[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-72003":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":20,"compositeScore":21,"rankGlobal":10,"rankLanguage":10,"license":22,"archived":23,"fork":23,"defaultBranch":24,"hasWiki":25,"hasPages":25,"topics":26,"createdAt":10,"pushedAt":10,"updatedAt":44,"readmeContent":45,"aiSummary":46,"trendingCount":16,"starSnapshotCount":16,"syncStatus":47,"lastSyncTime":48,"discoverSource":49},72003,"deep-searcher","zilliztech\u002Fdeep-searcher","zilliztech","Open Source Deep Research Alternative to Reason and Search on Private Data. Written in Python.","https:\u002F\u002Fzilliztech.github.io\u002Fdeep-searcher\u002F",null,"Python",7859,759,59,42,0,3,11,39,9,39.64,"Apache License 2.0",false,"master",true,[27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43],"agent","agentic-rag","claude","deep-research","deepseek","deepseek-r1","grok","grok3","llama4","llm","milvus","openai","qwen3","rag","reasoning-models","vector-database","zilliz","2026-06-12 02:02:57","![DeepSearcher](.\u002Fassets\u002Fpic\u002Flogo.png)\n\n\u003Cdiv align=\"center\">\n  \n[![License](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-Apache%202.0-blue.svg)](https:\u002F\u002Fopensource.org\u002Flicenses\u002FApache-2.0)\n[![DeepWiki](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDeepWiki-AI%20Docs-orange.svg)](https:\u002F\u002Fdeepwiki.com\u002Fzilliztech\u002Fdeep-searcher)\n[![Twitter](https:\u002F\u002Fimg.shields.io\u002Ftwitter\u002Furl\u002Fhttps\u002Ftwitter.com\u002Fzilliz_universe.svg?style=social&label=Follow%20%40Zilliz)](https:\u002F\u002Ftwitter.com\u002Fzilliz_universe)\n\u003Ca href=\"https:\u002F\u002Fdiscord.gg\u002FmKc3R95yE5\">\u003Cimg height=\"20\" src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDiscord-%235865F2.svg?style=for-the-badge&logo=discord&logoColor=white\" alt=\"discord\"\u002F>\u003C\u002Fa>\n\n\u003C\u002Fdiv>\n\n---\n\nDeepSearcher combines cutting-edge LLMs (OpenAI o3, Qwen3, DeepSeek, Grok 4, Claude 4 Sonnet, Llama 4, QwQ, etc.) and Vector Databases (Milvus, Zilliz Cloud etc.) to perform search, evaluation, and reasoning based on private data, providing highly accurate answer and comprehensive report. This project is suitable for enterprise knowledge management, intelligent Q&A systems, and information retrieval scenarios.\n\n![Architecture](.\u002Fassets\u002Fpic\u002Fdeep-searcher-arch.png)\n\n## 🚀 Features\n\n- **Private Data Search**: Maximizes the utilization of enterprise internal data while ensuring data security. When necessary, it can integrate online content for more accurate answers.\n- **Vector Database Management**: Supports Milvus and other vector databases, allowing data partitioning for efficient retrieval.\n- **Flexible Embedding Options**: Compatible with multiple embedding models for optimal selection.\n- **Multiple LLM Support**: Supports DeepSeek, OpenAI, and other large models for intelligent Q&A and content generation.\n- **Document Loader**: Supports local file loading, with web crawling capabilities under development.\n\n---\n\n## 🎉 Demo\n![demo](.\u002Fassets\u002Fpic\u002Fdemo.gif)\n\n\n## 📖 Quick Start\n\n### Installation\nInstall DeepSearcher using one of the following methods:\n\n#### Option 1: Using pip\nCreate and activate a virtual environment(Python 3.10 version is recommended).\n```bash\npython -m venv .venv\nsource .venv\u002Fbin\u002Factivate\n```\nInstall DeepSearcher\n```bash\npip install deepsearcher\n```\n\nFor optional dependencies, e.g., ollama:\n```bash\npip install \"deepsearcher[ollama]\"\n```\n\n#### Option 2: Install in Development Mode\nWe recommend using [uv](https:\u002F\u002Fgithub.com\u002Fastral-sh\u002Fuv) for faster and more reliable installation. Follow the [offical installation instructions](https:\u002F\u002Fdocs.astral.sh\u002Fuv\u002Fgetting-started\u002Finstallation\u002F) to install it.\n\nClone the repository and navigate to the project directory:\n```shell\ngit clone https:\u002F\u002Fgithub.com\u002Fzilliztech\u002Fdeep-searcher.git && cd deep-searcher\n```\nSynchronize and install dependencies:\n```shell\nuv sync\nsource .venv\u002Fbin\u002Factivate\n```\n\nFor more detailed development setup and optional dependency installation options, see [CONTRIBUTING.md](CONTRIBUTING.md#development-environment-setup-with-uv).\n\n### Quick start demo\n\nTo run this quick start demo, please prepare your `OPENAI_API_KEY` in your environment variables. If you change the LLM in the configuration, make sure to prepare the corresponding API key.\n\n```python\nfrom deepsearcher.configuration import Configuration, init_config\nfrom deepsearcher.online_query import query\n\nconfig = Configuration()\n\n# Customize your config here,\n# more configuration see the Configuration Details section below.\nconfig.set_provider_config(\"llm\", \"OpenAI\", {\"model\": \"o1-mini\"})\nconfig.set_provider_config(\"embedding\", \"OpenAIEmbedding\", {\"model\": \"text-embedding-ada-002\"})\ninit_config(config = config)\n\n# Load your local data\nfrom deepsearcher.offline_loading import load_from_local_files\nload_from_local_files(paths_or_directory=your_local_path)\n\n# (Optional) Load from web crawling (`FIRECRAWL_API_KEY` env variable required)\nfrom deepsearcher.offline_loading import load_from_website\nload_from_website(urls=website_url)\n\n# Query\nresult = query(\"Write a report about xxx.