicText(\n  config: LlmProviderConfig,\n  env: NodeJS.ProcessEnv,\n  baseUrl?: string,\n): Promise<LanguageModel> {\n  const { createAnthropic } = await import(\"@ai-sdk/anthropic\");\n  const anthropic = createAnthropic({\n    ...(env.ANTHROPIC_API_KEY ? { apiKey: env.ANTHROPIC_API_KEY } : {}),\n    ...(baseUrl ? { baseURL: baseUrl } : {}),\n  });\n  return anthropic(config.model);\n}\n\nasync function googleText(\n  config: LlmProviderConfig,\n  env: NodeJS.ProcessEnv,\n  baseUrl?: string,\n): Promise<LanguageModel> {\n  const { createGoogleGenerativeAI } = await import(\"@ai-sdk/google\");\n  const google = createGoogleGenerativeAI({\n    ...(env.GOOGLE_GENERATIVE_AI_API_KEY\n      ? { apiKey: env.GOOGLE_GENERATIVE_AI_API_KEY }\n      : {}),\n    ...(baseUrl ? { baseURL: baseUrl } : {}),\n  });\n  return google(config.model);\n}\n\nasync function googleEmbedding(\n  config: LlmEmbeddingConfig,\n  env: NodeJS.ProcessEnv,\n): Promise<EmbeddingModel<string>> {\n  const { createGoogleGenerativeAI } = await import(\"@ai-sdk/google\");\n  const google = createGoogleGenerativeAI(\n    env.GOOGLE_GENERATIVE_AI_API_KEY\n      ? { apiKey: env.GOOGLE_GENERATIVE_AI_API_KEY }\n      : {},\n  );\n  return google.textEmbeddingModel(config.model);\n}\n\nasync function ollamaText(\n  config: LlmProviderConfig,\n  env: NodeJS.ProcessEnv,\n  baseUrl?: string,\n): Promise<LanguageModel> {\n  const { createOllama } = await import(\"ollama-ai-provider\");\n  const base = baseUrl ?? env.OLLAMA_BASE_URL;\n  const ollama = createOllama(base ? { baseURL: `${base.replace(/\\/$/, \"\")}/api` } : {});\n  return ollama(config.model);\n}\n\nasync function ollamaEmbedding(\n  config: LlmEmbeddingConfig,\n  env: NodeJS.ProcessEnv,\n): Promise<EmbeddingModel<string>> {\n  const { createOllama } = await import(\"ollama-ai-provider\");\n  const base = env.OLLAMA_BASE_URL;\n  const ollama = createOllama(base ? { baseURL: `${base.replace(/\\/$/, \"\")}/api` } : {});\n  return ollama.embedding(config.model);\n}\n"],"mappings":";;;;;AACA,SAAS,WAAW,oBAAoB;AAUxC,eAAsB,gBAAgB,MAA+C;AACnF,QAAM,MAAM,KAAK,OAAO,QAAQ;AAChC,QAAM,WAA8B;AAAA,IAClC,GAAG,KAAK;AAAA,IACR,oBAAoB;AAAA,MAClB,UAAU,KAAK,OAAO,WAAW;AAAA,MACjC,OAAO,KAAK,OAAO,WAAW;AAAA,MAC9B,WAAW,KAAK,OAAO,WAAW;AAAA,IACpC;AAAA,EACF;AAEA,QAAM,kBAAkB,MAAM,qBAAqB,KAAK,QAAQ,GAAG;AACnE,QAAM,iBAAiB,MAAM,oBAAoB,KAAK,OAAO,YAAY,GAAG;AAE5E,iBAAe,QAAQ,OAAsC;AAC3D,QAAI,MAAM,WAAW,EAAG,QAAO,CAAC;AAChC,UAAM,EAAE,WAAW,IAAI,MAAM,UAAU,EAAE,OAAO,gBAAgB,QAAQ,MAAM,CAAC;AAC/E,WAAO;AAAA,EACT;AAEA,iBAAe,WAAW,SAA2C;AACnE,UAAM,EAAE,KAAK,IAAI,MAAM,aAAa;AAAA,MAClC,OAAO;AAAA,MACP,GAAI,QAAQ,WAAW,SAAY,EAAE,QAAQ,QAAQ,OAAO,IAAI,CAAC;AAAA,MACjE,UAAU,QAAQ;AAAA,MAClB,GAAI,QAAQ,gBAAgB,SAAY,EAAE,aAAa,QAAQ,YAAY,IAAI,CAAC;AAAA,MAChF,GAAI,QAAQ,cAAc,SAAY,EAAE,WAAW,QAAQ,UAAU,IAAI,CAAC;AAAA,IAC5E,CAAC;AACD,WAAO;AAAA,EACT;AAEA,QAAM,SAAoB;AAAA,IACxB,QAAQ;AAAA,IACR,OAAO;AAAA,IACP,UAAU;AAAA,IACV,MAAM,WAAW;AACf,YAAM,aAAa,KAAK,IAAI;AAC5B,YAAM,SAAS,MAAM,QAAQ,CAAC,qBAAqB,CAAC;AACpD,YAAM,iBAAiB,KAAK,IAAI,IAAI;AACpC,YAAM,WAAW,KAAK,IAAI;AAC1B,YAAM,MAAM,MAAM,WAAW;AAAA,QAC3B,UAAU,CAAC,EAAE,MAAM,QAAQ,SAAS,iCAAiC,CAAC;AAAA,QACtE,WAAW;AAAA,QACX,aAAa;AAAA,MACf,CAAC;AACD,YAAM,oBAAoB,KAAK,IAAI,IAAI;AACvC,aAAO;AAAA,QACL,WAAW,OAAO,CAAC,GAAG,UAAU;AAAA,QAChC,cAAc,OAAO,QAAQ,YAAY,IAAI,SAAS;AAAA,QACtD;AAAA,QACA;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACA,SAAO;AACT;AAMO,SAAS,eAAe,QAA2B;AACxD,SAAO,kBAAkB;AAAA,IACvB,UAAU,OAAO,WAAW;AAAA,IAC5B,OAAO,OAAO,WAAW;AAAA,IACzB,WAAW,OAAO,WAAW;AAAA,EAC/B,CAAC;AACH;AAEA,eAAe,qBACb,QACA,KACwB;AACxB,QAAM,WAAW,OAAO,WAAW;AACnC,MAAI,aAAa,SAAU,QAAO,WAAW,OAAO,YAAY,KAAK,OAAO,OAAO;AACnF,MAAI,aAAa,YAAa,QAAO,cAAc,OAAO,YAAY,KAAK,OAAO,OAAO;AACzF,MAAI,aAAa,SAAU,QAAO,WAAW,OAAO,YAAY,KAAK,OAAO,OAAO;AACnF,MAAI,aAAa,SAAU,QAAO,WAAW,OAAO,YAAY,KAAK,OAAO,OAAO;AACnF,QAAM,IAAI,MAAM,oCAAoC,QAAQ,EAAE;AAChE;AAEA,eAAe,oBACb,QACA,KACiC;AACjC,MAAI,OAAO,aAAa,SAAU,QAAO,gBAAgB,QAAQ,GAAG;AACpE,MAAI,OAAO,aAAa,SAAU,QAAO,gBAAgB,QAAQ,GAAG;AACpE,MAAI,OAAO,aAAa,SAAU,QAAO,gBAAgB,QAAQ,GAAG;AACpE,MAAI,OAAO,aAAa,aAAa;AAGnC,WAAO;AAAA,MACL,EAAE,UAAU,UAAU,OAAO,0BAA0B,WAAW,KAAK;AAAA,MACvE;AAAA,IACF;AAAA,EACF;AACA,QAAM,IAAI,MAAM,mCAAmC,OAAO,QAAQ,EAAE;AACtE;AAEA,eAAe,WACb,QACA,KACA,SACwB;AACxB,QAAM,EAAE,aAAa,IAAI,MAAM,OAAO,gBAAgB;AACtD,QAAM,SAAS,aAAa;AAAA,IAC1B,GAAI,IAAI,iBAAiB,EAAE,QAAQ,IAAI,eAAe,IAAI,CAAC;AAAA,IAC3D,GAAI,UAAU,EAAE,SAAS,QAAQ,IAAI,CAAC;AAAA,EACxC,CAAC;AACD,SAAO,OAAO,OAAO,KAAK;AAC5B;AAEA,eAAe,gBACb,QACA,KACiC;AACjC,QAAM,EAAE,aAAa,IAAI,MAAM,OAAO,gBAAgB;AACtD,QAAM,SAAS,aAAa,IAAI,iBAAiB,EAAE,QAAQ,IAAI,eAAe,IAAI,CAAC,CAAC;AACpF,SAAO,OAAO,UAAU,OAAO,KAAK;AACtC;AAEA,eAAe,cACb,QACA,KACA,SACwB;AACxB,QAAM,EAAE,gBAAgB,IAAI,MAAM,OAAO,mBAAmB;AAC5D,QAAM,YAAY,gBAAgB;AAAA,IAChC,GAAI,IAAI,oBAAoB,EAAE,QAAQ,IAAI,kBAAkB,IAAI,CAAC;AAAA,IACjE,GAAI,UAAU,EAAE,SAAS,QAAQ,IAAI,CAAC;AAAA,EACxC,CAAC;AACD,SAAO,UAAU,OAAO,KAAK;AAC/B;AAEA,eAAe,WACb,QACA,KACA,SACwB;AACxB,QAAM,EAAE,yBAAyB,IAAI,MAAM,OAAO,gBAAgB;AAClE,QAAM,SAAS,yBAAyB;AAAA,IACtC,GAAI,IAAI,+BACJ,EAAE,QAAQ,IAAI,6BAA6B,IAC3