doccupine 0.0.45 → 0.0.47

package/dist/templates/env.example.d.ts

@@ -1 +1 @@
-export declare const envExampleTemplate = "# LLM Provider Configuration\n# Choose your preferred LLM provider: openai, anthropic, or google\nLLM_PROVIDER=openai\n\n# API Keys (set the one matching your provider)\nOPENAI_API_KEY=your_openai_api_key_here\nANTHROPIC_API_KEY=your_anthropic_api_key_here\nGOOGLE_API_KEY=your_google_api_key_here\n\n# Optional: Override default models\n# OpenAI models: gpt-4.1-nano, gpt-5-mini, gpt-5-nano, gpt-5\n# Anthropic models: claude-sonnet-4-5-20250929, claude-haiku-4-5-20251001, claude-opus-4-5-20251101\n# Google models: gemini-2.5-flash-lite, gemini-2.5-pro, gemini-2.5-flash\n# LLM_CHAT_MODEL=gpt-4.1-nano\n\n# Optional: Override default embedding model\n# OpenAI: text-embedding-3-small, text-embedding-3-large\n# Google: text-embedding-004\n# Note: Anthropic doesn't provide embeddings, will fallback to OpenAI\n# LLM_EMBEDDING_MODEL=text-embedding-3-small\n\n# Optional: Set temperature (0-1, default: 1)\n# LLM_TEMPERATURE=1\n";
+export declare const envExampleTemplate = "# LLM Provider Configuration\n# Choose your preferred LLM provider: openai, anthropic, or google\nLLM_PROVIDER=openai\n\n# API Keys (set the one matching your provider)\nOPENAI_API_KEY=your_openai_api_key_here\nANTHROPIC_API_KEY=your_anthropic_api_key_here\nGOOGLE_API_KEY=your_google_api_key_here\n\n# Optional: Override default models\n# OpenAI models: gpt-4.1-nano, gpt-5-mini, gpt-5-nano, gpt-5\n# Anthropic models: claude-sonnet-4-5-20250929, claude-haiku-4-5-20251001, claude-opus-4-5-20251101\n# Google models: gemini-2.5-flash-lite, gemini-2.5-pro, gemini-2.5-flash\n# LLM_CHAT_MODEL=gpt-4.1-nano\n\n# Optional: Override default embedding model\n# OpenAI: text-embedding-3-small, text-embedding-3-large\n# Google: text-embedding-004\n# Note: Anthropic doesn't provide embeddings, will fallback to OpenAI\n# LLM_EMBEDDING_MODEL=text-embedding-3-small\n\n# Optional: Set temperature (0-1, default: 0)\n# LLM_TEMPERATURE=0\n";

package/dist/templates/env.example.js

@@ -19,6 +19,6 @@ GOOGLE_API_KEY=your_google_api_key_here
 # Note: Anthropic doesn't provide embeddings, will fallback to OpenAI
 # LLM_EMBEDDING_MODEL=text-embedding-3-small
 
-# Optional: Set temperature (0-1, default: 1)
-# LLM_TEMPERATURE=1
+# Optional: Set temperature (0-1, default: 0)
+# LLM_TEMPERATURE=0
 `;