\") # Your question here\n```\n### Configuration Details:\n#### LLM Configuration\n\n\u003Cpre>\u003Ccode>config.set_provider_config(\"llm\", \"(LLMName)\", \"(Arguments dict)\")\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>The \"LLMName\" can be one of the following: [\"DeepSeek\", \"OpenAI\", \"XAI\", \"SiliconFlow\", \"Aliyun\", \"PPIO\", \"TogetherAI\", \"Gemini\", \"Ollama\", \"Novita\", \"Jiekou.AI\"]\u003C\u002Fp>\n\u003Cp> The \"Arguments dict\" is a dictionary that contains the necessary arguments for the LLM class.\u003C\u002Fp>\n\n\u003Cdetails>\n  \u003Csummary>Example (OpenAI)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your OPENAI API KEY as an env variable \u003Ccode>OPENAI_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"llm\", \"OpenAI\", {\"model\": \"o1-mini\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> More details about OpenAI models: https:\u002F\u002Fplatform.openai.com\u002Fdocs\u002Fmodels \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (Qwen3 from Aliyun Bailian)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your Bailian API KEY as an env variable \u003Ccode>DASHSCOPE_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"llm\", \"Aliyun\", {\"model\": \"qwen-plus-latest\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> More details about Aliyun Bailian models: https:\u002F\u002Fbailian.console.aliyun.com \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\n\u003Cdetails>\n  \u003Csummary>Example (Qwen3 from OpenRouter)\u003C\u002Fsummary>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"llm\", \"OpenAI\", {\"model\": \"qwen\u002Fqwen3-235b-a22b:free\", \"base_url\": \"https:\u002F\u002Fopenrouter.ai\u002Fapi\u002Fv1\", \"api_key\": \"OPENROUTER_API_KEY\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> More details about OpenRouter models: https:\u002F\u002Fopenrouter.ai\u002Fqwen\u002Fqwen3-235b-a22b:free \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\n\u003Cdetails>\n  \u003Csummary>Example (DeepSeek from official)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your DEEPSEEK API KEY as an env variable \u003Ccode>DEEPSEEK_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"llm\", \"DeepSeek\", {\"model\": \"deepseek-reasoner\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> More details about DeepSeek: https:\u002F\u002Fapi-docs.deepseek.com\u002F \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (DeepSeek from SiliconFlow)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your SILICONFLOW API KEY as an env variable \u003Ccode>SILICONFLOW_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"llm\", \"SiliconFlow\", {\"model\": \"deepseek-ai\u002FDeepSeek-R1\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> More details about SiliconFlow: https:\u002F\u002Fdocs.siliconflow.cn\u002Fquickstart \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (DeepSeek from TogetherAI)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your TOGETHER API KEY as an env variable \u003Ccode>TOGETHER_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    For deepseek R1:\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"llm\", \"TogetherAI\", {\"model\": \"deepseek-ai\u002FDeepSeek-R1\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    For Llama 4:\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"llm\", \"TogetherAI\", {\"model\": \"meta-llama\u002FLlama-4-Scout-17B-16E-Instruct\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> You need to install together before running, execute: \u003Ccode>pip install together\u003C\u002Fcode>. More details about TogetherAI: https:\u002F\u002Fwww.together.ai\u002F \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (XAI Grok)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your XAI API KEY as an env variable \u003Ccode>XAI_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"llm\", \"XAI\", {\"model\": \"grok-4-0709\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> More details about XAI Grok: https:\u002F\u002Fdocs.x.ai\u002Fdocs\u002Foverview#featured-models \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (Claude)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your ANTHROPIC API KEY as an env variable \u003Ccode>ANTHROPIC_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"llm\", \"Anthropic\", {\"model\": \"claude-sonnet-4-0\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> More details about Anthropic Claude: https:\u002F\u002Fdocs.anthropic.com\u002Fen\u002Fhome \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (Google Gemini)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your GEMINI API KEY as an env variable \u003Ccode>GEMINI_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config('llm', 'Gemini', { 'model': 'gemini-2.0-flash' })\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> You need to install