From 5b4aa22bfa262215e52f6b1a9313ee85352ef222 Mon Sep 17 00:00:00 2001 From: William Fu-Hinthorn <13333726+hinthornw@users.noreply.github.com> Date: Tue, 17 Sep 2024 10:35:58 -0700 Subject: [PATCH 1/5] Rename to extraction_schema --- README.md | 7 ++++--- src/enrichment_agent/graph.py | 6 +++--- src/enrichment_agent/state.py | 2 +- src/enrichment_agent/tools.py | 2 +- 4 files changed, 9 insertions(+), 8 deletions(-) diff --git a/README.md b/README.md index 182c39b..ea1d040 100644 --- a/README.md +++ b/README.md @@ -2,6 +2,7 @@ [![CI](https://github.com/langchain-ai/data-enrichment/actions/workflows/unit-tests.yml/badge.svg)](https://github.com/langchain-ai/data-enrichment/actions/workflows/unit-tests.yml) [![Integration Tests](https://github.com/langchain-ai/data-enrichment/actions/workflows/integration-tests.yml/badge.svg)](https://github.com/langchain-ai/data-enrichment/actions/workflows/integration-tests.yml) +[![Open in - LangGraph Studio](https://img.shields.io/badge/Open_in-LangGraph_Studio-00324d.svg?logo=data:image/svg%2bxml;base64,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)](https://langgraph-studio.vercel.app/templates/open?githubUrl=https://github.com/langchain-ai/data-enrichment) This is a starter project to help you get started with developing a data enrichment agent using [LangGraph](https://github.com/langchain-ai/langgraph) in [LangGraph Studio](https://github.com/langchain-ai/langgraph-studio). @@ -13,13 +14,13 @@ It contains an example graph exported from `src/enrichment_agent/graph.py` that The enrichment agent: -1. Takes a research **topic** and requested **row_schema** as input +1. Takes a research **topic** and requested **extraction_schema** as input. The 2. Searches the web for relevant information 3. Reads and extracts key details from websites 4. Organizes the findings into the requested structured format 5. Validates the gathered information for completeness and accuracy -By default, it's set up to gather information based on the user-provided schema passed through the `row_schema` key in the state. +By default, it's set up to gather information based on the user-provided schema passed through the `extraction_schema` key in the state. ## Getting Started @@ -92,7 +93,7 @@ End setup instructions ## How to customize -1. **Customize research targets**: Provide a custom `row_schema` when calling the graph to gather different types of information. +1. **Customize research targets**: Provide a custom `extraction_schema` when calling the graph to gather different types of information. 2. **Select a different model**: We default to anthropic (sonnet-35). You can select a compatible chat model using `provider/model-name` via configuration. Example: `openai/gpt-4o-mini`. 3. **Customize the prompt**: We provide a default prompt in [prompts.py](./src/enrichment_agent/prompts.py). You can easily update this via configuration in the studio. diff --git a/src/enrichment_agent/graph.py b/src/enrichment_agent/graph.py index f5b82fe..7c9edb8 100644 --- a/src/enrichment_agent/graph.py +++ b/src/enrichment_agent/graph.py @@ -29,11 +29,11 @@ async def call_agent_model( info_tool = { "name": "Info", "description": "Call this when you have gathered all the relevant info", - "parameters": state.row_schema, + "parameters": state.extraction_schema, } p = configuration.prompt.format( - info=json.dumps(state.row_schema, indent=2), topic=state.topic + info=json.dumps(state.extraction_schema, indent=2), topic=state.topic ) messages = [HumanMessage(content=p)] + state.messages raw_model = init_model(config) @@ -82,7 +82,7 @@ async def reflect( ) -> Dict[str, Any]: """Validate the quality of the data enrichment agent's calls.""" p = prompts.MAIN_PROMPT.format( - info=json.dumps(state.row_schema, indent=2), topic=state.topic + info=json.dumps(state.extraction_schema, indent=2), topic=state.topic ) messages = [HumanMessage(content=p)] + state.messages[:-1] presumed_info = state.info diff --git a/src/enrichment_agent/state.py b/src/enrichment_agent/state.py index f7374b0..ada0b8f 100644 --- a/src/enrichment_agent/state.py +++ b/src/enrichment_agent/state.py @@ -19,7 +19,7 @@ class InputState: topic: str "The topic for which the agent is tasked to gather information." - row_schema: dict[str, Any] + extraction_schema: dict[str, Any] "The json schema defines the information the agent is tasked with filling out." info: Optional[dict[str, Any]] = field(default=None) diff --git a/src/enrichment_agent/tools.py b/src/enrichment_agent/tools.py index b746328..9e3b2e9 100644 --- a/src/enrichment_agent/tools.py +++ b/src/enrichment_agent/tools.py @@ -62,7 +62,7 @@ async def scrape_website( content = await response.text() p = _INFO_PROMPT.format( - info=json.dumps(state.row_schema, indent=2), + info=json.dumps(state.extraction_schema, indent=2), url=url, content=content, ) From 542acc4885484cfa94dc62b4aabf779856770ed9 Mon Sep 17 00:00:00 2001 From: William Fu-Hinthorn <13333726+hinthornw@users.noreply.github.com> Date: Tue, 17 Sep 2024 10:56:11 -0700 Subject: [PATCH 2/5] Update test --- src/enrichment_agent/graph.py | 2 +- .../48fccba2-fb6a-440c-96da-4439d876cf5e.yaml | 1443 +++++++++++++++++ tests/integration_tests/test_graph.py | 81 +- 3 files changed, 1522 insertions(+), 4 deletions(-) diff --git a/src/enrichment_agent/graph.py b/src/enrichment_agent/graph.py index 7c9edb8..281b59c 100644 --- a/src/enrichment_agent/graph.py +++ b/src/enrichment_agent/graph.py @@ -49,7 +49,7 @@ async def call_agent_model( break if info is not None: # The agent is submitting their answer; - # ensure it isnt' erroneously attempting to simultaneously perform research + # ensure it isn't erroneously attempting to simultaneously perform research response.tool_calls = [ next(tc for tc in response.tool_calls if tc["name"] == "Info") ] diff --git a/tests/casettes/48fccba2-fb6a-440c-96da-4439d876cf5e.yaml b/tests/casettes/48fccba2-fb6a-440c-96da-4439d876cf5e.yaml index a0dc275..59d90b6 100644 --- a/tests/casettes/48fccba2-fb6a-440c-96da-4439d876cf5e.yaml +++ b/tests/casettes/48fccba2-fb6a-440c-96da-4439d876cf5e.yaml @@ -3017,4 +3017,1447 @@ interactions: status: code: 200 message: OK +- request: + body: '{"max_tokens": 1024, "messages": [{"role": "user", "content": "You are + doing web research on behalf of a user. You are trying to figure out this information:\n\n\n{\n \"type\": + \"object\",\n \"properties\": {\n \"founder\": {\n \"type\": \"string\",\n \"description\": + \"The name of the company founder.\"\n },\n \"websiteUrl\": {\n \"type\": + \"string\",\n \"description\": \"Website URL of the company, e.g.: https://openai.com/, + or https://microsoft.com\"\n },\n \"products_sold\": {\n \"type\": + \"array\",\n \"items\": {\n \"type\": \"string\"\n },\n \"description\": + \"A list of products sold by the company.\"\n }\n },\n \"required\": [\n \"founder\",\n \"websiteUrl\",\n \"products_sold\"\n ]\n}\n\n\nYou + have access to the following tools:\n\n- `Search`: call a search tool and get + back some results\n- `ScrapeWebsite`: scrape a website and get relevant notes + about the given request. This will update the notes above.\n- `Info`: call this + when you are done and have gathered all the relevant info\n\nHere is the information + you have about the topic you are researching:\n\nTopic: LangChain"}], "model": + "claude-3-5-sonnet-20240620", "tool_choice": {"type": "any"}, "tools": [{"name": + "scrape_website", "description": "Scrape and summarize content from a given + URL.", "input_schema": {"type": "object", "properties": {"url": {"type": "string"}}, + "required": ["url"]}}, {"name": "search", "description": "Search for general + results. This function performs a search using the Tavily search engine, which + is designed\nto provide comprehensive, accurate, and trusted results. It''s + particularly useful\nfor answering questions about current events.", "input_schema": + {"type": "object", "properties": {"query": {"type": "string"}}, "required": + ["query"]}}, {"name": "Info", "description": "Call this when you have gathered + all the relevant info", "input_schema": {"type": "object", "properties": {"founder": + {"type": "string", "description": "The name of the company founder."}, "websiteUrl": + {"type": "string", "description": "Website URL of the company, e.g.: https://openai.com/, + or https://microsoft.com"}, "products_sold": {"type": "array", "items": {"type": + "string"}, "description": "A list of products sold by the company."