package/dist/templates/mdx/ai-assistant.mdx.d.ts

@@ -1 +1 @@
-export declare const aiAssistantMdxTemplate = "---\ntitle: \"AI Assistant\"\ndescription: \"Integrate AI capabilities into your Doccupine documentation using OpenAI, Anthropic, or Google Gemini.\"\ndate: \"2025-01-24\"\ncategory: \"Configuration\"\ncategoryOrder: 3\norder: 7\n---\n# AI Assistant\nDoccupine supports AI integration to enhance your documentation experience. You can use OpenAI, Anthropic, or Google Gemini to power AI features in your documentation site. The AI assistant uses your documentation content as context, allowing users to ask questions about your docs and receive accurate answers based on the documentation.\n\n## Setup\nTo enable AI features, create an `.env` file in the directory where your website is generated. By default, this is the `nextjs-app/` directory.\n\n## Configuration\nCreate an `.env` file with the following configuration options:\n\n```env\n# LLM Provider Configuration\n# Choose your preferred LLM provider: openai, anthropic, or google\nLLM_PROVIDER=openai\n\n# API Keys (set the one matching your provider)\nOPENAI_API_KEY=your_openai_api_key_here\nANTHROPIC_API_KEY=your_anthropic_api_key_here\nGOOGLE_API_KEY=your_google_api_key_here\n\n# Optional: Override default models\n# OpenAI models: gpt-4.1-nano, gpt-5-mini, gpt-5-nano, gpt-5\n# Anthropic models: claude-sonnet-4-5-20250929, claude-haiku-4-5-20251001, claude-opus-4-5-20251101\n# Google models: gemini-2.5-flash-lite, gemini-2.5-pro, gemini-2.5-flash\n# LLM_CHAT_MODEL=gpt-4.1-nano\n\n# Optional: Override default embedding model\n# OpenAI: text-embedding-3-small, text-embedding-3-large\n# Google: text-embedding-004\n# Note: Anthropic doesn't provide embeddings, will fallback to OpenAI\n# LLM_EMBEDDING_MODEL=text-embedding-3-small\n\n# Optional: Set temperature (0-1, default: 1)\n# LLM_TEMPERATURE=1\n```\n\n## Provider Selection\nSet `LLM_PROVIDER` to one of the following values:\n- `openai` - Use OpenAI's models (GPT-4.1, GPT-4.1-mini, GPT-4.1-nano)\n- `anthropic` - Use Anthropic's models (Claude Sonnet 4.5, Claude Haiku 4.5, Claude Opus 4.5)\n- `google` - Use Google's models (Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 2.5 Flash-Lite)\n\n## API Keys\nYou need to set the API key that matches your chosen provider:\n- For OpenAI: Set `OPENAI_API_KEY`\n- For Anthropic: Set `ANTHROPIC_API_KEY`\n- For Google: Set `GOOGLE_API_KEY`\n\n<Callout type=\"warning\">\n  Keep your API keys secure. Never commit your `.env` file to version control.\n</Callout>\n\n<Callout type=\"note\">\n  Doccupine automatically adds `.env` to your `.gitignore` file.\n</Callout>\n\n## Using Anthropic with OpenAI\nIf you want to use Anthropic as your LLM provider, you must also have an OpenAI API key set. Here's why:\n\n### The Situation\nAnthropic (Claude) does not provide an embeddings API. They only offer chat/completion models, not text embeddings.\n\nYour RAG (Retrieval-Augmented Generation) system has two components:\n- **Chat/Completion** - Generates answers, works with Anthropic.\n- **Embeddings** - Creates vector representations of text for search, Anthropic doesn't provide this.\n\nWhen using Anthropic as your `LLM_PROVIDER`, Doccupine will use Anthropic for chat/completion tasks, but will automatically fallback to OpenAI for embeddings. This means you need both API keys configured:\n\n```env\nLLM_PROVIDER=anthropic\nANTHROPIC_API_KEY=your_anthropic_api_key_here\nOPENAI_API_KEY=your_openai_api_key_here\n```\n\nThis hybrid approach allows you to leverage Anthropic's powerful chat models while still having access to embeddings functionality through OpenAI.\n\n## Optional Settings\n\n### Chat Model\nOverride the default chat model by uncommenting and setting `LLM_CHAT_MODEL`. You can use any available model from your chosen provider. Example models include:\n- **OpenAI**: `gpt-4.1-nano`, `gpt-4.1-mini`, `gpt-4.1`\n- **Anthropic**: `claude-sonnet-4-5-20250929`, `claude-haiku-4-5-20251001`, `claude-opus-4-5-20251101`\n- **Google**: `gemini-2.5-flash-lite`, `gemini-2.5-pro`, `gemini-2.5-flash`\n\nFor a complete list of available models, refer to the official documentation:\n- [OpenAI Models](https://platform.openai.com/docs/models)\n- [Anthropic Models](https://docs.anthropic.com/claude/docs/models-overview)\n- [Google Gemini Models](https://ai.google.dev/models/gemini)\n\n### Embedding Model\nOverride the default embedding model by uncommenting and setting `LLM_EMBEDDING_MODEL`:\n- **OpenAI**: `text-embedding-3-small`, `text-embedding-3-large`\n- **Google**: `text-embedding-004`\n- **Anthropic**: Anthropic doesn't provide embeddings. If you use Anthropic as your provider, Doccupine will fallback to OpenAI for embeddings.\n\n### Temperature\nControl the randomness of AI responses by setting `LLM_TEMPERATURE` to a value between 0 and 1:\n- `0` - More deterministic and focused responses (default)\n- `1` - More creative and varied responses";
+export declare const aiAssistantMdxTemplate = "---\ntitle: \"AI Assistant\"\ndescription: \"Integrate AI capabilities into your Doccupine documentation using OpenAI, Anthropic, or Google Gemini.\"\ndate: \"2025-01-24\"\ncategory: \"Configuration\"\ncategoryOrder: 3\norder: 7\n---\n# AI Assistant\nDoccupine supports AI integration to enhance your documentation experience. You can use OpenAI, Anthropic, or Google Gemini to power AI features in your documentation site. The AI assistant uses your documentation content as context, allowing users to ask questions about your docs and receive accurate answers based on the documentation.\n\n## Setup\nTo enable AI features, create an `.env` file in the directory where your website is generated. By default, this is the `nextjs-app/` directory.\n\n## Configuration\nCreate an `.env` file with the following configuration options:\n\n```env\n# LLM Provider Configuration\n# Choose your preferred LLM provider: openai, anthropic, or google\nLLM_PROVIDER=openai\n\n# API Keys (set the one matching your provider)\nOPENAI_API_KEY=your_openai_api_key_here\nANTHROPIC_API_KEY=your_anthropic_api_key_here\nGOOGLE_API_KEY=your_google_api_key_here\n\n# Optional: Override default models\n# OpenAI models: gpt-4.1-nano, gpt-5-mini, gpt-5-nano, gpt-5\n# Anthropic models: claude-sonnet-4-5-20250929, claude-haiku-4-5-20251001, claude-opus-4-5-20251101\n# Google models: gemini-2.5-flash-lite, gemini-2.5-pro, gemini-2.5-flash\n# LLM_CHAT_MODEL=gpt-4.1-nano\n\n# Optional: Override default embedding model\n# OpenAI: text-embedding-3-small, text-embedding-3-large\n# Google: text-embedding-004\n# Note: Anthropic doesn't provide embeddings, will fallback to OpenAI\n# LLM_EMBEDDING_MODEL=text-embedding-3-small\n\n# Optional: Set temperature (0-1, default: 0)\n# LLM_TEMPERATURE=0\n```\n\n## Provider Selection\nSet `LLM_PROVIDER` to one of the following values:\n- `openai` - Use OpenAI's models (GPT-4.1, GPT-4.1-mini, GPT-4.1-nano)\n- `anthropic` - Use Anthropic's models (Claude Sonnet 4.5, Claude Haiku 4.5, Claude Opus 4.5)\n- `google` - Use Google's models (Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 2.5 Flash-Lite)\n\n## API Keys\nYou need to set the API key that matches your chosen provider:\n- For OpenAI: Set `OPENAI_API_KEY`\n- For Anthropic: Set `ANTHROPIC_API_KEY`\n- For Google: Set `GOOGLE_API_KEY`\n\n<Callout type=\"warning\">\n  Keep your API keys secure. Never commit your `.env` file to version control.\n</Callout>\n\n<Callout type=\"note\">\n  Doccupine automatically adds `.env` to your `.gitignore` file.\n</Callout>\n\n## Using Anthropic with OpenAI\nIf you want to use Anthropic as your LLM provider, you must also have an OpenAI API key set. Here's why:\n\n### The Situation\nAnthropic (Claude) does not provide an embeddings API. They only offer chat/completion models, not text embeddings.\n\nYour RAG (Retrieval-Augmented Generation) system has two components:\n- **Chat/Completion** - Generates answers, works with Anthropic.\n- **Embeddings** - Creates vector representations of text for search, Anthropic doesn't provide this.\n\nWhen using Anthropic as your `LLM_PROVIDER`, Doccupine will use Anthropic for chat/completion tasks, but will automatically fallback to OpenAI for embeddings. This means you need both API keys configured:\n\n```env\nLLM_PROVIDER=anthropic\nANTHROPIC_API_KEY=your_anthropic_api_key_here\nOPENAI_API_KEY=your_openai_api_key_here\n```\n\nThis hybrid approach allows you to leverage Anthropic's powerful chat models while still having access to embeddings functionality through OpenAI.\n\n## Optional Settings\n\n### Chat Model\nOverride the default chat model by uncommenting and setting `LLM_CHAT_MODEL`. You can use any available model from your chosen provider. Example models include:\n- **OpenAI**: `gpt-4.1-nano`, `gpt-4.1-mini`, `gpt-4.1`\n- **Anthropic**: `claude-sonnet-4-5-20250929`, `claude-haiku-4-5-20251001`, `claude-opus-4-5-20251101`\n- **Google**: `gemini-2.5-flash-lite`, `gemini-2.5-pro`, `gemini-2.5-flash`\n\nFor a complete list of available models, refer to the official documentation:\n- [OpenAI Models](https://platform.openai.com/docs/models)\n- [Anthropic Models](https://docs.anthropic.com/claude/docs/models-overview)\n- [Google Gemini Models](https://ai.google.dev/models/gemini)\n\n### Embedding Model\nOverride the default embedding model by uncommenting and setting `LLM_EMBEDDING_MODEL`:\n- **OpenAI**: `text-embedding-3-small`, `text-embedding-3-large`\n- **Google**: `text-embedding-004`\n- **Anthropic**: Anthropic doesn't provide embeddings. If you use Anthropic as your provider, Doccupine will fallback to OpenAI for embeddings.\n\n### Temperature\nControl the randomness of AI responses by setting `LLM_TEMPERATURE` to a value between 0 and 1:\n- `0` - More deterministic and focused responses (default)\n- `1` - More creative and varied responses";