gemini before running, execute: \u003Ccode>pip install google-genai\u003C\u002Fcode>. More details about Gemini: https:\u002F\u002Fai.google.dev\u002Fgemini-api\u002Fdocs \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (DeepSeek from PPIO)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your PPIO API KEY as an env variable \u003Ccode>PPIO_API_KEY\u003C\u002Fcode>. You can create an API Key \u003Ca href=\"https:\u002F\u002Fppinfra.com\u002Fsettings\u002Fkey-management?utm_source=github_deep-searcher\">here\u003C\u002Fa>. \u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"llm\", \"PPIO\", {\"model\": \"deepseek\u002Fdeepseek-r1-turbo\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> More details about PPIO: https:\u002F\u002Fppinfra.com\u002Fdocs\u002Fget-started\u002Fquickstart.html?utm_source=github_deep-searcher \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (Claude Sonnet 4.5 from Jiekou.AI)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your Jiekou.AI API KEY as an env variable \u003Ccode>JIEKOU_API_KEY\u003C\u002Fcode>. You can create an API Key \u003Ca href=\"https:\u002F\u002Fjiekou.ai\u002Fsettings\u002Fkey-management?utm_source=github_deep-searcher\">here\u003C\u002Fa>. \u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"llm\", \"JiekouAI\", {\"model\": \"claude-sonnet-4-5-20250929\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> More details about Jiekou.AI: https:\u002F\u002Fdocs.jiekou.ai\u002Fdocs\u002Fsupport\u002Fquickstart?utm_source=github_deep-searcher \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (Ollama)\u003C\u002Fsummary>\n  \u003Cp> Follow \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fjmorganca\u002Follama\">these instructions\u003C\u002Fa> to set up and run a local Ollama instance:\u003C\u002Fp>\n  \u003Cp> \u003Ca href=\"https:\u002F\u002Follama.ai\u002Fdownload\">Download\u003C\u002Fa> and install Ollama onto the available supported platforms (including Windows Subsystem for Linux).\u003C\u002Fp>\n  \u003Cp> View a list of available models via the \u003Ca href=\"https:\u002F\u002Follama.ai\u002Flibrary\">model library\u003C\u002Fa>.\u003C\u002Fp>\n  \u003Cp> Fetch available LLM models via \u003Ccode>ollama pull <name-of-model>\u003C\u002Fcode>\u003C\u002Fp>\n  \u003Cp> Example: \u003Ccode>ollama pull qwen3\u003C\u002Fcode>\u003C\u002Fp>\n  \u003Cp> To chat directly with a model from the command line, use \u003Ccode>ollama run <name-of-model>\u003C\u002Fcode>.\u003C\u002Fp>\n  \u003Cp> By default, Ollama has a REST API for running and managing models on \u003Ca href=\"http:\u002F\u002Flocalhost:11434\">http:\u002F\u002Flocalhost:11434\u003C\u002Fa>.\u003C\u002Fp>\n  \u003Cpre>\u003Ccode>config.set_provider_config(\"llm\", \"Ollama\", {\"model\": \"qwen3\"})\u003C\u002Fcode>\u003C\u002Fpre>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (Volcengine)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your Volcengine API KEY as an env variable \u003Ccode>VOLCENGINE_API_KEY\u003C\u002Fcode>. You can create an API Key \u003Ca href=\"https:\u002F\u002Fconsole.volcengine.com\u002Fark\u002Fregion:ark+cn-beijing\u002FapiKey\">here\u003C\u002Fa>. \u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"llm\", \"Volcengine\", {\"model\": \"deepseek-r1-250120\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> More details about Volcengine: https:\u002F\u002Fwww.volcengine.com\u002Fdocs\u002F82379\u002F1099455?utm_source=github_deep-searcher \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (GLM)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your GLM API KEY as an env variable \u003Ccode>GLM_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"llm\", \"GLM\", {\"model\": \"glm-4-plus\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> You need to install zhipuai before running, execute: \u003Ccode>pip install zhipuai\u003C\u002Fcode>. More details about GLM: https:\u002F\u002Fbigmodel.cn\u002Fdev\u002Fwelcome \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (Amazon Bedrock)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your Amazon Bedrock API KEY as an env variable \u003Ccode>AWS_ACCESS_KEY_ID\u003C\u002Fcode> and \u003Ccode>AWS_SECRET_ACCESS_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"llm\", \"Bedrock\", {\"model\": \"us.deepseek.r1-v1:0\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> You need to install boto3 before running, execute: \u003Ccode>pip install boto3\u003C\u002Fcode>. More details about Amazon Bedrock: https:\u002F\u002Fdocs.aws.amazon.com\u002Fbedrock\u002F \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (IBM watsonx.ai)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your watsonx.ai credentials as env variables \u003Ccode>WATSONX_APIKEY\u003C\u002Fcode>, \u003Ccode>WATSONX_URL\u003C\u002Fcode>, and \u003Ccode>WATSONX_PROJECT_ID\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"llm\", \"watsonx\", {\"model\": \"us.deepseek.r1-v1:0\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> You need to install ibm-watsonx-ai before running, execute: \u003Ccode>pip install ibm-watsonx-ai\u003C\u002Fcode>. More details about IBM watsonx.ai: https:\u002F\u002Fwww.ibm.com\u002Fproducts\u002Fwatsonx-ai\u002Ffoundation-models \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\n#### Embedding Model Configuration\n\u003Cpre>\u003Ccode>config.set_provider_config(\"embedding\", \"(EmbeddingModelName)\", \"(Arguments dict)\")\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>The \"EmbeddingModelName\" can be one of the following: [\"MilvusEmbedding\", \"OpenAIEmbedding\", \"VoyageEmbedding\", \"SiliconflowEmbedding\", \"PPIOEmbedding\", \"NovitaEmbedding\", \"JiekouAIEmbedding\"]\u003C\u002Fp>\n\u003Cp> The \"Arguments dict\" is a dictionary that contains the necessary arguments for the embedding model class.