}}, "required": + ["founder", "websiteUrl", "products_sold"]}}]}' + headers: {} + method: POST + uri: https://api.anthropic.com/v1/messages + response: + body: + string: !!binary | + H4sIAAAAAAAAA1RQTUvDQBD9K2HOG0hrrWVv0oNQLLSI0ioS1t1pEk1mkswsUkP+u6Qi6HHevC/e + AFUAC40UeTbbvh/8njd3wbuy2D1t97vuuVUwoOcWJxaKuALBQM/1BDiRStTRxGk4YA0WfO1iwPQq + vU6FiVDTeTZfZMt5BgY8kyIp2Jfh11SZ6zzK5HqpMt0xz2YbpYfb074KpbvZ8HF96B6PXx4MkGsm + naDrfTmpqI0KdoAuYn8GC/eOinXpKko8N62jc3LiSAF7k3zim1SKJnEUkrbnEL0KjOOrAVFu8x6d + MP1vdXkIdhHJI1iKdW0gXoaww096rvyBJGBXq6UBjvoXWyzH8RsAAP//AwC9EaJ+ZwEAAA== + headers: + CF-Cache-Status: + - DYNAMIC + CF-RAY: + - 8c4af82538cb230a-SJC + Connection: + - keep-alive + Content-Encoding: + - gzip + Content-Type: + - application/json + Date: + - Tue, 17 Sep 2024 17:55:50 GMT + Server: + - cloudflare + Transfer-Encoding: + - chunked + X-Robots-Tag: + - none + anthropic-ratelimit-requests-limit: + - '4000' + anthropic-ratelimit-requests-remaining: + - '3999' + anthropic-ratelimit-requests-reset: + - '2024-09-17T17:55:54Z' + anthropic-ratelimit-tokens-limit: + - '400000' + anthropic-ratelimit-tokens-remaining: + - '400000' + anthropic-ratelimit-tokens-reset: + - '2024-09-17T17:55:50Z' + request-id: + - req_01FAWNKeAwKQRx8VwSLtkKFB + via: + - 1.1 google + x-cloud-trace-context: + - 469797dbc792f48da3e4b25f9fe3a121 + status: + code: 200 + message: OK +- request: + body: '{"max_tokens": 1024, "messages": [{"role": "user", "content": "You are + doing web research on behalf of a user. You are trying to figure out this information:\n\n\n{\n \"type\": + \"object\",\n \"properties\": {\n \"founder\": {\n \"type\": \"string\",\n \"description\": + \"The name of the company founder.\"\n },\n \"websiteUrl\": {\n \"type\": + \"string\",\n \"description\": \"Website URL of the company, e.g.: https://openai.com/, + or https://microsoft.com\"\n },\n \"products_sold\": {\n \"type\": + \"array\",\n \"items\": {\n \"type\": \"string\"\n },\n \"description\": + \"A list of products sold by the company.\"\n }\n },\n \"required\": [\n \"founder\",\n \"websiteUrl\",\n \"products_sold\"\n ]\n}\n\n\nYou + have access to the following tools:\n\n- `Search`: call a search tool and get + back some results\n- `ScrapeWebsite`: scrape a website and get relevant notes + about the given request. This will update the notes above.\n- `Info`: call this + when you are done and have gathered all the relevant info\n\nHere is the information + you have about the topic you are researching:\n\nTopic: LangChain"}, {"role": + "assistant", "content": [{"type": "tool_use", "name": "search", "input": {"query": + "LangChain company founder, website, and products"}, "id": "toolu_01JtnSAfQidha7JoYCXqUYzc"}]}, + {"role": "user", "content": [{"type": "tool_result", "content": "[{\"url\": + \"https://www.cbinsights.com/company/langchain\", \"content\": \"It was founded + in 2023 and is based in San Francisco, California. LangChain specializes in + the development of large language model (LLM) applications, providing a suite + of products that support developers throughout the application lifecycle. Use + the CB Insights Platform to explore LangChain''s full profile.\"}, {\"url\": + \"https://en.wikipedia.org/wiki/LangChain\", \"content\": \"In October 2023 + LangChain introduced LangServe, a deployment tool designed to facilitate the + transition from LCEL (LangChain Expression Language) prototypes to production-ready + applications.[5]\\nIntegrations[edit]\\nAs of March 2023, LangChain included + integrations with systems including Amazon, Google, and Microsoft Azure cloud + storage; API wrappers for news, movie information, and weather; Bash for summarization, + syntax and semantics checking, and execution of shell scripts; multiple web + scraping subsystems and templates; few-shot learning prompt generation support; + finding and summarizing \\\"todo\\\" tasks in code; Google Drive documents, + spreadsheets, and presentations summarization, extraction, and creation; Google + Search and Microsoft Bing web search; OpenAI, Anthropic, and Hugging Face language + models; iFixit repair guides and wikis search and summarization; MapReduce for + question answering, combining documents, and question generation; N-gram overlap + scoring; PyPDF, pdfminer, fitz, and pymupdf for PDF file text extraction and + manipulation; Python and JavaScript code generation, analysis, and debugging; + Milvus vector database[6] to store and retrieve vector embeddings; Weaviate + vector database[7] to cache embedding and data objects; Redis cache database + storage; Python RequestsWrapper and other methods for API requests; SQL and + NoSQL databases including JSON support; Streamlit, including for logging; text + mapping for k-nearest neighbors search; time zone conversion and calendar operations; + tracing and recording stack symbols in threaded and asynchronous subprocess + runs; and the Wolfram Alpha website and SDK.[8] As a language model integration + framework, LangChain''s use-cases largely overlap with those of language models + in general, including document analysis and summarization, chatbots, and code + analysis.[2]\\nHistory[edit]\\nLangChain was launched in October 2022 as an + open source project by Harrison Chase, while working at machine learning startup + Robust Intelligence. In April 2023, LangChain had incorporated and the new startup + raised over $20 million in funding at a valuation of at least $200 million from + venture firm Sequoia Capital, a week after announcing a $10 million seed investment + from Benchmark.[3][4]\\n The project quickly garnered popularity, with improvements + from hundreds of contributors on GitHub, trending discussions on Twitter, lively + activity on the project''s Discord server, many YouTube tutorials, and meetups + in San Francisco and London. As of April 2023, it can read from more than 50 + document types and data sources.[9]\\nReferences[edit]\\nExternal links[edit]\"}, + {\"url\": \"https://www.forbes.com/companies/langchain/\", \"content\": \"LangChain. + Founded in late 2022, LangChain was an instant hit in Silicon Valley for its + software that enables anyone to build apps powered by large language models + like GPT-4 in as little as two ...\"}, {\"url\": \"https://tedai-sanfrancisco.ted.com/speakers-1/harrison-chase-/co-founder-and-ceo,-langchain/\", + \"content\": \"Harrison Chase, a Harvard graduate in statistics and computer + science, co-founded LangChain to streamline the development of Language Model + applications with open-source Python/Typescript packages. Chase''s experience + includes heading the Machine Learning team at Robust Intelligence, focusing + on the testing and validation of machine learning models, and leading the entity + linking team at Kensho ...\"}, {\"url\": \"https://www.crunchbase.com/organization/langchain\", + \"content\": \"Summary. LangChain may be growing as evidenced by securing $25 + million in funding, which is a significant financial milestone for the company. + Additionally, the launch of a new platform to facilitate the full lifecycle + of Large Language Model applications suggests expansion in their product offerings. + The integration of Gemini Pro API, enabling ...\"}, {\"url\": \"https://craft.co/langchain\", + \"content\": \"LangChain has 5 employees at their 1 location and $35 m in total + funding,. See insights on LangChain including office locations, competitors, + revenue, financials, executives, subsidiaries and more at Craft. ... Harrison + Chase, Co-Founder and Chief Executive Officer. Linkedin;\"}, {\"url\": \"https://www.langchain.com/about\", + \"content\": \"Ready to start shipping\\nreliable GenAI apps faster?\\nLangChain + and LangSmith are critical parts of the reference\\narchitecture to get you + from prototype to production. Announcing the General Availability of LangSmith\\nToday, + we\\u2019re thrilled to announce the general availability of LangSmith \\u2014 + our solution for LLM application development, monitoring, and testing.\\n We + build products that enable developers to go from an idea to working code in + an afternoon and in the hands of users in days or weeks. The biggest developer + community in \\u00a0GenAI\\nLearn alongside the 100K+ practitioners\\nwho are + pushing the industry forward.\\n And we built LangSmith to support all stages + of the AI engineering lifecycle, to get applications into production faster.