package/dist/templates/mdx/ai-assistant.mdx.js

@@ -37,8 +37,8 @@ GOOGLE_API_KEY=your_google_api_key_here
 # Note: Anthropic doesn't provide embeddings, will fallback to OpenAI
 # LLM_EMBEDDING_MODEL=text-embedding-3-small
 
-# Optional: Set temperature (0-1, default: 1)
-# LLM_TEMPERATURE=1
+# Optional: Set temperature (0-1, default: 0)
+# LLM_TEMPERATURE=0
 \`\`\`
 
 ## Provider Selection

package/dist/templates/mdx/model-context-protocol.mdx.d.ts

@@ -1 +1 @@
-export declare const mcpMdxTemplate = "---\ntitle: \"Model Context Protocol\"\ndescription: \"Connect your Doccupine documentation to AI tools with an MCP server for enhanced AI-powered documentation search.\"\ndate: \"2025-01-24\"\ncategory: \"Configuration\"\ncategoryOrder: 3\norder: 8\n---\n# Model Context Protocol\nConnect your documentation to AI tools with a hosted MCP server.\n\nDoccupine automatically generates a Model Context Protocol (MCP) server from your documentation, making your content accessible to AI applications like Claude, Cursor, VS Code, and other MCP-compatible tools. Your MCP server exposes semantic search capabilities, allowing AI tools to query your documentation directly and provide accuraxte, context-aware answers.\n\n## About MCP servers\n\nThe Model Context Protocol (MCP) is an open protocol that creates standardized connections between AI applications and external services, like documentation. Doccupine generates an MCP server from your documentation, preparing your content for the broader AI ecosystem where any MCP client can connect to your documentation.\n\nYour MCP server exposes search and retrieval tools for AI applications to query your documentation. Your users must connect your MCP server to their preferred AI tools to access your documentation.\n\n### How MCP servers work\n\nWhen an AI tool has your documentation MCP server connected, the AI tool can search your documentation directly instead of making a generic web search in response to a user's prompt. Your MCP server provides access to all indexed content from your documentation site.\n\n* The LLM can proactively search your documentation while generating a response, not just when explicitly asked.\n* The LLM determines when to use the search tool based on the context of the conversation and the relevance of your documentation.\n* Each tool call happens during the generation process, so the LLM searches up-to-date information from your documentation to generate its response.\n\n### MCP compared to web search\n\nAI tools can search the web, but MCP provides distinct advantages for documentation.\n\n* **Direct source access**: Web search depends on what search engines have indexed, which may be stale or incomplete. MCP searches your current indexed documentation directly.\n* **Integrated workflow**: MCP allows the AI to search during response generation rather than performing a separate web search.\n* **Semantic search**: MCP uses vector embeddings for semantic similarity search, providing more relevant results than keyword-based web search.\n* **No search noise**: SEO and ranking algorithms influence web search results. MCP goes straight to your documentation content.\n\n## Access your MCP server\n\nDoccupine automatically generates an MCP server for your documentation and hosts it at your documentation URL with the `/api/mcp` path. For example, if your documentation is hosted at `https://example.com`, your MCP server is available at `https://example.com/api/mcp`.\n\nThe MCP server provides both a GET endpoint to discover available tools and a POST endpoint to execute tool calls.\n\n### API Endpoints\n\n#### GET /api/mcp\n\nReturns information about available tools and the current index status.\n\n**Response:**\n```json\n{\n  \"tools\": [\n    {\n      \"name\": \"search_docs\",\n      \"description\": \"Search through the documentation content using semantic search...\",\n      \"inputSchema\": { ... }\n    },\n    ...\n  ],\n  \"index\": {\n    \"ready\": true,\n    \"chunkCount\": 150\n  }\n}\n```\n\n#### POST /api/mcp\n\nExecutes an MCP tool call.\n\n**Request Body:**\n```json\n{\n  \"tool\": \"search_docs\",\n  \"params\": {\n    \"query\": \"how to deploy\",\n    \"limit\": 6\n  }\n}\n```\n\n**Response:**\n```json\n{\n  \"content\": [\n    {\n      \"path\": \"app/deployment/page.tsx\",\n      \"uri\": \"docs://deployment\",\n      \"score\": \"0.892\",\n      \"text\": \"Deploy your Doccupine site as a Next.js application...\"\n    },\n    ...\n  ]\n}\n```\n\n## Available Tools\n\nYour MCP server exposes three tools for interacting with your documentation:\n\n### search_docs\n\nSearch through the documentation content using semantic search. Returns relevant chunks of documentation based on the query using vector embeddings and cosine similarity.\n\n**Parameters:**\n- `query` (required): The search query to find relevant documentation\n- `limit` (optional): Maximum number of results to return (default: 6)\n\n**Example:**\n```json\n{\n  \"tool\": \"search_docs\",\n  \"params\": {\n    \"query\": \"how to configure AI assistant\",\n    \"limit\": 5\n  }\n}\n```\n\n### get_doc\n\nGet the full content of a specific documentation page by its path.\n\n**Parameters:**\n- `path` (required): The file path to the documentation page (e.g., `app/getting-started/page.tsx`)\n\n**Example:**\n```json\n{\n  \"tool\": \"get_doc\",\n  \"params\": {\n    \"path\": \"app/deployment/page.tsx\"\n  }\n}\n```\n\n### list_docs\n\nList all available documentation pages, optionally filtered by directory.\n\n**Parameters:**\n- `directory` (optional): Optional directory to filter results (e.g., `components`)\n\n**Example:**\n```json\n{\n  \"tool\": \"list_docs\",\n  \"params\": {\n    \"directory\": \"configuration\"\n  }\n}\n```\n\n## How it works\n\nDoccupine's MCP server uses semantic search powered by vector embeddings to provide accurate, context-aware search results.\n\n### Indexing Process\n\n1. **Document Discovery**: The server scans your `app/` directory for all `page.tsx`, `page.ts`, `page.jsx`, and `page.js` files.\n2. **Content Extraction**: It extracts content from `const content =` declarations in your page files.\n3. **Chunking**: Large documents are split into chunks of approximately 1500 characters with 200 character overlap for better context preservation.\n4. **Embedding Generation**: Each chunk is converted to a vector embedding using your configured LLM provider's embedding model.\n5. **Index Building**: All embeddings are stored in memory for fast similarity search.\n\n### Search Process\n\n1. **Query Embedding**: The search query is converted to a vector embedding using the same embedding model.\n2. **Similarity Calculation**: Cosine similarity is calculated between the query embedding and all document chunk embeddings.\n3. **Ranking**: Results are sorted by similarity score (highest first).\n4. **Response**: The top N results (based on the limit parameter) are returned with their paths, URIs, scores, and text content.