C,CAAC;AAAA,IACL,GAAI,UAAU,EAAE,SAAS,QAAQ,IAAI,CAAC;AAAA,EACxC,CAAC;AACD,SAAO,OAAO,OAAO,KAAK;AAC5B;AAEA,eAAe,gBACb,QACA,KACiC;AACjC,QAAM,EAAE,yBAAyB,IAAI,MAAM,OAAO,gBAAgB;AAClE,QAAM,SAAS;AAAA,IACb,IAAI,+BACA,EAAE,QAAQ,IAAI,6BAA6B,IAC3C,CAAC;AAAA,EACP;AACA,SAAO,OAAO,mBAAmB,OAAO,KAAK;AAC/C;AAEA,eAAe,WACb,QACA,KACA,SACwB;AACxB,QAAM,EAAE,aAAa,IAAI,MAAM,OAAO,oBAAoB;AAC1D,QAAM,OAAO,WAAW,IAAI;AAC5B,QAAM,SAAS,aAAa,OAAO,EAAE,SAAS,GAAG,KAAK,QAAQ,OAAO,EAAE,CAAC,OAAO,IAAI,CAAC,CAAC;AACrF,SAAO,OAAO,OAAO,KAAK;AAC5B;AAEA,eAAe,gBACb,QACA,KACiC;AACjC,QAAM,EAAE,aAAa,IAAI,MAAM,OAAO,oBAAoB;AAC1D,QAAM,OAAO,IAAI;AACjB,QAAM,SAAS,aAAa,OAAO,EAAE,SAAS,GAAG,KAAK,QAAQ,OAAO,EAAE,CAAC,OAAO,IAAI,CAAC,CAAC;AACrF,SAAO,OAAO,UAAU,OAAO,KAAK;AACtC;","names":[]}

package/dist/chunk-F5QKPRNW.js

@@ -0,0 +1,359 @@
+import {
+  EDGE_KINDS,
+  NODE_KINDS
+} from "./chunk-XGPZDCQ4.js";
+
+// ../graph-db/src/client.ts
+import { mkdir } from "fs/promises";
+import { homedir } from "os";
+import { dirname, resolve } from "path";
+import * as kuzu from "kuzu";
+
+// ../graph-db/src/schema.ts
+var SYMBOL_COLUMNS = [
+  "id",
+  "kind",
+  "repoId",
+  "name",
+  "path",
+  "lineStart",
+  "lineEnd",
+  "signature",
+  "leadingComment",
+  "isExported",
+  // File-specific
+  "language",
+  "sizeBytes",
+  "contentHash",
+  // Function-specific
+  "isAsync",
+  "isArrow",
+  // Route-specific
+  "method",
+  "routePath",
+  "framework",
+  // Embedding
+  "embeddingNamespace"
+];
+var EDGE_COLUMNS = ["line"];
+function buildSchemaStatements(opts) {
+  const columnDefs = [
+    "id STRING",
+    "kind STRING",
+    "repoId STRING",
+    "name STRING",
+    "path STRING",
+    "lineStart INT64",
+    "lineEnd INT64",
+    "signature STRING",
+    "leadingComment STRING",
+    "isExported BOOLEAN",
+    "language STRING",
+    "sizeBytes INT64",
+    "contentHash STRING",
+    "isAsync BOOLEAN",
+    "isArrow BOOLEAN",
+    "method STRING",
+    "routePath STRING",
+    "framework STRING",
+    "embeddingNamespace STRING",
+    `embedding FLOAT[${opts.embeddingDimension}]`,
+    "PRIMARY KEY (id)"
+  ];
+  const statements = [
+    `CREATE NODE TABLE IF NOT EXISTS Symbol(${columnDefs.join(", ")})`
+  ];
+  for (const kind of EDGE_KINDS) {
+    statements.push(
+      `CREATE REL TABLE IF NOT EXISTS ${kind}(FROM Symbol TO Symbol, line INT64)`
+    );
+  }
+  return statements;
+}
+function buildVectorIndexStatements() {
+  return [
+    "INSTALL VECTOR",
+    "LOAD EXTENSION VECTOR",