\\n\"}, + {\"url\": \"https://www.bluelabellabs.com/blog/what-is-langchain/\", \"content\": + \"Similar setups to AutoTweeter would allow businesses to parse past and present + customer queries to understand drop-off points and build the appropriate flows + to solve customer problems on the spot, instead of spitting them into a queue + to wait for a human and plot how they\\u2019re going to steal from your business.\\n + Rather than creating a bot to tweet in your tone on your behalf about your love + for processed snacks, hockey, and political happenings, you could train such + a system on your brand voice, knowledge bases, and other resources to create + a highly valuable feature for your digital touchpoints.\\n Here, you essentially + create a curated ML system that\\u2019s tailored to your brand with specific + issues and requests from your customers \\u2013 with a little bit of dev magic, + you can quickly build a solution that integrates features like payment systems, + tech support, ordering, account management, and so much more.\\n Creating a + bot for social media that can post just like you\\nSocial media already has + enough bots engaging with people as if they\\u2019re real people, but for the + sake of making a functional example, our co-founder Bobby created a bot called + AutoTweeter for the social media platform we all still call Twitter.\\n How + does LangChain work?\\nLangchain essentially serves as a middleman between other + LLMs, data sources, and API-accessible functions \\u2013 more precisely, it\\u2019s + a comprehensive open-source library of components that allow users to build + better (i.e., more custom) solutions by interconnecting different resources.\\n\"}, + {\"url\": \"https://www.latent.space/p/langchain\", \"content\": \"And so you + can do a lot of things like reflection and kind of like making sure that you''re + on the right track and really put more safeguards and thinking about these agents + as opposed to relying on like chas and interface, like the bot we have that''s + on GitHub answering questions on the issues, I think probably gives better answers + than the bots that we have that are on chat on the website. [00:50:32]\\nHarrison: + And then between kind of like allowing for, I think one helpful thing that we + did in terms of like distinguishing between basically the base interfaces and + then more complex stuff is part of the separation around the docs is there''s + like the components piece, which has the model IO, the retrieval, the agents, + the callbacks, things like that. And then like, I know a lot of people are doing + this, but like a personal assistant for like managing like email calendar, basic + stuff, which I think is, I think that''s like a fantastic application for these + like agent like things, because if you think about personal assistants today, + you usually interact, I don''t have one, but I''m told you interact with them + over email. Show Notes\\nLangChain\\nLangChain\\u2019s Berkshire Hathaway Homepage\\nAbstractions + tweet\\nLangSmith\\nLangSmith Cookbooks repo\\nLangChain Retrieval blog\\nEvaluating + CSV Question/Answering blog and YouTube\\nMultiOn Partner blog\\nHarvard Sports + Analytics Collective\\nEvaluating RAG Webinar\\nawesome-langchain:\\nLLM Math + Chain\\nSelf-Ask\\nLangChain Hub UI\\n\\u201cLangChain is Pointless\\u201d\\nHarrison\\u2019s + links\\nsports - estimating player compatibility in the NBA\\nearly interest + in prompt injections\\nGitHub\\nTwitter\\nTimestamps\\n[00:00:00] Introduction\\n[00:00:48] + Harrison''s background and how sports led him into ML\\n[00:04:54] Memory (for + long running chats, whether a simple Buffer, Knowledge Graph, Summary, or Vector + Store)\\nUse-Cases: compositions of Components\\nChains: combining a PromptTemplate, + LLM Model and optional OutputParser\\nwith Router, Sequential, and Transform + Chains for advanced usecases\\nsavable, sharable schemas that can be loaded + from LangChainHub\\nAgents: a chain that has access to a suite of tools, of + nondeterministic8 length because the LLM is used as a reasoning engine to determine + which actions to take and in which order.\"}, {\"url\": \"https://www.langchain.com/\", + \"content\": \"Join Discord\\nContribute on GitHub\\n5.3M\\nMonthly Downloads\\n67k\\nGithub + Stars\\n1800+\\nContributors\\n34k+\\nApps Powered\\nReady to build?\\nGo To + Docs\\nGo To Docs\\nGo To Docs\\nContact Sales\\nContact Sales\\nContact Sales\\nProducts + and use-cases\\nLangChain\\nLangSmith\\nRetrieval\\nAgents\\nInspiration\\nCode\\nGitHub\\nLangChain + Hub\\nPython Docs\\nJS/TS Docs\\nSocial\\nTwitter\\nDiscord\\nBlog\\nLinkedIn\\nYouTube\\nTerms + of Service\\nSign up for our newsletter\\nProducts and use-cases\\nLangChain\\nLangSmith\\nRetrieval\\nAgents\\nInspiration\\nCode\\nGitHub\\nLangChain + Hub\\nPython Docs\\nJS/TS Docs\\nSocial\\nTwitter\\nDiscord\\nBlog\\nLinkedIn\\nYouTube\\nTerms + of Service\\nSign up for our newsletter\\nProducts and use-cases\\nLangChain\\nLangSmith\\nRetrieval\\nAgents\\nInspiration\\nCode\\nGitHub\\nLangChain + Hub\\nPython Docs\\nJS/TS Docs\\nSocial\\nTwitter\\nDiscord\\nBlog\\nLinkedIn\\nYouTube\\nTerms + of Service\\nSign up for our newsletter Chatbots\\nQ&A Over Docs\\nSummarization\\nCopilots\\nWorkflow + Automation\\nDocument Analysis\\nCustom Search\\nHarness the power\\u2013and + wrangle the complexity\\u2013 of LLMs\\nGet from proof-of-concept to production + on one platform\\nRely on a set of tools built especially for non-deterministic + models\\nLeverage new cognitive architectures and battle-tested orchestration\\nThanks + to our contributors and partners for making LangChain awesome\\nWhen you build + with LangChain, you build alongside a community of developers advancing the + frontier of what\\u2019s possible with LLMs.\\n \\ud83e\\udd9c\\ud83d\\udd17 + LangChain\\nLangSmith\\nLangServe\\nAgents\\nRetrieval\\nEvaluation\\nBlog\\nDocs\\n\\ud83e\\udd9c\\ud83d\\udd17 + LangChain\\n\\ud83e\\udd9c\\ud83d\\udd17 LangChain\\nLangChain Expression Language + makes it easy to create custom chains.\\n Go To Docs\\nGo To Docs\\nGo To Docs\\nContact + Sales\\nglobal corporations, STARTUPS, and TINKERERS build with LangChain\\nglobal + corporations, STARTUPS, and TINKERERS build with LangChain\\nglobal corporations, + STARTUPS, and TINKERERS build with LangChain\\nA complete set of powerful building + blocks\\nA complete set of powerful building blocks\\n Try it out here\\nGet + your LLM application from prototype to production\\nGet your LLM application + from prototype to production\\nBuild context-aware, reasoning applications with + LangChain\\u2019s flexible abstractions and AI-first toolkit.\\n\"}]", "tool_use_id": + "toolu_01JtnSAfQidha7JoYCXqUYzc", "is_error": false}]}], "model": "claude-3-5-sonnet-20240620", + "tool_choice": {"type": "any"}, "tools": [{"name": "scrape_website", "description": + "Scrape and summarize content from a given URL.", "input_schema": {"type": "object", + "properties": {"url": {"type": "string"}}, "required": ["url"]}}, {"name": "search", + "description": "Search for general results. This function performs a search + using the Tavily search engine, which is designed\nto provide comprehensive, + accurate, and trusted results. It''s particularly useful\nfor answering questions + about current events.", "input_schema": {"type": "object", "properties": {"query": + {"type": "string"}}, "required": ["query"]}}, {"name": "Info", "description": + "Call this when you have gathered all the relevant info", "input_schema": {"type": + "object", "properties": {"founder": {"type": "string", "description": "The name + of the company founder."}, "websiteUrl": {"type": "string", "description": "Website + URL of the company, e.g.: https://openai.com/, or https://microsoft.com"}, "products_sold": + {"type": "array", "items": {"type": "string"}, "description": "A list of products + sold by the company."