\n\n<Callout type=\"note\">\n  The index is built automatically when the first search is performed. The index is stored in memory and persists for the lifetime of the server process. If you update your documentation, restart the server to rebuild the index.\n</Callout>\n\n## Use your MCP server\n\nYour users must connect your MCP server to their preferred AI tools.\n\n1. Make your MCP server URL publicly available.\n2. Users copy your MCP server URL and add it to their tools.\n3. Users access your documentation through their tools.\n\nThese are some of the ways you can help your users connect to your MCP server:\n\n<Tabs>\n  <TabContent title=\"Claude\">\n    <Steps>\n      <Step title=\"Get your MCP server URL.\">\n        Your MCP server URL is available at `https://your-domain.com/api/mcp`.\n      </Step>\n      <Step title=\"Publish your MCP server URL for your users.\">\n        Create a guide for your users that includes your MCP server URL and the steps to connect it to Claude.\n        1. Navigate to the [Connectors](https://claude.ai/settings/connectors) page in the Claude settings.\n        2. Select **Add custom connector**.\n        3. Add your MCP server name and URL.\n        4. Select **Add**.\n        5. When using Claude, select the attachments button (the plus icon).\n        6. Select your MCP server.\n      </Step>\n    </Steps>\n    See the [Model Context Protocol documentation](https://modelcontextprotocol.io/docs/tutorials/use-remote-mcp-server#connecting-to-a-remote-mcp-server) for more details.\n  </TabContent>\n  <TabContent title=\"Claude Code\">\n    <Steps>\n      <Step title=\"Get your MCP server URL.\">\n        Your MCP server URL is available at `https://your-domain.com/api/mcp`.\n      </Step>\n      <Step title=\"Publish your MCP server URL for your users.\">\n        Create a guide for your users that includes your MCP server URL and the command to connect it to Claude Code.\n        ```bash\n        claude mcp add --transport http <name> <url>\n        ```\n      </Step>\n    </Steps>\n    See the [Claude Code documentation](https://docs.anthropic.com/en/docs/claude-code/mcp#installing-mcp-servers) for more details.\n  </TabContent>\n  <TabContent title=\"Cursor\">\n    <Steps>\n      <Step title=\"Get your MCP server URL.\">\n        Your MCP server URL is available at `https://your-domain.com/api/mcp`.\n      </Step>\n      <Step title=\"Publish your MCP server URL for your users.\">\n        Create a guide for your users that includes your MCP server URL and the steps to connect it to Cursor.\n        1. Use <kbd>Command</kbd> + <kbd>Shift</kbd> + <kbd>P</kbd> (<kbd>Ctrl</kbd> + <kbd>Shift</kbd> + <kbd>P</kbd> on Windows) to open the command palette.\n        2. Search for \"Open MCP settings\".\n        3. Select **Add custom MCP**. This opens the `mcp.json` file.\n        4. In `mcp.json`, configure your server:\n        ```json\n        {\n          \"mcpServers\": {\n            \"<your-mcp-server-name>\": {\n              \"url\": \"<your-mcp-server-url>\"\n            }\n          }\n        }\n        ```\n      </Step>\n    </Steps>\n    See the [Cursor documentation](https://docs.cursor.com/en/context/mcp#installing-mcp-servers) for more details.\n  </TabContent>\n  <TabContent title=\"VS Code\">\n    <Steps>\n      <Step title=\"Get your MCP server URL.\">\n        Your MCP server URL is available at `https://your-domain.com/api/mcp`.\n      </Step>\n      <Step title=\"Publish your MCP server URL for your users.\">\n        Create a guide for your users that includes your MCP server URL and the steps to connect it to VS Code.\n        1. Create a `.vscode/mcp.json` file.\n        2. In `mcp.json`, configure your server:\n        ```json\n        {\n          \"servers\": {\n            \"<your-mcp-server-name>\": {\n              \"type\": \"http\",\n              \"url\": \"<your-mcp-server-url>\"\n            }\n          }\n        }\n        ```\n      </Step>\n    </Steps>\n    See the [VS Code documentation](https://code.visualstudio.com/docs/copilot/chat/mcp-servers) for more details.\n  </TabContent>\n</Tabs>\n\n## Requirements\n\nTo use the MCP server, you need to have the AI Assistant configured. The MCP server uses the same LLM configuration for generating embeddings.\n\n<Callout type=\"warning\">\n  The MCP server requires an LLM provider to be configured for generating embeddings. Make sure you have set up your AI Assistant with a valid API key before using the MCP server.\n</Callout>\n\nSee the [AI Assistant documentation](/ai-assistant) for configuration details.\n\n## Content indexing\n\nYour MCP server searches content extracted from your page files. The server automatically discovers and indexes all `page.tsx`, `page.ts`, `page.jsx`, and `page.js` files in your `app/` directory.\n\n### Content extraction\n\nThe server extracts content from `const content =` declarations in your page files. Make sure your documentation pages export a `content` constant with your markdown or MDX content.\n\n**Example:**\n```tsx\nexport const content = `\n# Getting Started\n\nWelcome to the documentation...\n`;\n```\n\n### Excluded directories\n\nThe following directories are automatically excluded from indexing:\n- `node_modules`\n- `.next`\n- `.git`\n- `api`\n\n## Troubleshooting\n\n### Index not building\n\nIf the index is not building, check:\n- Your LLM provider is configured correctly in your `.env` file\n- You have a valid API key set\n- Your documentation pages export a `content` constant\n\n### No search results\n\nIf searches return no results:\n- Verify that your documentation pages are in the `app/` directory\n- Check that your pages export a `content` constant\n- Ensure the index has been built (check the `index.ready` status via GET `/api/mcp`)\n\n### Slow search performance\n\nThe first search may be slower as it builds the index. Subsequent searches are fast as they use the in-memory index. If performance is consistently slow:\n- Check your embedding API response times\n- Consider reducing the number of documentation pages\n- Verify your server has sufficient memory\n\n## Best practices\n\n* **Keep content up-to-date**: Restart your server after updating documentation to rebuild the index with fresh content.\n* **Use descriptive queries**: More specific queries yield better semantic search results.\n* **Monitor index status**: Use the GET endpoint to check if your index is ready before performing searches.\n* **Optimize content structure**: Well-structured markdown with clear headings improves search relevance.";