\u003C\u002Fp>\n\n\u003Cdetails>\n  \u003Csummary>Example (OpenAI embedding)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your OpenAI API KEY as an env variable \u003Ccode>OPENAI_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"embedding\", \"OpenAIEmbedding\", {\"model\": \"text-embedding-3-small\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> More details about OpenAI models: https:\u002F\u002Fplatform.openai.com\u002Fdocs\u002Fguides\u002Fembeddings\u002Fuse-cases \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (OpenAI embedding Azure)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your OpenAI API KEY as an env variable \u003Ccode>OPENAI_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"embedding\", \"OpenAIEmbedding\", {\n    \"model\": \"text-embedding-ada-002\",\n    \"azure_endpoint\": \"https:\u002F\u002F\u003Cyouraifoundry>.openai.azure.com\u002F\",\n    \"api_version\": \"2023-05-15\"\n})\u003C\u002Fcode>\u003C\u002Fpre>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (Pymilvus built-in embedding model)\u003C\u002Fsummary>\n    \u003Cp> Use the built-in embedding model in Pymilvus, you can set the model name as \u003Ccode>\"default\"\u003C\u002Fcode>, \u003Ccode>\"BAAI\u002Fbge-base-en-v1.5\"\u003C\u002Fcode>, \u003Ccode>\"BAAI\u002Fbge-large-en-v1.5\"\u003C\u002Fcode>, \u003Ccode>\"jina-embeddings-v3\"\u003C\u002Fcode>, etc. \u003Cbr\u002F>\n    See [milvus_embedding.py](deepsearcher\u002Fembedding\u002Fmilvus_embedding.py) for more details.  \u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"embedding\", \"MilvusEmbedding\", {\"model\": \"BAAI\u002Fbge-base-en-v1.5\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"embedding\", \"MilvusEmbedding\", {\"model\": \"jina-embeddings-v3\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> For Jina's embedding model, you need\u003Ccode>JINAAI_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cp> You need to install pymilvus model before running, execute: \u003Ccode>pip install pymilvus.model\u003C\u002Fcode>. More details about Pymilvus: https:\u002F\u002Fmilvus.io\u002Fdocs\u002Fembeddings.md \u003C\u002Fp>\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (VoyageAI embedding)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your VOYAGE API KEY as an env variable \u003Ccode>VOYAGE_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"embedding\", \"VoyageEmbedding\", {\"model\": \"voyage-3\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> You need to install voyageai before running, execute: \u003Ccode>pip install voyageai\u003C\u002Fcode>. More details about VoyageAI: https:\u002F\u002Fdocs.voyageai.com\u002Fembeddings\u002F \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (Amazon Bedrock embedding)\u003C\u002Fsummary>\n  \u003Cpre>\u003Ccode>config.set_provider_config(\"embedding\", \"BedrockEmbedding\", {\"model\": \"amazon.titan-embed-text-v2:0\"})\u003C\u002Fcode>\u003C\u002Fpre>\n  \u003Cp> You need to install boto3 before running, execute: \u003Ccode>pip install boto3\u003C\u002Fcode>. More details about Amazon Bedrock: https:\u002F\u002Fdocs.aws.amazon.com\u002Fbedrock\u002F \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (Novita AI embedding)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your Novita AI API KEY as an env variable \u003Ccode>NOVITA_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"embedding\", \"NovitaEmbedding\", {\"model\": \"baai\u002Fbge-m3\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> More details about Novita AI: https:\u002F\u002Fnovita.ai\u002Fdocs\u002Fapi-reference\u002Fmodel-apis-llm-create-embeddings?utm_source=github_deep-searcher&utm_medium=github_readme&utm_campaign=link \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (Siliconflow embedding)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your Siliconflow API KEY as an env variable \u003Ccode>SILICONFLOW_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"embedding\", \"SiliconflowEmbedding\", {\"model\": \"BAAI\u002Fbge-m3\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> More details about Siliconflow: https:\u002F\u002Fdocs.siliconflow.cn\u002Fen\u002Fapi-reference\u002Fembeddings\u002Fcreate-embeddings \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (Volcengine embedding)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your Volcengine API KEY as an env variable \u003Ccode>VOLCENGINE_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"embedding\", \"VolcengineEmbedding\", {\"model\": \"doubao-embedding-text-240515\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> More details about Volcengine: https:\u002F\u002Fwww.volcengine.com\u002Fdocs\u002F82379\u002F1302003 \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (GLM embedding)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your GLM API KEY as an env variable \u003Ccode>GLM_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"embedding\", \"GLMEmbedding\", {\"model\": \"embedding-3\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> You need to install zhipuai before running, execute: \u003Ccode>pip install zhipuai\u003C\u002Fcode>. More details about GLM: https:\u002F\u002Fbigmodel.cn\u002Fdev\u002Fwelcome \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (Google Gemini embedding)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your Gemini API KEY as an env variable \u003Ccode>GEMINI_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"embedding\", \"GeminiEmbedding\", {\"model\": \"text-embedding-004\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> You need to install gemini before running, execute: \u003Ccode>pip install google-genai\u003C\u002Fcode>. More details about Gemini: https:\u002F\u002Fai.google.dev\u002Fgemini-api\u002Fdocs \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (Ollama embedding)\u003C\u002Fsummary>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"embedding\", \"OllamaEmbedding\", {\"model\": \"bge-m3\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> You need to install ollama before running, execute: \u003Ccode>pip install ollama\u003C\u002Fcode>. More details about Ollama Python SDK: https:\u002F\u002Fgithub.com\u002Follama\u002Follama-python \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (PPIO embedding)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your PPIO API KEY as an env variable \u003Ccode>PPIO_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"embedding\", \"PPIOEmbedding\", {\"model\": \"baai\u002Fbge-m3\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> More details about PPIO: https:\u002F\u002Fppinfra.com\u002Fdocs\u002Fget-started\u002Fquickstart.html?utm_source=github_deep-searcher \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (Jiekou.AI embedding)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your Jiekou.AI API KEY as an env variable \u003Ccode>JIEKOU_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"embedding\", \"JiekouAIEmbedding\", {\"model\": \"qwen\u002Fqwen3-embedding-8b\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> More details about Jiekou.AI: https:\u002F\u002Fdocs.jiekou.ai\u002Fdocs\u002Fsupport\u002Fquickstart?utm_source=github_deep-searcher \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (FastEmbed embedding)\u003C\u002Fsummary>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"embedding\", \"FastEmbedEmbedding\", {\"model\": \"intfloat\u002Fmultilingual-e5-large\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> You need to install fastembed before running, execute: \u003Ccode>pip install fastembed\u003C\u002Fcode>. More details about fastembed: https:\u002F\u002Fgithub.com\u002Fqdrant\u002Ffastembed \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\n\u003Cdetails>\n  \u003Csummary>Example (IBM watsonx.ai embedding)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your WatsonX credentials as env variables \u003Ccode>WATSONX_APIKEY\u003C\u002Fcode>, \u003Ccode>WATSONX_URL\u003C\u002Fcode>, and \u003Ccode>WATSONX_PROJECT_ID\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"embedding\", \"WatsonXEmbedding\", {\"model\": \"ibm\u002Fslate-125m-english-rtrvr-v2\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"embedding\", \"WatsonXEmbedding\", {\"model\": \"sentence-transformers\u002Fall-minilm-l6-v2\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> You need to install ibm-watsonx-ai before running, execute: \u003Ccode>pip install ibm-watsonx-ai\u003C\u002Fcode>. More details about IBM watsonx.ai: https:\u002F\u002Fwww.ibm.com\u002Fproducts\u002Fwatsonx-ai\u002Ffoundation-models \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n#### Vector Database Configuration\n\u003Cpre>\u003Ccode>config.set_provider_config(\"vector_db\", \"(VectorDBName)\", \"(Arguments dict)\")\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>The \"VectorDBName\" can be one of the following: [\"Milvus\"] (Under development)\u003C\u002Fp>\n\u003Cp> The \"Arguments dict\" is a dictionary that contains the necessary arguments for the Vector Database class.\u003C\u002Fp>\n\n\u003Cdetails>\n  \u003Csummary>Example (Milvus)\u003C\u002Fsummary>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"vector_db\", \"Milvus\", {\"uri\": \".\u002Fmilvus.db\", \"token\": \"\"})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> More details about Milvus Config:\u003C\u002Fp>\n    \u003Cul>\n        \u003Cli>\n            Setting the \u003Ccode>uri\u003C\u002Fcode> as a local file, e.g. \u003Ccode>.