}}, "required": ["founder", "websiteUrl", "products_sold"]}}]}' + headers: {} + method: POST + uri: https://api.anthropic.com/v1/messages + response: + body: + string: !!binary | + H4sIAAAAAAAAA1SP3UrDQBCF32WuN81P21j3TigtFBGxiKhI2CZjG7KZTTOzRBvy7pJIL3o5Z+ac + +U4PZQEaaj5mUZzstunm4523fr9cn6rvx2q3whQUyG+D4xUymyOCgtbZUTDMJYshAQW1K9CChtwa + X2AwD5YBOyKUIImSRZQmESjIHQmSgP7sr6HinM08j6kTyjj7LIpXL2d6Pcx/0uXlbXP/vF9fHpIn + BgVk6tHHeWsazDo8cCmTmxovoHvw7chxEmlYh2HXdTNr6JifTEmz3NUhDMOXAhbXZC0adnRLMS0Y + zx4pR9DkrVXgp+K6//+SiauQGPQijhMFzsuNeDcMfwAAAP//AwAZCFJOWAEAAA== + headers: + CF-Cache-Status: + - DYNAMIC + CF-RAY: + - 8c4af82f7cc9fa72-SJC + Connection: + - keep-alive + Content-Encoding: + - gzip + Content-Type: + - application/json + Date: + - Tue, 17 Sep 2024 17:55:52 GMT + Server: + - cloudflare + Transfer-Encoding: + - chunked + X-Robots-Tag: + - none + anthropic-ratelimit-requests-limit: + - '4000' + anthropic-ratelimit-requests-remaining: + - '3999' + anthropic-ratelimit-requests-reset: + - '2024-09-17T17:55:54Z' + anthropic-ratelimit-tokens-limit: + - '400000' + anthropic-ratelimit-tokens-remaining: + - '400000' + anthropic-ratelimit-tokens-reset: + - '2024-09-17T17:55:52Z' + request-id: + - req_0112C84Bvx9Tnzv4NTYA8K4t + via: + - 1.1 google + x-cloud-trace-context: + - efac6bac769f8e88ae6445e24b167f86 + status: + code: 200 + message: OK +- request: + body: '{"max_tokens": 1024, "messages": [{"role": "user", "content": "You are + doing web research on behalf of a user. You are trying to figure out this information:\n\n\n{\n \"type\": + \"object\",\n \"properties\": {\n \"founder\": {\n \"type\": \"string\",\n \"description\": + \"The name of the company founder.\"\n },\n \"websiteUrl\": {\n \"type\": + \"string\",\n \"description\": \"Website URL of the company, e.g.: https://openai.com/, + or https://microsoft.com\"\n },\n \"products_sold\": {\n \"type\": + \"array\",\n \"items\": {\n \"type\": \"string\"\n },\n \"description\": + \"A list of products sold by the company.\"\n }\n },\n \"required\": [\n \"founder\",\n \"websiteUrl\",\n \"products_sold\"\n ]\n}\n\n\nYou + have access to the following tools:\n\n- `Search`: call a search tool and get + back some results\n- `ScrapeWebsite`: scrape a website and get relevant notes + about the given request. This will update the notes above.\n- `Info`: call this + when you are done and have gathered all the relevant info\n\nHere is the information + you have about the topic you are researching:\n\nTopic: LangChain"}, {"role": + "assistant", "content": [{"type": "tool_use", "name": "search", "input": {"query": + "LangChain company founder, website, and products"}, "id": "toolu_01JtnSAfQidha7JoYCXqUYzc"}]}, + {"role": "user", "content": [{"type": "tool_result", "content": "[{\"url\": + \"https://www.cbinsights.com/company/langchain\", \"content\": \"It was founded + in 2023 and is based in San Francisco, California. LangChain specializes in + the development of large language model (LLM) applications, providing a suite + of products that support developers throughout the application lifecycle. Use + the CB Insights Platform to explore LangChain''s full profile.\"}, {\"url\": + \"https://en.wikipedia.org/wiki/LangChain\", \"content\": \"In October 2023 + LangChain introduced LangServe, a deployment tool designed to facilitate the + transition from LCEL (LangChain Expression Language) prototypes to production-ready + applications.[5]\\nIntegrations[edit]\\nAs of March 2023, LangChain included + integrations with systems including Amazon, Google, and Microsoft Azure cloud + storage; API wrappers for news, movie information, and weather; Bash for summarization, + syntax and semantics checking, and execution of shell scripts; multiple web + scraping subsystems and templates; few-shot learning prompt generation support; + finding and summarizing \\\"todo\\\" tasks in code; Google Drive documents, + spreadsheets, and presentations summarization, extraction, and creation; Google + Search and Microsoft Bing web search; OpenAI, Anthropic, and Hugging Face language + models; iFixit repair guides and wikis search and summarization; MapReduce for + question answering, combining documents, and question generation; N-gram overlap + scoring; PyPDF, pdfminer, fitz, and pymupdf for PDF file text extraction and + manipulation; Python and JavaScript code generation, analysis, and debugging; + Milvus vector database[6] to store and retrieve vector embeddings; Weaviate + vector database[7] to cache embedding and data objects; Redis cache database + storage; Python RequestsWrapper and other methods for API requests; SQL and + NoSQL databases including JSON support; Streamlit, including for logging; text + mapping for k-nearest neighbors search; time zone conversion and calendar operations; + tracing and recording stack symbols in threaded and asynchronous subprocess + runs; and the Wolfram Alpha website and SDK.[8] As a language model integration + framework, LangChain''s use-cases largely overlap with those of language models + in general, including document analysis and summarization, chatbots, and code + analysis.[2]\\nHistory[edit]\\nLangChain was launched in October 2022 as an + open source project by Harrison Chase, while working at machine learning startup + Robust Intelligence. In April 2023, LangChain had incorporated and the new startup + raised over $20 million in funding at a valuation of at least $200 million from + venture firm Sequoia Capital, a week after announcing a $10 million seed investment + from Benchmark.[3][4]\\n The project quickly garnered popularity, with improvements + from hundreds of contributors on GitHub, trending discussions on Twitter, lively + activity on the project''s Discord server, many YouTube tutorials, and meetups + in San Francisco and London. As of April 2023, it can read from more than 50 + document types and data sources.[9]\\nReferences[edit]\\nExternal links[edit]\"}, + {\"url\": \"https://www.forbes.com/companies/langchain/\", \"content\": \"LangChain. + Founded in late 2022, LangChain was an instant hit in Silicon Valley for its + software that enables anyone to build apps powered by large language models + like GPT-4 in as little as two ...\"}, {\"url\": \"https://tedai-sanfrancisco.ted.com/speakers-1/harrison-chase-/co-founder-and-ceo,-langchain/\", + \"content\": \"Harrison Chase, a Harvard graduate in statistics and computer + science, co-founded LangChain to streamline the development of Language Model + applications with open-source Python/Typescript packages. Chase''s experience + includes heading the Machine Learning team at Robust Intelligence, focusing + on the testing and validation of machine learning models, and leading the entity + linking team at Kensho ...\"}, {\"url\": \"https://www.crunchbase.com/organization/langchain\", + \"content\": \"Summary. LangChain may be growing as evidenced by securing $25 + million in funding, which is a significant financial milestone for the company. + Additionally, the launch of a new platform to facilitate the full lifecycle + of Large Language Model applications suggests expansion in their product offerings. + The integration of Gemini Pro API, enabling ...\"}, {\"url\": \"https://craft.co/langchain\", + \"content\": \"LangChain has 5 employees at their 1 location and $35 m in total + funding,. See insights on LangChain including office locations, competitors, + revenue, financials, executives, subsidiaries and more at Craft. ... Harrison + Chase, Co-Founder and Chief Executive Officer. Linkedin;\"}, {\"url\": \"https://www.langchain.com/about\", + \"content\": \"Ready to start shipping\\nreliable GenAI apps faster?\\nLangChain + and LangSmith are critical parts of the reference\\narchitecture to get you + from prototype to production. Announcing the General Availability of LangSmith\\nToday, + we\\u2019re thrilled to announce the general availability of LangSmith \\u2014 + our solution for LLM application development, monitoring, and testing.\\n We + build products that enable developers to go from an idea to working code in + an afternoon and in the hands of users in days or weeks. The biggest developer + community in \\u00a0GenAI\\nLearn alongside the 100K+ practitioners\\nwho are + pushing the industry forward.\\n And we built LangSmith to support all stages + of the AI engineering lifecycle, to get applications into production faster.\\n\"}, + {\"url\": \"https://www.bluelabellabs.com/blog/what-is-langchain/\", \"content\": + \"Similar setups to AutoTweeter would allow businesses to parse past and present + customer queries to understand drop-off points and build the appropriate flows + to solve customer problems on the spot, instead of spitting them into a queue + to wait for a human and plot how they\\u2019re going to steal from your business.