+export declare const mcpMdxTemplate = "---\ntitle: \"Model Context Protocol\"\ndescription: \"Connect your Doccupine documentation to AI tools with an MCP server for enhanced AI-powered documentation search.\"\ndate: \"2025-01-24\"\ncategory: \"Configuration\"\ncategoryOrder: 3\norder: 8\n---\n# Model Context Protocol\nConnect your documentation to AI tools with a hosted MCP server.\n\nDoccupine automatically generates a Model Context Protocol (MCP) server from your documentation, making your content accessible to AI applications like Claude, Cursor, VS Code, and other MCP-compatible tools. Your MCP server exposes semantic search capabilities, allowing AI tools to query your documentation directly and provide accuraxte, context-aware answers.\n\n## About MCP servers\n\nThe Model Context Protocol (MCP) is an open protocol that creates standardized connections between AI applications and external services, like documentation. Doccupine generates an MCP server from your documentation, preparing your content for the broader AI ecosystem where any MCP client can connect to your documentation.\n\nYour MCP server exposes search and retrieval tools for AI applications to query your documentation. Your users must connect your MCP server to their preferred AI tools to access your documentation.\n\n### How MCP servers work\n\nWhen an AI tool has your documentation MCP server connected, the AI tool can search your documentation directly instead of making a generic web search in response to a user's prompt. Your MCP server provides access to all indexed content from your documentation site.\n\n* The LLM can proactively search your documentation while generating a response, not just when explicitly asked.\n* The LLM determines when to use the search tool based on the context of the conversation and the relevance of your documentation.\n* Each tool call happens during the generation process, so the LLM searches up-to-date information from your documentation to generate its response.\n\n### MCP compared to web search\n\nAI tools can search the web, but MCP provides distinct advantages for documentation.\n\n* **Direct source access**: Web search depends on what search engines have indexed, which may be stale or incomplete. MCP searches your current indexed documentation directly.\n* **Integrated workflow**: MCP allows the AI to search during response generation rather than performing a separate web search.\n* **Semantic search**: MCP uses vector embeddings for semantic similarity search, providing more relevant results than keyword-based web search.\n* **No search noise**: SEO and ranking algorithms influence web search results. MCP goes straight to your documentation content.\n\n## Access your MCP server\n\nDoccupine automatically generates an MCP server for your documentation and hosts it at your documentation URL with the `/api/mcp` path. For example, if your documentation is hosted at `https://example.com`, your MCP server is available at `https://example.com/api/mcp`.\n\nThe MCP server provides both a GET endpoint to discover available tools and a POST endpoint to execute tool calls.\n\n### API Endpoints\n\n#### GET /api/mcp\n\nReturns information about available tools and the current index status.\n\n**Response:**\n```json\n{\n  \"tools\": [\n    {\n      \"name\": \"search_docs\",\n      \"description\": \"Search through the documentation content using semantic search...\",\n      \"inputSchema\": { ... }\n    },\n    ...\n  ],\n  \"index\": {\n    \"ready\": true,\n    \"chunkCount\": 150\n  }\n}\n```\n\n#### POST /api/mcp\n\nExecutes an MCP tool call.\n\n**Request Body:**\n```json\n{\n  \"tool\": \"search_docs\",\n  \"params\": {\n    \"query\": \"how to deploy\",\n    \"limit\": 6\n  }\n}\n```\n\n**Response:**\n```json\n{\n  \"content\": [\n    {\n      \"path\": \"app/deployment/page.tsx\",\n      \"uri\": \"docs://deployment\",\n      \"score\": \"0.892\",\n      \"text\": \"Deploy your Doccupine site as a Next.js application...\"\n    },\n    ...\n  ]\n}\n```\n\n## Available Tools\n\nYour MCP server exposes three tools for interacting with your documentation:\n\n### search_docs\n\nSearch through the documentation content using semantic search. Returns relevant chunks of documentation based on the query using vector embeddings and cosine similarity.\n\n**Parameters:**\n- `query` (required): The search query to find relevant documentation\n- `limit` (optional): Maximum number of results to return (default: 6)\n\n**Example:**\n```json\n{\n  \"tool\": \"search_docs\",\n  \"params\": {\n    \"query\": \"how to configure AI assistant\",\n    \"limit\": 5\n  }\n}\n```\n\n### get_doc\n\nGet the full content of a specific documentation page by its path.\n\n**Parameters:**\n- `path` (required): The file path to the documentation page (e.g., `app/getting-started/page.tsx`)\n\n**Example:**\n```json\n{\n  \"tool\": \"get_doc\",\n  \"params\": {\n    \"path\": \"app/deployment/page.tsx\"\n  }\n}\n```\n\n### list_docs\n\nList all available documentation pages, optionally filtered by directory.\n\n**Parameters:**\n- `directory` (optional): Optional directory to filter results (e.g., `components`)\n\n**Example:**\n```json\n{\n  \"tool\": \"list_docs\",\n  \"params\": {\n    \"directory\": \"configuration\"\n  }\n}\n```\n\n## How it works\n\nDoccupine's MCP server uses semantic search powered by vector embeddings to provide accurate, context-aware search results.\n\n### Indexing Process\n\n1. **Document Discovery**: The server scans your `app/` directory for all `page.tsx`, `page.ts`, `page.jsx`, and `page.js` files.\n2. **Content Extraction**: It extracts content from `const content =` declarations in your page files.\n3. **Chunking**: Large documents are split into chunks of approximately 1500 characters with 200 character overlap for better context preservation.\n4. **Embedding Generation**: Each chunk is converted to a vector embedding using your configured LLM provider's embedding model.\n5. **Index Building**: All embeddings are stored in memory for fast similarity search.\n\n### Search Process\n\n1. **Query Embedding**: The search query is converted to a vector embedding using the same embedding model.\n2. **Similarity Calculation**: Cosine similarity is calculated between the query embedding and all document chunk embeddings.\n3. **Ranking**: Results are sorted by similarity score (highest first).\n4. **Response**: The top N results (based on the limit parameter) are returned with their paths, URIs, scores, and text content.\n\n<Callout type=\"note\">\n  The index is built automatically when the first search is performed. The index is stored in memory and persists for the lifetime of the server process. If you update your documentation, restart the server to rebuild the index.\n</Callout>\n\n## Use your MCP server\n\nYour users must connect your MCP server to their preferred AI tools.\n\n1. Make your MCP server URL publicly available.\n2. Users copy your MCP server URL and add it to their tools.\n3. Users access your documentation through their tools.\n\nThese are some of the ways you can help your users connect to your MCP server:\n\n<Tabs>\n  <TabContent title=\"Claude\">\n    <Steps>\n      <Step title=\"Get your MCP server URL.\">\n        Your MCP server URL is available at `https://your-domain.com/api/mcp`.\n      </Step>\n      <Step title=\"Publish your MCP server URL for your users.\">\n        Create a guide for your users that includes your MCP server URL and the steps to connect it to Claude.