\u002Fmilvus.db\u003C\u002Fcode>, is the most convenient method, as it automatically utilizes \u003Ca href=\"https:\u002F\u002Fmilvus.io\u002Fdocs\u002Fmilvus_lite.md\" target=\"_blank\">Milvus Lite\u003C\u002Fa> to store all data in this file.\n        \u003C\u002Fli>\n    \u003C\u002Ful>\n    \u003Cul>\n      \u003Cli>\n          If you have a large-scale dataset, you can set up a more performant Milvus server using \n          \u003Ca href=\"https:\u002F\u002Fmilvus.io\u002Fdocs\u002Fquickstart.md\" target=\"_blank\">Docker or Kubernetes\u003C\u002Fa>. \n          In this setup, use the server URI, e.g., \u003Ccode>http:\u002F\u002Flocalhost:19530\u003C\u002Fcode>, as your \u003Ccode>uri\u003C\u002Fcode>. \n          You can also use any other connection parameters supported by Milvus such as \u003Ccode>host\u003C\u002Fcode>, \u003Ccode>user\u003C\u002Fcode>, \u003Ccode>password\u003C\u002Fcode>, or \u003Ccode>secure\u003C\u002Fcode>.\n        \u003C\u002Fli>\n    \u003C\u002Ful>\n    \u003Cul>\n        \u003Cli>\n            If you want to use \u003Ca href=\"https:\u002F\u002Fzilliz.com\u002Fcloud\" target=\"_blank\">Zilliz Cloud\u003C\u002Fa>, \n            the fully managed cloud service for Milvus, adjust the \u003Ccode>uri\u003C\u002Fcode> and \u003Ccode>token\u003C\u002Fcode> \n            according to the \u003Ca href=\"https:\u002F\u002Fdocs.zilliz.com\u002Fdocs\u002Fon-zilliz-cloud-console#free-cluster-details\" \n            target=\"_blank\">Public Endpoint and API Key\u003C\u002Fa> in Zilliz Cloud.\n        \u003C\u002Fli>\n    \u003C\u002Ful>\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (AZURE AI Search)\u003C\u002Fsummary>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"vector_db\", \"AzureSearch\", {\n    \"endpoint\": \"https:\u002F\u002F\u003Cyourazureaisearch>.search.windows.net\",\n    \"index_name\": \"\u003Cyourindex>\",\n    \"api_key\": \"\u003Cyourkey>\",\n    \"vector_field\": \"\"\n})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> More details about Milvus Config:\u003C\u002Fp>\n\n\u003C\u002Fdetails>\n\n#### File Loader Configuration\n\u003Cpre>\u003Ccode>config.set_provider_config(\"file_loader\", \"(FileLoaderName)\", \"(Arguments dict)\")\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>The \"FileLoaderName\" can be one of the following: [\"PDFLoader\", \"TextLoader\", \"UnstructuredLoader\"]\u003C\u002Fp>\n\u003Cp> The \"Arguments dict\" is a dictionary that contains the necessary arguments for the File Loader class.\u003C\u002Fp>\n\n\u003Cdetails>\n  \u003Csummary>Example (Unstructured)\u003C\u002Fsummary>\n    \u003Cp>You can use Unstructured in two ways:\u003C\u002Fp>\n    \u003Cul>\n      \u003Cli>With API: Set environment variables \u003Ccode>UNSTRUCTURED_API_KEY\u003C\u002Fcode> and \u003Ccode>UNSTRUCTURED_API_URL\u003C\u002Fcode>\u003C\u002Fli>\n      \u003Cli>Without API: Use the local processing mode by simply not setting these environment variables\u003C\u002Fli>\n    \u003C\u002Ful>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"file_loader\", \"UnstructuredLoader\", {})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cul>\n      \u003Cli>Currently supported file types: [\"pdf\"] (Under development)\u003C\u002Fli>\n      \u003Cli>Installation requirements:\n        \u003Cul>\n          \u003Cli>Install ingest pipeline: \u003Ccode>pip install unstructured-ingest\u003C\u002Fcode>\u003C\u002Fli>\n          \u003Cli>For all document formats: \u003Ccode>pip install \"unstructured[all-docs]\"\u003C\u002Fcode>\u003C\u002Fli>\n          \u003Cli>For specific formats (e.g., PDF only): \u003Ccode>pip install \"unstructured[pdf]\"\u003C\u002Fcode>\u003C\u002Fli>\n        \u003C\u002Ful>\n      \u003C\u002Fli>\n      \u003Cli>More information:\n        \u003Cul>\n          \u003Cli>Unstructured documentation: \u003Ca href=\"https:\u002F\u002Fdocs.unstructured.io\u002Fingestion\u002Foverview\">https:\u002F\u002Fdocs.unstructured.io\u002Fingestion\u002Foverview\u003C\u002Fa>\u003C\u002Fli>\n          \u003Cli>Installation guide: \u003Ca href=\"https:\u002F\u002Fdocs.unstructured.io\u002Fopen-source\u002Finstallation\u002Ffull-installation\">https:\u002F\u002Fdocs.unstructured.io\u002Fopen-source\u002Finstallation\u002Ffull-installation\u003C\u002Fa>\u003C\u002Fli>\n        \u003C\u002Ful>\n      \u003C\u002Fli>\n    \u003C\u002Ful>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (Docling)\u003C\u002Fsummary>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"file_loader\", \"DoclingLoader\", {})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> Currently supported file types: please refer to the Docling documentation: https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fusage\u002Fsupported_formats\u002F#supported-output-formats \u003C\u002Fp>\n    \u003Cp> You need to install docling before running, execute: \u003Ccode>pip install docling\u003C\u002Fcode>. More details about Docling: https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002F \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n#### Web Crawler Configuration\n\u003Cpre>\u003Ccode>config.set_provider_config(\"web_crawler\", \"(WebCrawlerName)\", \"(Arguments dict)\")\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>The \"WebCrawlerName\" can be one of the following: [\"FireCrawlCrawler\", \"Crawl4AICrawler\", \"JinaCrawler\"]\u003C\u002Fp>\n\u003Cp> The \"Arguments dict\" is a dictionary that contains the necessary arguments for the Web Crawler class.