\\n + Rather than creating a bot to tweet in your tone on your behalf about your love + for processed snacks, hockey, and political happenings, you could train such + a system on your brand voice, knowledge bases, and other resources to create + a highly valuable feature for your digital touchpoints.\\n Here, you essentially + create a curated ML system that\\u2019s tailored to your brand with specific + issues and requests from your customers \\u2013 with a little bit of dev magic, + you can quickly build a solution that integrates features like payment systems, + tech support, ordering, account management, and so much more.\\n Creating a + bot for social media that can post just like you\\nSocial media already has + enough bots engaging with people as if they\\u2019re real people, but for the + sake of making a functional example, our co-founder Bobby created a bot called + AutoTweeter for the social media platform we all still call Twitter.\\n How + does LangChain work?\\nLangchain essentially serves as a middleman between other + LLMs, data sources, and API-accessible functions \\u2013 more precisely, it\\u2019s + a comprehensive open-source library of components that allow users to build + better (i.e., more custom) solutions by interconnecting different resources.\\n\"}, + {\"url\": \"https://www.latent.space/p/langchain\", \"content\": \"And so you + can do a lot of things like reflection and kind of like making sure that you''re + on the right track and really put more safeguards and thinking about these agents + as opposed to relying on like chas and interface, like the bot we have that''s + on GitHub answering questions on the issues, I think probably gives better answers + than the bots that we have that are on chat on the website. [00:50:32]\\nHarrison: + And then between kind of like allowing for, I think one helpful thing that we + did in terms of like distinguishing between basically the base interfaces and + then more complex stuff is part of the separation around the docs is there''s + like the components piece, which has the model IO, the retrieval, the agents, + the callbacks, things like that. And then like, I know a lot of people are doing + this, but like a personal assistant for like managing like email calendar, basic + stuff, which I think is, I think that''s like a fantastic application for these + like agent like things, because if you think about personal assistants today, + you usually interact, I don''t have one, but I''m told you interact with them + over email. Show Notes\\nLangChain\\nLangChain\\u2019s Berkshire Hathaway Homepage\\nAbstractions + tweet\\nLangSmith\\nLangSmith Cookbooks repo\\nLangChain Retrieval blog\\nEvaluating + CSV Question/Answering blog and YouTube\\nMultiOn Partner blog\\nHarvard Sports + Analytics Collective\\nEvaluating RAG Webinar\\nawesome-langchain:\\nLLM Math + Chain\\nSelf-Ask\\nLangChain Hub UI\\n\\u201cLangChain is Pointless\\u201d\\nHarrison\\u2019s + links\\nsports - estimating player compatibility in the NBA\\nearly interest + in prompt injections\\nGitHub\\nTwitter\\nTimestamps\\n[00:00:00] Introduction\\n[00:00:48] + Harrison''s background and how sports led him into ML\\n[00:04:54] Memory (for + long running chats, whether a simple Buffer, Knowledge Graph, Summary, or Vector + Store)\\nUse-Cases: compositions of Components\\nChains: combining a PromptTemplate, + LLM Model and optional OutputParser\\nwith Router, Sequential, and Transform + Chains for advanced usecases\\nsavable, sharable schemas that can be loaded + from LangChainHub\\nAgents: a chain that has access to a suite of tools, of + nondeterministic8 length because the LLM is used as a reasoning engine to determine + which actions to take and in which order.\"}, {\"url\": \"https://www.langchain.com/\", + \"content\": \"Join Discord\\nContribute on GitHub\\n5.3M\\nMonthly Downloads\\n67k\\nGithub + Stars\\n1800+\\nContributors\\n34k+\\nApps Powered\\nReady to build?\\nGo To + Docs\\nGo To Docs\\nGo To Docs\\nContact Sales\\nContact Sales\\nContact Sales\\nProducts + and use-cases\\nLangChain\\nLangSmith\\nRetrieval\\nAgents\\nInspiration\\nCode\\nGitHub\\nLangChain + Hub\\nPython Docs\\nJS/TS Docs\\nSocial\\nTwitter\\nDiscord\\nBlog\\nLinkedIn\\nYouTube\\nTerms + of Service\\nSign up for our newsletter\\nProducts and use-cases\\nLangChain\\nLangSmith\\nRetrieval\\nAgents\\nInspiration\\nCode\\nGitHub\\nLangChain + Hub\\nPython Docs\\nJS/TS Docs\\nSocial\\nTwitter\\nDiscord\\nBlog\\nLinkedIn\\nYouTube\\nTerms + of Service\\nSign up for our newsletter\\nProducts and use-cases\\nLangChain\\nLangSmith\\nRetrieval\\nAgents\\nInspiration\\nCode\\nGitHub\\nLangChain + Hub\\nPython Docs\\nJS/TS Docs\\nSocial\\nTwitter\\nDiscord\\nBlog\\nLinkedIn\\nYouTube\\nTerms + of Service\\nSign up for our newsletter Chatbots\\nQ&A Over Docs\\nSummarization\\nCopilots\\nWorkflow + Automation\\nDocument Analysis\\nCustom Search\\nHarness the power\\u2013and + wrangle the complexity\\u2013 of LLMs\\nGet from proof-of-concept to production + on one platform\\nRely on a set of tools built especially for non-deterministic + models\\nLeverage new cognitive architectures and battle-tested orchestration\\nThanks + to our contributors and partners for making LangChain awesome\\nWhen you build + with LangChain, you build alongside a community of developers advancing the + frontier of what\\u2019s possible with LLMs.\\n \\ud83e\\udd9c\\ud83d\\udd17 + LangChain\\nLangSmith\\nLangServe\\nAgents\\nRetrieval\\nEvaluation\\nBlog\\nDocs\\n\\ud83e\\udd9c\\ud83d\\udd17 + LangChain\\n\\ud83e\\udd9c\\ud83d\\udd17 LangChain\\nLangChain Expression Language + makes it easy to create custom chains.\\n Go To Docs\\nGo To Docs\\nGo To Docs\\nContact + Sales\\nglobal corporations, STARTUPS, and TINKERERS build with LangChain\\nglobal + corporations, STARTUPS, and TINKERERS build with LangChain\\nglobal corporations, + STARTUPS, and TINKERERS build with LangChain\\nA complete set of powerful building + blocks\\nA complete set of powerful building blocks\\n Try it out here\\nGet + your LLM application from prototype to production\\nGet your LLM application + from prototype to production\\nBuild context-aware, reasoning applications with + LangChain\\u2019s flexible abstractions and AI-first toolkit.\\n\"}]", "tool_use_id": + "toolu_01JtnSAfQidha7JoYCXqUYzc", "is_error": false}]}, {"role": "assistant", + "content": [{"type": "tool_use", "name": "scrape_website", "input": {"url": + "https://www.langchain.com/"}, "id": "toolu_018RqnUb3x65zWF9PSDzA2Ns"}]}, {"role": + "user", "content": [{"type": "tool_result", "content": "Here are some key notes + about the LangChain website based on the content:\n\n1. LangChain offers a suite + of products for building, running, and managing LLM applications:\n - LangChain: + A framework for building with LLMs\n - LangGraph: A framework for building + controllable agentic workflows\n - LangSmith: A platform for debugging, testing, + deploying and monitoring LLM apps\n\n2. The website highlights that LangChain + products can be used independently or together.\n\n3. Key stats mentioned:\n - + 15M+ monthly downloads\n - 100K+ apps powered\n - 75K+ GitHub stars\n - + 3K+ contributors\n\n4. The site emphasizes LangChain''s large developer community + (1M+ developers).\n\n5. There are testimonials from companies like Retool, Rakuten, + and Elastic about using LangChain products.\n\n6. The website offers resources + like documentation, GitHub links, blog, and community pages.\n\n7. There''s + a newsletter signup form in the footer.\n\n8. The site promotes both free signup + and enterprise demos.\n\n9. Use cases across industries like financial services, + fintech, and technology are highlighted.\n\n10. The website design uses a lot + of green coloring, matching LangChain''s branding.", "tool_use_id": "toolu_018RqnUb3x65zWF9PSDzA2Ns", + "is_error": false}]}], "model": "claude-3-5-sonnet-20240620", "tool_choice": + {"type": "any"}, "tools": [{"name": "scrape_website", "description": "Scrape + and summarize content from a given URL.", "input_schema": {"type": "object", + "properties": {"url": {"type": "string"}}, "required": ["url"]}}, {"name": "search", + "description": "Search for general results. This function performs a search + using the Tavily search engine, which is designed\nto provide comprehensive, + accurate, and trusted results. It''s particularly useful\nfor answering questions + about current events.", "input_schema": {"type": "object", "properties": {"query": + {"type": "string"}}, "required": ["query"]}}, {"name": "Info", "description": + "Call this when you have gathered all the relevant info", "input_schema": {"type": + "object", "properties": {"founder": {"type": "string", "description": "The name + of the company founder."