\n        1. Navigate to the [Connectors](https://claude.ai/settings/connectors) page in the Claude settings.\n        2. Select **Add custom connector**.\n        3. Add your MCP server name and URL.\n        4. Select **Add**.\n        5. When using Claude, select the attachments button (the plus icon).\n        6. Select your MCP server.\n      </Step>\n    </Steps>\n    See the [Model Context Protocol documentation](https://modelcontextprotocol.io/docs/tutorials/use-remote-mcp-server#connecting-to-a-remote-mcp-server) for more details.\n  </TabContent>\n  <TabContent title=\"Claude Code\">\n    <Steps>\n      <Step title=\"Get your MCP server URL.\">\n        Your MCP server URL is available at `https://your-domain.com/api/mcp`.\n      </Step>\n      <Step title=\"Publish your MCP server URL for your users.\">\n        Create a guide for your users that includes your MCP server URL and the command to connect it to Claude Code.\n        ```bash\n        claude mcp add --transport http <name> <url>\n        ```\n      </Step>\n    </Steps>\n    See the [Claude Code documentation](https://docs.anthropic.com/en/docs/claude-code/mcp#installing-mcp-servers) for more details.\n  </TabContent>\n  <TabContent title=\"Cursor\">\n    <Steps>\n      <Step title=\"Get your MCP server URL.\">\n        Your MCP server URL is available at `https://your-domain.com/api/mcp`.\n      </Step>\n      <Step title=\"Publish your MCP server URL for your users.\">\n        Create a guide for your users that includes your MCP server URL and the steps to connect it to Cursor.\n        1. Use <kbd>Command</kbd> + <kbd>Shift</kbd> + <kbd>P</kbd> (<kbd>Ctrl</kbd> + <kbd>Shift</kbd> + <kbd>P</kbd> on Windows) to open the command palette.\n        2. Search for \"Open MCP settings\".\n        3. Select **Add custom MCP**. This opens the `mcp.json` file.\n        4. In `mcp.json`, configure your server:\n        ```json\n        {\n          \"mcpServers\": {\n            \"<your-mcp-server-name>\": {\n              \"url\": \"<your-mcp-server-url>\"\n            }\n          }\n        }\n        ```\n      </Step>\n    </Steps>\n    See the [Cursor documentation](https://docs.cursor.com/en/context/mcp#installing-mcp-servers) for more details.\n  </TabContent>\n  <TabContent title=\"VS Code\">\n    <Steps>\n      <Step title=\"Get your MCP server URL.\">\n        Your MCP server URL is available at `https://your-domain.com/api/mcp`.\n      </Step>\n      <Step title=\"Publish your MCP server URL for your users.\">\n        Create a guide for your users that includes your MCP server URL and the steps to connect it to VS Code.\n        1. Create a `.vscode/mcp.json` file.\n        2. In `mcp.json`, configure your server:\n        ```json\n        {\n          \"servers\": {\n            \"<your-mcp-server-name>\": {\n              \"type\": \"http\",\n              \"url\": \"<your-mcp-server-url>\"\n            }\n          }\n        }\n        ```\n      </Step>\n    </Steps>\n    See the [VS Code documentation](https://code.visualstudio.com/docs/copilot/chat/mcp-servers) for more details.\n  </TabContent>\n</Tabs>\n\n## Requirements\n\nTo use the MCP server, you need to have the AI Assistant configured. The MCP server uses the same LLM configuration for generating embeddings.\n\n<Callout type=\"warning\">\n  The MCP server requires an LLM provider to be configured for generating embeddings. Make sure you have set up your AI Assistant with a valid API key before using the MCP server.\n</Callout>\n\nSee the [AI Assistant documentation](/ai-assistant) for configuration details.\n\n## Content indexing\n\nYour MCP server searches content extracted from your page files. The server automatically discovers and indexes all `page.tsx`, `page.ts`, `page.jsx`, and `page.js` files in your `app/` directory.\n\n### Content extraction\n\nThe server extracts content from `const content =` declarations in your page files. Make sure your documentation pages export a `content` constant with your markdown or MDX content.\n\n**Example:**\n```tsx\nexport const content = `\n# Getting Started\n\nWelcome to the documentation...\n`;\n```\n\n### Excluded directories\n\nThe following directories are automatically excluded from indexing:\n- `node_modules`\n- `.next`\n- `.git`\n- `api`\n\n## Troubleshooting\n\n### Index not building\n\nIf the index is not building, check:\n- Your LLM provider is configured correctly in your `.env` file\n- You have a valid API key set\n- Your documentation pages export a `content` constant\n\n### No search results\n\nIf searches return no results:\n- Verify that your documentation pages are in the `app/` directory\n- Check that your pages export a `content` constant\n- Ensure the index has been built (check the `index.ready` status via GET `/api/mcp`)\n\n### Slow search performance\n\nThe first search may be slower as it builds the index. Subsequent searches are fast as they use the in-memory index. If performance is consistently slow:\n- Check your embedding API response times\n- Consider reducing the number of documentation pages\n- Verify your server has sufficient memory\n\n### Cloudflare blocking MCP requests\n\nIf you use Cloudflare as a proxy in front of your documentation site, Cloudflare's bot protection may block server-to-server requests from AI tools like Claude.ai. This can cause MCP connections to fail silently or return errors.\n\nThere are two ways to resolve this:\n\n**Option 1: Disable the Cloudflare proxy (simplest)**\n\nIn your Cloudflare DNS settings, click the orange cloud icon next to your domain record to switch it to \"DNS only\" (grey cloud). This disables Cloudflare's proxy and bot protection for your domain, allowing MCP requests to reach your server directly.\n\n**Option 2: Add a Cloudflare WAF exception (keeps your custom domain proxied)**\n\nIn Cloudflare dashboard:\n\n1. Go to **Security > WAF**.\n2. Click **Create rule**.\n3. Set it up as:\n   - **Rule name**: Allow MCP API\n   - **Field**: URI Path\n   - **Operator**: starts with\n   - **Value**: `/api/mcp`\n   - **Action**: Skip -- then check all remaining custom rules, Rate limiting rules, and Bot Fight Mode / Super Bot Fight Mode.\n4. Deploy the rule and make sure it is ordered first (above other rules).\n\n<Callout type=\"warning\">\n  Also check **Security > Bots** in your Cloudflare dashboard. If \"Bot Fight Mode\" or \"Super Bot Fight Mode\" is enabled, that is very likely what is blocking server-to-server requests from AI tools.\n</Callout>\n\n## Best practices\n\n* **Keep content up-to-date**: Restart your server after updating documentation to rebuild the index with fresh content.\n* **Use descriptive queries**: More specific queries yield better semantic search results.\n* **Monitor index status**: Use the GET endpoint to check if your index is ready before performing searches.\n* **Optimize content structure**: Well-structured markdown with clear headings improves search relevance.";