\u003C\u002Fp>\n\n\u003Cdetails>\n  \u003Csummary>Example (FireCrawl)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your FireCrawl API KEY as an env variable \u003Ccode>FIRECRAWL_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"web_crawler\", \"FireCrawlCrawler\", {})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> More details about FireCrawl: https:\u002F\u002Fdocs.firecrawl.dev\u002Fintroduction \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (Crawl4AI)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have run \u003Ccode>crawl4ai-setup\u003C\u002Fcode> in your environment.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"web_crawler\", \"Crawl4AICrawler\", {\"browser_config\": {\"headless\": True, \"verbose\": True}})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> You need to install crawl4ai before running, execute: \u003Ccode>pip install crawl4ai\u003C\u002Fcode>. More details about Crawl4AI: https:\u002F\u002Fdocs.crawl4ai.com\u002F \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (Jina Reader)\u003C\u002Fsummary>\n    \u003Cp> Make sure you have prepared your Jina Reader API KEY as an env variable \u003Ccode>JINA_API_TOKEN\u003C\u002Fcode> or \u003Ccode>JINAAI_API_KEY\u003C\u002Fcode>.\u003C\u002Fp>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"web_crawler\", \"JinaCrawler\", {})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> More details about Jina Reader: https:\u002F\u002Fjina.ai\u002Freader\u002F \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n  \u003Csummary>Example (Docling)\u003C\u002Fsummary>\n    \u003Cpre>\u003Ccode>config.set_provider_config(\"web_crawler\", \"DoclingCrawler\", {})\u003C\u002Fcode>\u003C\u002Fpre>\n    \u003Cp> Currently supported file types: please refer to the Docling documentation: https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002Fusage\u002Fsupported_formats\u002F#supported-output-formats \u003C\u002Fp>\n    \u003Cp> You need to install docling before running, execute: \u003Ccode>pip install docling\u003C\u002Fcode>. More details about Docling: https:\u002F\u002Fdocling-project.github.io\u002Fdocling\u002F \u003C\u002Fp>\n\u003C\u002Fdetails>\n\n\n### Python CLI Mode\n#### Load\n```shell\ndeepsearcher load \"your_local_path_or_url\"\n# load into a specific collection\ndeepsearcher load \"your_local_path_or_url\" --collection_name \"your_collection_name\" --collection_desc \"your_collection_description\"\n```\nExample loading from local file:\n```shell\ndeepsearcher load \"\u002Fpath\u002Fto\u002Fyour\u002Flocal\u002Ffile.pdf\"\n# or more files at once\ndeepsearcher load \"\u002Fpath\u002Fto\u002Fyour\u002Flocal\u002Ffile1.pdf\" \"\u002Fpath\u002Fto\u002Fyour\u002Flocal\u002Ffile2.md\"\n```\nExample loading from url (*Set `FIRECRAWL_API_KEY` in your environment variables, see [FireCrawl](https:\u002F\u002Fdocs.firecrawl.dev\u002Fintroduction) for more details*):\n\n```shell\ndeepsearcher load \"https:\u002F\u002Fwww.wikiwand.com\u002Fen\u002Farticles\u002FDeepSeek\"\n```\n\n#### Query\n```shell\ndeepsearcher query \"Write a report about xxx.\"\n```\n\nMore help information\n```shell\ndeepsearcher --help\n```\nFor more help information about a specific subcommand, you can use `deepsearcher [subcommand] --help`.\n```shell\ndeepsearcher load --help\ndeepsearcher query --help\n```\n\n### Deployment\n\n#### Configure modules\n\nYou can configure all arguments by modifying [config.yaml](.\u002Fconfig.yaml) to set up your system with default modules.\nFor example, set your `OPENAI_API_KEY` in the `llm` section of the YAML file.\n\n#### Start service\nThe main script will run a FastAPI service with default address `localhost:8000`.\n\n```shell\n$ python main.py\n```\n\n#### Access via browser\n\nYou can open url http:\u002F\u002Flocalhost:8000\u002Fdocs in browser to access the web service.\nClick on the button \"Try it out\", it allows you to fill the parameters and directly interact with the API.\n\n\n---\n\n## ❓ Q&A\n\n**Q1**: Why I failed to parse LLM output format \u002F How to select the LLM?\n\n\n**A1**: Small LLMs struggle to follow the prompt to generate a desired response, which usually cause the format parsing problem. A better practice is to use large reasoning models e.g. deepseek-r1 671b, OpenAI o-series, Claude 4 sonnet, etc. as your LLM. \n\n---\n\n**Q2**: \nOSError: We couldn't connect to 'https:\u002F\u002Fhuggingface.co' to load this file, couldn't find it in the cached files and it looks like GPTCache\u002Fparaphrase-albert-small-v2 is not the path to a directory containing a file named config.json.\nCheckout your internet connection or see how to run the library in offline mode at 'https:\u002F\u002Fhuggingface.co\u002Fdocs\u002Ftransformers\u002Finstallation#offline-mode'.