+    "CALL CREATE_VECTOR_INDEX('Symbol', 'embedding_idx', 'embedding', metric := 'cosine')"
+  ];
+}
+var DEFAULT_EMBEDDING_DIMENSION = 1536;
+
+// ../graph-db/src/client.ts
+function defaultDbPath() {
+  return resolve(homedir(), ".codegraph", "graph");
+}
+var GraphDb = class {
+  dbPath;
+  embeddingDimension;
+  db = null;
+  conn = null;
+  vectorIndexReady = false;
+  /**
+   * Cache of `conn.prepare()` results keyed by Cypher source. Kuzu's Node SDK requires a
+   * prepared statement for any parameterized query - reusing the prepared object keeps
+   * UNWIND-batched upserts fast.
+   */
+  preparedCache = /* @__PURE__ */ new Map();
+  constructor(opts = {}) {
+    this.dbPath = opts.dbPath ?? defaultDbPath();
+    this.embeddingDimension = opts.embeddingDimension ?? DEFAULT_EMBEDDING_DIMENSION;
+  }
+  async connect() {
+    if (this.conn) return;
+    await mkdir(dirname(this.dbPath), { recursive: true });
+    this.db = new kuzu.Database(this.dbPath);
+    this.conn = new kuzu.Connection(this.db);
+  }
+  async close() {
+    this.preparedCache.clear();
+    try {
+      this.conn?.close?.();
+    } catch {
+    }
+    try {
+      this.db?.close?.();
+    } catch {
+    }
+    this.conn = null;
+    this.db = null;
+    this.vectorIndexReady = false;
+  }
+  requireConn() {
+    if (!this.conn) {
+      throw new Error("GraphDb not connected. Call connect() first.");
+    }
+    return this.conn;
+  }
+  /**
+   * Idempotent migration. Runs all DDL through `IF NOT EXISTS` guards and tolerates the
+   * vector-index "already exists" error from re-runs.
+   */
+  async migrate() {
+    await this.connect();
+    const schemaStmts = buildSchemaStatements({ embeddingDimension: this.embeddingDimension });
+    for (const stmt of schemaStmts) {
+      await this.exec(stmt);
+    }
+    for (const stmt of buildVectorIndexStatements()) {
+      try {
+        await this.exec(stmt);
+      } catch (err) {
+        const message = err instanceof Error ? err.message : String(err);
+        if (isAlreadyExistsError(message)) continue;
+        if (/extension/i.test(message) && /(not found|missing|unsupported|disabled)/i.test(message)) {
+          console.warn(
+            `[codegraph] vector extension unavailable; semantic search disabled. Underlying: ${message}`
+          );
+          return;
+        }
+        throw new Error(`migrate failed on \`${stmt}\`: ${message}`);
+      }
+    }
+    this.vectorIndexReady = true;
+  }
+  /**
+   * Typed Cypher escape hatch.