package/dist/templates/mdx/model-context-protocol.mdx.js

@@ -323,6 +323,34 @@ The first search may be slower as it builds the index. Subsequent searches are f
 - Consider reducing the number of documentation pages
 - Verify your server has sufficient memory
 
+### Cloudflare blocking MCP requests
+
+If you use Cloudflare as a proxy in front of your documentation site, Cloudflare's bot protection may block server-to-server requests from AI tools like Claude.ai. This can cause MCP connections to fail silently or return errors.
+
+There are two ways to resolve this:
+
+**Option 1: Disable the Cloudflare proxy (simplest)**
+
+In your Cloudflare DNS settings, click the orange cloud icon next to your domain record to switch it to "DNS only" (grey cloud). This disables Cloudflare's proxy and bot protection for your domain, allowing MCP requests to reach your server directly.
+
+**Option 2: Add a Cloudflare WAF exception (keeps your custom domain proxied)**
+
+In Cloudflare dashboard:
+
+1. Go to **Security > WAF**.
+2. Click **Create rule**.
+3. Set it up as:
+   - **Rule name**: Allow MCP API
+   - **Field**: URI Path
+   - **Operator**: starts with
+   - **Value**: \`/api/mcp\`
+   - **Action**: Skip -- then check all remaining custom rules, Rate limiting rules, and Bot Fight Mode / Super Bot Fight Mode.
+4. Deploy the rule and make sure it is ordered first (above other rules).
+
+<Callout type="warning">
+  Also check **Security > Bots** in your Cloudflare dashboard. If "Bot Fight Mode" or "Super Bot Fight Mode" is enabled, that is very likely what is blocking server-to-server requests from AI tools.
+</Callout>
+
 ## Best practices
 
 * **Keep content up-to-date**: Restart your server after updating documentation to rebuild the index with fresh content.