}, "websiteUrl": {"type": "string", "description": "Website + URL of the company, e.g.: https://openai.com/, or https://microsoft.com"}, "products_sold": + {"type": "array", "items": {"type": "string"}, "description": "A list of products + sold by the company."}}, "required": ["founder", "websiteUrl", "products_sold"]}}]}' + headers: {} + method: POST + uri: https://api.anthropic.com/v1/messages + response: + body: + string: !!binary | + H4sIAAAAAAAAA1RQ0UrDQBD8FdnnS5vGVuy9aSlG24JgtQ+lhEvumgQve/H2jlRC/l0uKujb7OzM + MLs91BI4NFRm8Wzx8vZ0kYc277b3u7pMVoJKImDgPlsVVIpIlAoYWKMDIYhqcgIdMGiMVBo4FFp4 + qaLraBGRQVQuSuJkHt8kMTAoDDqFDvix/w11xujMU0gdq4TZZ/Hs0O7yVO736/Ri73JK5eY5WW+A + AYom+B7xbIIHW++A93A2HqWywCEV1tZk8GpViTG3UznVTr3a0K9yriU+nXZdN9ECy6ISNU4K00yB + QWuN9IWjjIyWwI+wFViuggLYiB+saKsf/NLUroLTMJwYkDNtZpUgg/9vGhekPrzCQgFHrzUDP76R + 99/tM2feFRLw+fw2ZmC8+0sul8PwBQAA//8DADLSBIWmAQAA + headers: + CF-Cache-Status: + - DYNAMIC + CF-RAY: + - 8c4af83c19ca9e50-SJC + Connection: + - keep-alive + Content-Encoding: + - gzip + Content-Type: + - application/json + Date: + - Tue, 17 Sep 2024 17:55:54 GMT + Server: + - cloudflare + Transfer-Encoding: + - chunked + X-Robots-Tag: + - none + anthropic-ratelimit-requests-limit: + - '4000' + anthropic-ratelimit-requests-remaining: + - '3999' + anthropic-ratelimit-requests-reset: + - '2024-09-17T17:56:54Z' + anthropic-ratelimit-tokens-limit: + - '400000' + anthropic-ratelimit-tokens-remaining: + - '400000' + anthropic-ratelimit-tokens-reset: + - '2024-09-17T17:55:54Z' + request-id: + - req_01MdGZEXErzbkdgwN1gdHZ3f + via: + - 1.1 google + x-cloud-trace-context: + - 609e1261acac28312845cf27b91fb6e0 + status: + code: 200 + message: OK +- request: + body: '{"max_tokens": 1024, "messages": [{"role": "user", "content": "You are + doing web research on behalf of a user. You are trying to figure out this information:\n\n\n{\n \"type\": + \"object\",\n \"properties\": {\n \"founder\": {\n \"type\": \"string\",\n \"description\": + \"The name of the company founder.\"\n },\n \"websiteUrl\": {\n \"type\": + \"string\",\n \"description\": \"Website URL of the company, e.g.: https://openai.com/, + or https://microsoft.com\"\n },\n \"products_sold\": {\n \"type\": + \"array\",\n \"items\": {\n \"type\": \"string\"\n },\n \"description\": + \"A list of products sold by the company.\"\n }\n },\n \"required\": [\n \"founder\",\n \"websiteUrl\",\n \"products_sold\"\n ]\n}\n\n\nYou + have access to the following tools:\n\n- `Search`: call a search tool and get + back some results\n- `ScrapeWebsite`: scrape a website and get relevant notes + about the given request. This will update the notes above.\n- `Info`: call this + when you are done and have gathered all the relevant info\n\nHere is the information + you have about the topic you are researching:\n\nTopic: LangChain"}, {"role": + "assistant", "content": [{"type": "tool_use", "name": "search", "input": {"query": + "LangChain company founder, website, and products"}, "id": "toolu_01JtnSAfQidha7JoYCXqUYzc"}]}, + {"role": "user", "content": [{"type": "tool_result", "content": "[{\"url\": + \"https://www.cbinsights.com/company/langchain\", \"content\": \"It was founded + in 2023 and is based in San Francisco, California. LangChain specializes in + the development of large language model (LLM) applications, providing a suite + of products that support developers throughout the application lifecycle. Use + the CB Insights Platform to explore LangChain''s full profile.\"}, {\"url\": + \"https://en.wikipedia.org/wiki/LangChain\", \"content\": \"In October 2023 + LangChain introduced LangServe, a deployment tool designed to facilitate the + transition from LCEL (LangChain Expression Language) prototypes to production-ready + applications.[5]\\nIntegrations[edit]\\nAs of March 2023, LangChain included + integrations with systems including Amazon, Google, and Microsoft Azure cloud + storage; API wrappers for news, movie information, and weather; Bash for summarization, + syntax and semantics checking, and execution of shell scripts; multiple web + scraping subsystems and templates; few-shot learning prompt generation support; + finding and summarizing \\\"todo\\\" tasks in code; Google Drive documents, + spreadsheets, and presentations summarization, extraction, and creation; Google + Search and Microsoft Bing web search; OpenAI, Anthropic, and Hugging Face language + models; iFixit repair guides and wikis search and summarization; MapReduce for + question answering, combining documents, and question generation; N-gram overlap + scoring; PyPDF, pdfminer, fitz, and pymupdf for PDF file text extraction and + manipulation; Python and JavaScript code generation, analysis, and debugging; + Milvus vector database[6] to store and retrieve vector embeddings; Weaviate + vector database[7] to cache embedding and data objects; Redis cache database + storage; Python RequestsWrapper and other methods for API requests; SQL and + NoSQL databases including JSON support; Streamlit, including for logging; text + mapping for k-nearest neighbors search; time zone conversion and calendar operations; + tracing and recording stack symbols in threaded and asynchronous subprocess + runs; and the Wolfram Alpha website and SDK.[8] As a language model integration + framework, LangChain''s use-cases largely overlap with those of language models + in general, including document analysis and summarization, chatbots, and code + analysis.[2]\\nHistory[edit]\\nLangChain was launched in October 2022 as an + open source project by Harrison Chase, while working at machine learning startup + Robust Intelligence. In April 2023, LangChain had incorporated and the new startup + raised over $20 million in funding at a valuation of at least $200 million from + venture firm Sequoia Capital, a week after announcing a $10 million seed investment + from Benchmark.[3][4]\\n The project quickly garnered popularity, with improvements + from hundreds of contributors on GitHub, trending discussions on Twitter, lively + activity on the project''s Discord server, many YouTube tutorials, and meetups + in San Francisco and London. As of April 2023, it can read from more than 50 + document types and data sources.[9]\\nReferences[edit]\\nExternal links[edit]\"}, + {\"url\": \"https://www.forbes.com/companies/langchain/\", \"content\": \"LangChain. + Founded in late 2022, LangChain was an instant hit in Silicon Valley for its + software that enables anyone to build apps powered by large language models + like GPT-4 in as little as two ...\"}, {\"url\": \"https://tedai-sanfrancisco.ted.com/speakers-1/harrison-chase-/co-founder-and-ceo,-langchain/\", + \"content\": \"Harrison Chase, a Harvard graduate in statistics and computer + science, co-founded LangChain to streamline the development of Language Model + applications with open-source Python/Typescript packages. Chase''s experience + includes heading the Machine Learning team at Robust Intelligence, focusing + on the testing and validation of machine learning models, and leading the entity + linking team at Kensho ...\"}, {\"url\": \"https://www.crunchbase.com/organization/langchain\", + \"content\": \"Summary. LangChain may be growing as evidenced by securing $25 + million in funding, which is a significant financial milestone for the company. + Additionally, the launch of a new platform to facilitate the full lifecycle + of Large Language Model applications suggests expansion in their product offerings. + The integration of Gemini Pro API, enabling ...\"}, {\"url\": \"https://craft.co/langchain\", + \"content\": \"LangChain has 5 employees at their 1 location and $35 m in total + funding,. See insights on LangChain including office locations, competitors, + revenue, financials, executives, subsidiaries and more at Craft. ... Harrison + Chase, Co-Founder and Chief Executive Officer. Linkedin;\"}, {\"url\": \"https://www.langchain.com/about\", + \"content\": \"Ready to start shipping\\nreliable GenAI apps faster?\\nLangChain + and LangSmith are critical parts of the reference\\narchitecture to get you + from prototype to production. Announcing the General Availability of LangSmith\\nToday, + we\\u2019re thrilled to announce the general availability of LangSmith \\u2014 + our solution for LLM application development, monitoring, and testing.