+   *
+   * Kuzu returns BIGINT columns as native BigInt; we coerce to plain `number` when safe so
+   * downstream JSON serialization (MCP responses, snapshot tests) does not need bespoke
+   * handling.
+   */
+  async query(cypher, params = {}) {
+    const result = await this.runQuery(cypher, params);
+    const raw = await collectAll(result);
+    const data = raw.map((row) => normalizeRow(row));
+    const headers = raw.length > 0 ? Object.keys(raw[0] ?? {}) : [];
+    return { data, headers, metadata: [] };
+  }
+  /** Fire-and-forget DDL/exec. */
+  async exec(cypher, params = {}) {
+    await this.runQuery(cypher, params);
+  }
+  /**
+   * Bridge to Kuzu's two execution paths:
+   * - `conn.query(stmt)` for unparameterized statements (the second positional arg is a
+   *   `progressCallback`, NOT params - mistaking that is the #1 way to confuse the API).
+   * - `conn.prepare(stmt) + conn.execute(prepared, params)` for anything with `$name`
+   *   placeholders. We cache the prepared statement so UNWIND batches reuse it.
+   */
+  async runQuery(cypher, params) {
+    const conn = this.requireConn();
+    if (Object.keys(params).length === 0) {
+      return conn.query(cypher);
+    }
+    let prepared = this.preparedCache.get(cypher);
+    if (!prepared) {
+      prepared = await conn.prepare(cypher);
+      if (!prepared.isSuccess()) {
+        throw new Error(prepared.getErrorMessage());
+      }
+      this.preparedCache.set(cypher, prepared);
+    }
+    return conn.execute(prepared, params);
+  }
+  /**
+   * Upserts nodes via batched UNWIND + MERGE. Kuzu requires every column we SET to exist
+   * in the schema, so each row is normalized to include every `SYMBOL_COLUMNS` field (NULL
+   * for fields not present).
+   */
+  async upsertNodes(nodes) {
+    if (nodes.length === 0) return;
+    await this.connect();
+    const BATCH = 200;
+    const setClause = SYMBOL_COLUMNS.filter((c) => c !== "id").map((c) => `n.${c} = r.${c}`).join(", ");
+    const cypher = `UNWIND $batch AS r MERGE (n:Symbol {id: r.id}) SET ${setClause}`;
+    for (let i = 0; i < nodes.length; i += BATCH) {
+      const slice = nodes.slice(i, i + BATCH);
+      const payload = slice.map(buildSymbolRow);
+      await this.exec(cypher, { batch: payload });
+    }
+  }
+  /**
+   * Upserts edges. Both endpoints must already exist as `Symbol` nodes; rows where the
+   * MATCH fails are silently dropped, matching Cypher semantics.
+   *
+   * Uses CREATE because the orchestrator wipes the repo's slice before writing, so
+   * duplicates can't pre-exist within a single index pass.
+   */
+  async upsertEdges(edges) {
+    if (edges.length === 0) return;
+    await this.connect();
+    const byKind = /* @__PURE__ */ new Map();
+    for (const e of edges) {
+      const bucket = byKind.get(e.kind);
+      if (bucket) bucket.push(e);
+      else byKind.set(e.kind, [e]);
+    }
+    const BATCH = 500;
+    for (const [kind, batch] of byKind) {
+      const cypher = `UNWIND $batch AS r MATCH (a:Symbol {id: r.fromId}) MATCH (b:Symbol {id: r.toId}) CREATE (a)-[e:${kind} {line: r.line}]->(b)`;
+      for (let i = 0; i < batch.length; i += BATCH) {
+        const slice = batch.slice(i, i + BATCH);
+        const payload = slice.map((e) => ({
+          fromId: e.fromId,
+          toId: e.toId,
+          line: typeof e.line === "number" ? e.line : null
+        }));
+        await this.exec(cypher, { batch: payload });
+      }
+    }
+  }
+  /**
+   * Deletes all nodes (and incident edges via DETACH DELETE) for a repo. If `paths` is
+   * provided, restricts the delete to nodes whose `path` is in the list - used by
+   * incremental re-indexing.