package/dist/templates/package.js

@@ -13,7 +13,7 @@ export const packageJsonTemplate = JSON.stringify({
         "@langchain/google-genai": "^2.1.18",
         "@langchain/openai": "^1.2.7",
         "@modelcontextprotocol/sdk": "^1.26.0",
-        langchain: "^1.2.23",
+        langchain: "^1.2.24",
         next: "16.1.6",
         react: "19.2.4",
         "react-dom": "19.2.4",

package/dist/templates/services/llm/config.d.ts

@@ -1 +1 @@
-export declare const llmConfigTemplate = "import type {\n  LLMConfig,\n  LLMProvider,\n  ProviderDefaults,\n} from \"@/services/llm/types\";\nconst PROVIDER_DEFAULTS: ProviderDefaults = {\n  openai: {\n    chat: \"gpt-4.1-nano\",\n    embedding: \"text-embedding-3-small\",\n  },\n  anthropic: {\n    chat: \"claude-sonnet-4-5-20250929\",\n    embedding: \"text-embedding-3-small\", // Fallback to OpenAI\n  },\n  google: {\n    chat: \"gemini-2.5-flash-lite\",\n    embedding: \"text-embedding-004\",\n  },\n};\nfunction validateAPIKeys(provider: LLMProvider): void {\n  const requiredKeys: Record<LLMProvider, string> = {\n    openai: \"OPENAI_API_KEY\",\n    anthropic: \"ANTHROPIC_API_KEY\",\n    google: \"GOOGLE_API_KEY\",\n  };\n  const keyName = requiredKeys[provider];\n  const keyValue = process.env[keyName];\n  if (!keyValue) {\n    throw new Error(\n      `Missing API key for ${provider}. Please set ${keyName} in your environment variables.`,\n    );\n  }\n  if (provider === \"anthropic\" && !process.env.OPENAI_API_KEY) {\n    console.warn(\n      \"Anthropic provider requires OPENAI_API_KEY for embeddings. Please set OPENAI_API_KEY in your environment variables.\",\n    );\n  }\n}\nexport function getLLMConfig(): LLMConfig {\n  const provider = (process.env.LLM_PROVIDER || \"openai\") as LLMProvider;\n  if (![\"openai\", \"anthropic\", \"google\"].includes(provider)) {\n    throw new Error(\n      `Invalid LLM_PROVIDER: ${provider}. Must be one of: openai, anthropic, google`,\n    );\n  }\n  validateAPIKeys(provider);\n  const defaults = PROVIDER_DEFAULTS[provider];\n  return {\n    provider,\n    chatModel: process.env.LLM_CHAT_MODEL || defaults.chat,\n    embeddingModel: process.env.LLM_EMBEDDING_MODEL || defaults.embedding,\n    temperature: parseFloat(process.env.LLM_TEMPERATURE || \"1\"),\n  };\n}\nexport function getLLMDisplayName(config?: LLMConfig): string {\n  const c = config || getLLMConfig();\n  const providerNames: Record<LLMProvider, string> = {\n    openai: \"OpenAI\",\n    anthropic: \"Anthropic\",\n    google: \"Google\",\n  };\n  return `${providerNames[c.provider]} (${c.chatModel})`;\n}\nexport function isLLMConfigured(): boolean {\n  try {\n    getLLMConfig();\n    return true;\n  } catch {\n    return false;\n  }\n}\n";
+export declare const llmConfigTemplate = "import type {\n  LLMConfig,\n  LLMProvider,\n  ProviderDefaults,\n} from \"@/services/llm/types\";\nconst PROVIDER_DEFAULTS: ProviderDefaults = {\n  openai: {\n    chat: \"gpt-4.1-nano\",\n    embedding: \"text-embedding-3-small\",\n  },\n  anthropic: {\n    chat: \"claude-sonnet-4-5-20250929\",\n    embedding: \"text-embedding-3-small\", // Fallback to OpenAI\n  },\n  google: {\n    chat: \"gemini-2.5-flash-lite\",\n    embedding: \"text-embedding-004\",\n  },\n};\nfunction validateAPIKeys(provider: LLMProvider): void {\n  const requiredKeys: Record<LLMProvider, string> = {\n    openai: \"OPENAI_API_KEY\",\n    anthropic: \"ANTHROPIC_API_KEY\",\n    google: \"GOOGLE_API_KEY\",\n  };\n  const keyName = requiredKeys[provider];\n  const keyValue = process.env[keyName];\n  if (!keyValue) {\n    throw new Error(\n      `Missing API key for ${provider}. Please set ${keyName} in your environment variables.`,\n    );\n  }\n  if (provider === \"anthropic\" && !process.env.OPENAI_API_KEY) {\n    console.warn(\n      \"Anthropic provider requires OPENAI_API_KEY for embeddings. Please set OPENAI_API_KEY in your environment variables.\",\n    );\n  }\n}\nexport function getLLMConfig(): LLMConfig {\n  const provider = (process.env.LLM_PROVIDER || \"openai\") as LLMProvider;\n  if (![\"openai\", \"anthropic\", \"google\"].includes(provider)) {\n    throw new Error(\n      `Invalid LLM_PROVIDER: ${provider}. Must be one of: openai, anthropic, google`,\n    );\n  }\n  validateAPIKeys(provider);\n  const defaults = PROVIDER_DEFAULTS[provider];\n  return {\n    provider,\n    chatModel: process.env.LLM_CHAT_MODEL || defaults.chat,\n    embeddingModel: process.env.LLM_EMBEDDING_MODEL || defaults.embedding,\n    temperature: parseFloat(process.env.LLM_TEMPERATURE || \"0\"),\n  };\n}\nexport function getLLMDisplayName(config?: LLMConfig): string {\n  const c = config || getLLMConfig();\n  const providerNames: Record<LLMProvider, string> = {\n    openai: \"OpenAI\",\n    anthropic: \"Anthropic\",\n    google: \"Google\",\n  };\n  return `${providerNames[c.provider]} (${c.chatModel})`;\n}\nexport function isLLMConfigured(): boolean {\n  try {\n    getLLMConfig();\n    return true;\n  } catch {\n    return false;\n  }\n}\n";

package/dist/templates/services/llm/config.js

@@ -49,7 +49,7 @@ export function getLLMConfig(): LLMConfig {
     provider,
     chatModel: process.env.LLM_CHAT_MODEL || defaults.chat,
     embeddingModel: process.env.LLM_EMBEDDING_MODEL || defaults.embedding,
-    temperature: parseFloat(process.env.LLM_TEMPERATURE || "1"),
+    temperature: parseFloat(process.env.LLM_TEMPERATURE || "0"),
   };
 }
 export function getLLMDisplayName(config?: LLMConfig): string {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "doccupine",
-  "version": "0.0.45",
+  "version": "0.0.47",
   "description": "Document management system that allows you to store, organize, and share your documentation with ease. AI-ready.",
   "main": "dist/index.js",
   "bin": {