\\n We + build products that enable developers to go from an idea to working code in + an afternoon and in the hands of users in days or weeks. The biggest developer + community in \\u00a0GenAI\\nLearn alongside the 100K+ practitioners\\nwho are + pushing the industry forward.\\n And we built LangSmith to support all stages + of the AI engineering lifecycle, to get applications into production faster.\\n\"}, + {\"url\": \"https://www.bluelabellabs.com/blog/what-is-langchain/\", \"content\": + \"Similar setups to AutoTweeter would allow businesses to parse past and present + customer queries to understand drop-off points and build the appropriate flows + to solve customer problems on the spot, instead of spitting them into a queue + to wait for a human and plot how they\\u2019re going to steal from your business.\\n + Rather than creating a bot to tweet in your tone on your behalf about your love + for processed snacks, hockey, and political happenings, you could train such + a system on your brand voice, knowledge bases, and other resources to create + a highly valuable feature for your digital touchpoints.\\n Here, you essentially + create a curated ML system that\\u2019s tailored to your brand with specific + issues and requests from your customers \\u2013 with a little bit of dev magic, + you can quickly build a solution that integrates features like payment systems, + tech support, ordering, account management, and so much more.\\n Creating a + bot for social media that can post just like you\\nSocial media already has + enough bots engaging with people as if they\\u2019re real people, but for the + sake of making a functional example, our co-founder Bobby created a bot called + AutoTweeter for the social media platform we all still call Twitter.\\n How + does LangChain work?\\nLangchain essentially serves as a middleman between other + LLMs, data sources, and API-accessible functions \\u2013 more precisely, it\\u2019s + a comprehensive open-source library of components that allow users to build + better (i.e., more custom) solutions by interconnecting different resources.\\n\"}, + {\"url\": \"https://www.latent.space/p/langchain\", \"content\": \"And so you + can do a lot of things like reflection and kind of like making sure that you''re + on the right track and really put more safeguards and thinking about these agents + as opposed to relying on like chas and interface, like the bot we have that''s + on GitHub answering questions on the issues, I think probably gives better answers + than the bots that we have that are on chat on the website. [00:50:32]\\nHarrison: + And then between kind of like allowing for, I think one helpful thing that we + did in terms of like distinguishing between basically the base interfaces and + then more complex stuff is part of the separation around the docs is there''s + like the components piece, which has the model IO, the retrieval, the agents, + the callbacks, things like that. And then like, I know a lot of people are doing + this, but like a personal assistant for like managing like email calendar, basic + stuff, which I think is, I think that''s like a fantastic application for these + like agent like things, because if you think about personal assistants today, + you usually interact, I don''t have one, but I''m told you interact with them + over email. Show Notes\\nLangChain\\nLangChain\\u2019s Berkshire Hathaway Homepage\\nAbstractions + tweet\\nLangSmith\\nLangSmith Cookbooks repo\\nLangChain Retrieval blog\\nEvaluating + CSV Question/Answering blog and YouTube\\nMultiOn Partner blog\\nHarvard Sports + Analytics Collective\\nEvaluating RAG Webinar\\nawesome-langchain:\\nLLM Math + Chain\\nSelf-Ask\\nLangChain Hub UI\\n\\u201cLangChain is Pointless\\u201d\\nHarrison\\u2019s + links\\nsports - estimating player compatibility in the NBA\\nearly interest + in prompt injections\\nGitHub\\nTwitter\\nTimestamps\\n[00:00:00] Introduction\\n[00:00:48] + Harrison''s background and how sports led him into ML\\n[00:04:54] Memory (for + long running chats, whether a simple Buffer, Knowledge Graph, Summary, or Vector + Store)\\nUse-Cases: compositions of Components\\nChains: combining a PromptTemplate, + LLM Model and optional OutputParser\\nwith Router, Sequential, and Transform + Chains for advanced usecases\\nsavable, sharable schemas that can be loaded + from LangChainHub\\nAgents: a chain that has access to a suite of tools, of + nondeterministic8 length because the LLM is used as a reasoning engine to determine + which actions to take and in which order.\"}, {\"url\": \"https://www.langchain.com/\", + \"content\": \"Join Discord\\nContribute on GitHub\\n5.3M\\nMonthly Downloads\\n67k\\nGithub + Stars\\n1800+\\nContributors\\n34k+\\nApps Powered\\nReady to build?\\nGo To + Docs\\nGo To Docs\\nGo To Docs\\nContact Sales\\nContact Sales\\nContact Sales\\nProducts + and use-cases\\nLangChain\\nLangSmith\\nRetrieval\\nAgents\\nInspiration\\nCode\\nGitHub\\nLangChain + Hub\\nPython Docs\\nJS/TS Docs\\nSocial\\nTwitter\\nDiscord\\nBlog\\nLinkedIn\\nYouTube\\nTerms + of Service\\nSign up for our newsletter\\nProducts and use-cases\\nLangChain\\nLangSmith\\nRetrieval\\nAgents\\nInspiration\\nCode\\nGitHub\\nLangChain + Hub\\nPython Docs\\nJS/TS Docs\\nSocial\\nTwitter\\nDiscord\\nBlog\\nLinkedIn\\nYouTube\\nTerms + of Service\\nSign up for our newsletter\\nProducts and use-cases\\nLangChain\\nLangSmith\\nRetrieval\\nAgents\\nInspiration\\nCode\\nGitHub\\nLangChain + Hub\\nPython Docs\\nJS/TS Docs\\nSocial\\nTwitter\\nDiscord\\nBlog\\nLinkedIn\\nYouTube\\nTerms + of Service\\nSign up for our newsletter Chatbots\\nQ&A Over Docs\\nSummarization\\nCopilots\\nWorkflow + Automation\\nDocument Analysis\\nCustom Search\\nHarness the power\\u2013and + wrangle the complexity\\u2013 of LLMs\\nGet from proof-of-concept to production + on one platform\\nRely on a set of tools built especially for non-deterministic + models\\nLeverage new cognitive architectures and battle-tested orchestration\\nThanks + to our contributors and partners for making LangChain awesome\\nWhen you build + with LangChain, you build alongside a community of developers advancing the + frontier of what\\u2019s possible with LLMs.\\n \\ud83e\\udd9c\\ud83d\\udd17 + LangChain\\nLangSmith\\nLangServe\\nAgents\\nRetrieval\\nEvaluation\\nBlog\\nDocs\\n\\ud83e\\udd9c\\ud83d\\udd17 + LangChain\\n\\ud83e\\udd9c\\ud83d\\udd17 LangChain\\nLangChain Expression Language + makes it easy to create custom chains.\\n Go To Docs\\nGo To Docs\\nGo To Docs\\nContact + Sales\\nglobal corporations, STARTUPS, and TINKERERS build with LangChain\\nglobal + corporations, STARTUPS, and TINKERERS build with LangChain\\nglobal corporations, + STARTUPS, and TINKERERS build with LangChain\\nA complete set of powerful building + blocks\\nA complete set of powerful building blocks\\n Try it out here\\nGet + your LLM application from prototype to production\\nGet your LLM application + from prototype to production\\nBuild context-aware, reasoning applications with + LangChain\\u2019s flexible abstractions and AI-first toolkit.\\n\"}]", "tool_use_id": + "toolu_01JtnSAfQidha7JoYCXqUYzc", "is_error": false}]}, {"role": "assistant", + "content": [{"type": "tool_use", "name": "scrape_website", "input": {"url": + "https://www.langchain.com/"}, "id": "toolu_018RqnUb3x65zWF9PSDzA2Ns"}]}, {"role": + "user", "content": [{"type": "tool_result", "content": "Here are some key notes + about the LangChain website based on the content:\n\n1. LangChain offers a suite + of products for building, running, and managing LLM applications:\n - LangChain: + A framework for building with LLMs\n - LangGraph: A framework for building + controllable agentic workflows\n - LangSmith: A platform for debugging, testing, + deploying and monitoring LLM apps\n\n2. The website highlights that LangChain + products can be used independently or together.\n\n3. Key stats mentioned:\n - + 15M+ monthly downloads\n - 100K+ apps powered\n - 75K+ GitHub stars\n - + 3K+ contributors\n\n4. The site emphasizes LangChain''s large developer community + (1M+ developers).\n\n5. There are testimonials from companies like Retool, Rakuten, + and Elastic about using LangChain products.