+   */
+  async deleteByRepo(repoId, paths) {
+    await this.connect();
+    if (paths && paths.length > 0) {
+      await this.exec(
+        "MATCH (n:Symbol) WHERE n.repoId = $repoId AND n.path IN $paths DETACH DELETE n",
+        { repoId, paths }
+      );
+      return;
+    }
+    await this.exec("MATCH (n:Symbol) WHERE n.repoId = $repoId DETACH DELETE n", { repoId });
+  }
+  /**
+   * Returns counts of nodes (per kind) and edges (per kind) for a repo, plus the share of
+   * non-File nodes that carry an embedding.
+   */
+  async stats(repoId) {
+    await this.connect();
+    const nodes = {};
+    for (const kind of NODE_KINDS) {
+      const r = await this.query(
+        "MATCH (n:Symbol) WHERE n.repoId = $repoId AND n.kind = $kind RETURN count(n) AS count",
+        { repoId, kind }
+      );
+      nodes[kind] = Number(r.data[0]?.count ?? 0);
+    }
+    const edges = {};
+    for (const kind of EDGE_KINDS) {
+      const r = await this.query(
+        `MATCH (a:Symbol)-[r:${kind}]->(b:Symbol)
+           WHERE a.repoId = $repoId AND b.repoId = $repoId
+           RETURN count(r) AS count`,
+        { repoId }
+      );
+      edges[kind] = Number(r.data[0]?.count ?? 0);
+    }
+    const cov = await this.query(
+      `MATCH (n:Symbol)
+         WHERE n.repoId = $repoId AND n.kind <> 'File'
+         RETURN count(n) AS total,
+                count(n.embedding) AS embedded`,
+      { repoId }
+    );
+    const row = cov.data[0];
+    const total = Number(row?.total ?? 0);
+    const embedded = Number(row?.embedded ?? 0);
+    const coverage = total === 0 ? 0 : embedded / total;
+    return { nodes, edges, embeddingCoverage: coverage };
+  }
+  /** True once `migrate()` confirmed the vector extension is loaded. */
+  hasVectorIndex() {
+    return this.vectorIndexReady;
+  }
+};
+function buildSymbolRow(node) {
+  const src = node;
+  const row = {};
+  for (const col of SYMBOL_COLUMNS) {
+    const value = src[col];
+    row[col] = value === void 0 ? null : value;
+  }
+  row.id = node.id;
+  row.kind = node.kind;
+  return row;
+}
+function normalizeRow(row) {
+  if (row instanceof Map) {
+    const out = {};
+    for (const [k, v] of row) {
+      out[String(k)] = coerceValue(v);
+    }
+    return out;
+  }
+  if (row && typeof row === "object") {
+    const src = row;
+    const out = {};
+    for (const k of Object.keys(src)) {
+      out[k] = coerceValue(src[k]);
+    }
+    return out;
+  }
+  return { value: coerceValue(row) };
+}
+function coerceValue(value) {
+  if (typeof value === "bigint") {
+    if (value <= BigInt(Number.MAX_SAFE_INTEGER) && value >= BigInt(Number.MIN_SAFE_INTEGER)) {
+      return Number(value);
+    }
+    return value.toString();
+  }
+  if (Array.isArray(value)) return value.map(coerceValue);
+  return value;
+}
+async function collectAll(result) {
+  const target = Array.isArray(result) ? result[result.length - 1] : result;
+  if (!target) return [];
+  const getAll = target.getAll;
+  if (typeof getAll !== "function") return [];
+  return getAll.call(target);
+}
+function isAlreadyExistsError(message) {
+  return /already exists/i.test(message) || /already loaded/i.test(message) || /already installed/i.test(message) || /duplicate (table|index)/i.test(message);
+}
+
+export {
+  SYMBOL_COLUMNS,
+  EDGE_COLUMNS,
+  buildSchemaStatements,
+  buildVectorIndexStatements,
+  DEFAULT_EMBEDDING_DIMENSION,
+  defaultDbPath,
+  GraphDb
+};
+//# sourceMappingURL=chunk-F5QKPRNW.js.map