\n\n**A2**: This is mainly due to abnormal access to huggingface, which may be a network or permission problem. You can try the following two methods:\n1. If there is a network problem, set up a proxy, try adding the following environment variable.\n```bash\nexport HF_ENDPOINT=https:\u002F\u002Fhf-mirror.com\n```\n2. If there is a permission problem, set up a personal token, try adding the following environment variable.\n```bash\nexport HUGGING_FACE_HUB_TOKEN=xxxx\n```\n\n---\n\n**Q3**: DeepSearcher doesn't run in Jupyter notebook.\n\n**A3**: Install `nest_asyncio` and then put this code block in front of your jupyter notebook.\n\n```\npip install nest_asyncio\n```\n\n```\nimport nest_asyncio\nnest_asyncio.apply()\n```\n\n---\n\n## 🔧 Module Support\n\n### 🔹 Embedding Models\n- [Open-source embedding models](https:\u002F\u002Fmilvus.io\u002Fdocs\u002Fembeddings.md)\n- [OpenAI](https:\u002F\u002Fplatform.openai.com\u002Fdocs\u002Fguides\u002Fembeddings\u002Fuse-cases) (`OPENAI_API_KEY` env variable required)\n- [VoyageAI](https:\u002F\u002Fdocs.voyageai.com\u002Fembeddings\u002F) (`VOYAGE_API_KEY` env variable required)\n- [Amazon Bedrock](https:\u002F\u002Fdocs.aws.amazon.com\u002Fbedrock\u002F) (`AWS_ACCESS_KEY_ID` and `AWS_SECRET_ACCESS_KEY` env variable required)\n- [FastEmbed](https:\u002F\u002Fqdrant.github.io\u002Ffastembed\u002F)\n- [PPIO](https:\u002F\u002Fppinfra.com\u002Fmodel-api\u002Fproduct\u002Fllm-api?utm_source=github_deep-searcher) (`PPIO_API_KEY` env variable required)\n- [Novita AI](https:\u002F\u002Fnovita.ai\u002Fdocs\u002Fapi-reference\u002Fmodel-apis-llm-create-embeddings?utm_source=github_deep-searcher&utm_medium=github_readme&utm_campaign=link) (`NOVITA_API_KEY` env variable required)\n- [IBM watsonx.ai](https:\u002F\u002Fwww.ibm.com\u002Fproducts\u002Fwatsonx-ai\u002Ffoundation-models#ibmembedding) (`WATSONX_APIKEY`, `WATSONX_URL`, `WATSONX_PROJECT_ID` env variables required)\n- [Jiekou.AI](https:\u002F\u002Fjiekou.ai\u002F?utm_source=github_deep-searcher) (`JIEKOU_API_KEY` env variable required)\n\n### 🔹 LLM Support\n- [OpenAI](https:\u002F\u002Fplatform.openai.com\u002Fdocs\u002Fmodels) (`OPENAI_API_KEY` env variable required)\n- [DeepSeek](https:\u002F\u002Fapi-docs.deepseek.com\u002F) (`DEEPSEEK_API_KEY` env variable required)\n- [XAI Grok](https:\u002F\u002Fx.ai\u002Fapi) (`XAI_API_KEY` env variable required)\n- [Anthropic Claude](https:\u002F\u002Fdocs.anthropic.com\u002Fen\u002Fhome) (`ANTHROPIC_API_KEY` env variable required)\n- [SiliconFlow Inference Service](https:\u002F\u002Fdocs.siliconflow.cn\u002Fen\u002Fuserguide\u002Fintroduction) (`SILICONFLOW_API_KEY` env variable required)\n- [PPIO](https:\u002F\u002Fppinfra.com\u002Fmodel-api\u002Fproduct\u002Fllm-api?utm_source=github_deep-searcher) (`PPIO_API_KEY` env variable required)\n- [TogetherAI Inference Service](https:\u002F\u002Fdocs.together.ai\u002Fdocs\u002Fintroduction) (`TOGETHER_API_KEY` env variable required)\n- [Google Gemini](https:\u002F\u002Fai.google.dev\u002Fgemini-api\u002Fdocs) (`GEMINI_API_KEY` env variable required)\n- [SambaNova Cloud Inference Service](https:\u002F\u002Fdocs.together.ai\u002Fdocs\u002Fintroduction) (`SAMBANOVA_API_KEY` env variable required)\n- [Ollama](https:\u002F\u002Follama.com\u002F)\n- [Novita AI](https:\u002F\u002Fnovita.ai\u002Fdocs\u002Fguides\u002Fintroduction?utm_source=github_deep-searcher&utm_medium=github_readme&utm_campaign=link) (`NOVITA_API_KEY` env variable required)\n- [IBM watsonx.ai](https:\u002F\u002Fwww.ibm.com\u002Fproducts\u002Fwatsonx-ai\u002Ffoundation-models#ibmfm) (`WATSONX_APIKEY`, `WATSONX_URL`, `WATSONX_PROJECT_ID` env variable required)\n- [Jiekou.AI](https:\u002F\u002Fjiekou.ai\u002F?utm_source=github_deep-searcher) (`JIEKOU_API_KEY` env variable required)\n\n### 🔹 Document Loader\n- Local File\n  - PDF(with txt\u002Fmd) loader\n  - [Unstructured](https:\u002F\u002Funstructured.io\u002F) (under development) (`UNSTRUCTURED_API_KEY` and `UNSTRUCTURED_URL` env variables required)\n- Web Crawler\n  - [FireCrawl](https:\u002F\u002Fdocs.firecrawl.dev\u002Fintroduction) (`FIRECRAWL_API_KEY` env variable required)\n  - [Jina Reader](https:\u002F\u002Fjina.ai\u002Freader\u002F) (`JINA_API_TOKEN` env variable required)\n  - [Crawl4AI](https:\u002F\u002Fdocs.crawl4ai.com\u002F) (You should run command `crawl4ai-setup` for the first time)\n\n### 🔹 Vector Database Support\n- [Milvus](https:\u002F\u002Fmilvus.io\u002F) and [Zilliz Cloud](https:\u002F\u002Fwww.zilliz.com\u002F) (fully managed Milvus)\n- [Qdrant](https:\u002F\u002Fqdrant.tech\u002F)\n\n---\n## 📊 Evaluation \nSee the [Evaluation](.\u002Fevaluation) directory for more details.\n\n---\n## 📌 Future Plans\n- Enhance web crawling functionality\n- Support more vector databases (e.g., FAISS...)\n- Add support for additional large models\n- Provide RESTful API interface (**DONE**)\n\nWe welcome contributions! Star & Fork the project and help us build a more powerful DeepSearcher! 🎯\n","DeepSearcher是一个基于私有数据进行搜索、评估和推理的开源深度研究工具，使用Python编写。该项目结合了先进的大语言模型（如OpenAI o3, Qwen3等）与向量数据库（如Milvus），提供精准答案及全面报告。其核心技术特点包括支持多种嵌入模型、兼容多款大型语言模型以及具备文档加载功能。适用于企业知识管理、智能问答系统及信息检索场景。",2,"2026-06-11 03:39:53","high_star"]