\n\n6. The website offers resources + like documentation, GitHub links, blog, and community pages.\n\n7. There''s + a newsletter signup form in the footer.\n\n8. The site promotes both free signup + and enterprise demos.\n\n9. Use cases across industries like financial services, + fintech, and technology are highlighted.\n\n10. The website design uses a lot + of green coloring, matching LangChain''s branding.", "tool_use_id": "toolu_018RqnUb3x65zWF9PSDzA2Ns", + "is_error": false}, {"type": "text", "text": "I am thinking of calling the info + tool with the info below. Is this good? Give your reasoning as well. You can + encourage the Assistant to look at specific URLs if that seems relevant, or + do more searches.\nIf you don''t think it is good, you should be very specific + about what could be improved.\n\n{\n \"founder\": \"Harrison Chase\",\n \"websiteUrl\": + \"https://www.langchain.com/\",\n \"products_sold\": [\n \"LangChain\",\n \"LangGraph\",\n \"LangSmith\"\n ]\n}"}]}], + "model": "claude-3-5-sonnet-20240620", "tool_choice": {"type": "tool", "name": + "InfoIsSatisfactory"}, "tools": [{"name": "InfoIsSatisfactory", "description": + "Validate whether the current extracted info is satisfactory and complete.", + "input_schema": {"properties": {"reason": {"description": "First, provide reasoning + for why this is either good or bad as a final result. Must include at least + 3 reasons.", "items": {"type": "string"}, "type": "array"}, "is_satisfactory": + {"description": "After providing your reasoning, provide a value indicating + whether the result is satisfactory. If not, you will continue researching.", + "type": "boolean"}}, "required": ["reason", "is_satisfactory"], "type": "object"}}]}' + headers: {} + method: POST + uri: https://api.anthropic.com/v1/messages + response: + body: + string: !!binary | + H4sIAAAAAAAAAwAAAP//ZJNhbxQ5DIb/ipUvB1LabodeEfO1Uq/oeujUFgRCaOVNPDs+MvEQe7pa + Vf3vKDNdKNy3ySR+/Np+/eA4utYNul2vTu8+dvjx+vbqP/3nrgnXpxO9+jrtnXe2H6m+IlXckvOu + SKo/UJXVMJvzbpBIybUuJJwiHb06+vNIJWeyo2bVnK3Om5XzLkg2yubazw8HqImk9aSVOkup52m9 + Ov27ubm5uvxwtjt/d/etuWxefxr+fbd13mUcatzb3MlbvUVj7TCYlCqU8ziZax9cIVTJrv3s7noC + zp2UAY0lw1jkniNFCHJPRQFTAusJCn2buFCEjilFhRedTDlS8bCjjbLR+5I8YI6VEKdgulZJ8SWg + go4UuGOKwHmGSeEtZ0wzldSOnZ+FPDH/UKhVeLjCUlglw0WPSh5YIUgpFCztgSNlW7AbVIogC1wJ + S+ihkE7J1MOIxThMCUvaQ1dkmF9dStmQzoIvypRDXxlA2QqTHvQ8lQbvb67hRW82antystvtjhPm + beiR83GQ4eRlFYYhTAWNZuSAFnrSpdiu48CY4Brz9qIG/eBajwa72qFQcKR4yPtLCyGxGnAOaYqk + Pyl+/vyr4Ngvja/H24Gt97DrOfSApc42EsVZyFBTH9BVFtV5bvbzZZBhxLyvZUiJnLdgsrRzkfZD + 85NHq9Qr2dF9tYD1VAgG3vYGG4IiMkAnBXiodqKBsh1G///SfLXIrJ3KPc3tqAEseTGMyvD7UJce + H9IFycpxLgYzYIxcgzEdch27L96xrvX5MrQdJqXHxy/eqcm4PmzE84WbL7RaNAdybZ5S8m6ad7x9 + WJZpbfKVsrr27Oz8tXcy2fOfTfPm8fE7AAAA//8DACQdP65EBAAA + headers: + CF-Cache-Status: + - DYNAMIC + CF-RAY: + - 8c4af84d7c57251c-SJC + Connection: + - keep-alive + Content-Encoding: + - gzip + Content-Type: + - application/json + Date: + - Tue, 17 Sep 2024 17:55:59 GMT + Server: + - cloudflare + Transfer-Encoding: + - chunked + X-Robots-Tag: + - none + anthropic-ratelimit-requests-limit: + - '4000' + anthropic-ratelimit-requests-remaining: + - '3999' + anthropic-ratelimit-requests-reset: + - '2024-09-17T17:56:54Z' + anthropic-ratelimit-tokens-limit: + - '400000' + anthropic-ratelimit-tokens-remaining: + - '400000' + anthropic-ratelimit-tokens-reset: + - '2024-09-17T17:55:59Z' + request-id: + - req_01YNF4M4GMfi6ycDLHzK62ey + via: + - 1.1 google + x-cloud-trace-context: + - 16bf9f43b729a79b26fb35cc24c5dc38 + status: + code: 200 + message: OK +- request: + body: '{"max_tokens": 1024, "messages": [{"role": "user", "content": "You are + doing web research on behalf of a user. You are trying to figure out this information:\n\n\n{\n \"type\": + \"object\",\n \"properties\": {\n \"providers\": {\n \"type\": \"array\",\n \"items\": + {\n \"type\": \"object\",\n \"properties\": {\n \"name\": + {\n \"type\": \"string\",\n \"description\": \"Company + name\"\n },\n \"technology_summary\": {\n \"type\": + \"string\",\n \"description\": \"Brief summary of their chip technology + for LLM training\"\n },\n \"current_market_share\": {\n \"type\": + \"string\",\n \"description\": \"Estimated current market share percentage + or position\"\n },\n \"future_outlook\": {\n \"type\": + \"string\",\n \"description\": \"Brief paragraph on future prospects + and developments\"\n }\n },\n \"required\": [\n \"name\",\n \"technology_summary\",\n \"current_market_share\",\n \"future_outlook\"\n ]\n },\n \"description\": + \"List of top chip providers for LLM Training\"\n },\n \"overall_market_trends\": + {\n \"type\": \"string\",\n \"description\": \"Brief paragraph on + general trends in the LLM chip market\"\n }\n },\n \"required\": [\n \"providers\",\n \"overall_market_trends\"\n ]\n}\n\n\nYou + have access to the following tools:\n\n- `Search`: call a search tool and get + back some results\n- `ScrapeWebsite`: scrape a website and get relevant notes + about the given request. This will update the notes above.\n- `Info`: call this + when you are done and have gathered all the relevant info\n\nHere is the information + you have about the topic you are researching:\n\nTopic: Top 5 chip providers + for LLM training"}], "model": "claude-3-5-sonnet-20240620", "tool_choice": {"type": + "any"}, "tools": [{"name": "scrape_website", "description": "Scrape and summarize + content from a given URL.", "input_schema": {"type": "object", "properties": + {"url": {"type": "string"}}, "required": ["url"]}}, {"name": "search", "description": + "Search for general results. This function performs a search using the Tavily + search engine, which is designed\nto provide comprehensive, accurate, and trusted + results. It''s particularly useful\nfor answering questions about current events.", + "input_schema": {"type": "object", "properties": {"query": {"type": "string"}}, + "required": ["query"]}}, {"name": "Info", "description": "Call this when you + have gathered all the relevant info", "input_schema": {"type": "object", "properties": + {"providers": {"type": "array", "items": {"type": "object", "properties": {"name": + {"type": "string", "description": "Company name"}, "technology_summary": {"type": + "string", "description": "Brief summary of their chip technology for LLM training"}, + "current_market_share": {"type": "string", "description": "Estimated current + market share percentage or position"}, "future_outlook": {"type": "string", + "description": "Brief paragraph on future prospects and developments"}}, "required": + ["name", "technology_summary", "current_market_share", "future_outlook"]}, "description": + "List of top chip providers for LLM Training"}, "overall_market_trends": {"type": + "string", "description": "Brief paragraph on general trends in the LLM chip + market"}}, "required": ["providers", "overall_market_trends"]}}]}' + headers: {} + method: POST + uri: https://api.anthropic.com/v1/messages + response: + body: + string: !!binary | + H4sIAAAAAAAAA1SQUUvDQBCE/8qxzxdI2qa296ZYEKwgtFhRSjiSbRKb7Ka3d4UQ8t8lFUEfZ2Zn + +NgB6gIMtFJmcfJ6eHi+D+WFvpLdZjX/eHeLu9UaNPi+w+kKRWyJoMFxMxlWpBZvyYOGlgtswEDe + 2FBgNI/SSJgIfTSLZ4t4OYtBQ87kkTyYz+F31DM3WZBp9YYy6ZDFyaE89ZsWffu2k/062V6f0sfQ + gQay7dQTtC6vphZ1wYMZ4BLQ9WBgz51KVV7VneocX+sCnagTO7XdvijvbE01laq17oxeSWUdKkuF + 8phXxA2XPYzjUYN47jKHVpj+Y94CwUtAyhEMhabREG6fMcMPTub5jCRgkiRZauDg/5ppPI7fAAAA + //8DAJ9/8BZ5AQAA + headers: + CF-Cache-Status: + - DYNAMIC + CF-RAY: + - 8c4b1ff70a182368-SJC + Connection: + - keep-alive + Content-Encoding: + - gzip + Content-Type: + - application/json + Date: + - Tue, 17 Sep 2024 18:23:01 GMT + Server: + - cloudflare + Transfer-Encoding: + - chunked + X-Robots-Tag: + - none + anthropic-ratelimit-requests-limit: + - '4000' + anthropic-ratelimit-requests-remaining: + - '3999' + anthropic-ratelimit-requests-reset: + - '2024-09-17T18:23:54Z' + anthropic-ratelimit-tokens-limit: + - '400000' + anthropic-ratelimit-tokens-remaining: + - '400000' + anthropic-ratelimit-tokens-reset: + - '2024-09-17T18:23:01Z' + request-id: + - req_01CMZwM2zGhoHhfBYEQ6ThJM + via: + - 1.1 google + x-cloud-trace-context: + - cedecb2fdef29557856f15dda0f5c1f6 + status: + code: 200 + message: OK +- request: + body: '{"max_tokens": 1024, "messages": [{"role": "user", "content": "You are + doing web research on behalf of a user. You are trying to figure out this information:\n\n\n{\n \"type\": + \"object\",\n \"properties\": {\n \"providers\": {\n \"type\": \"array\",\n \"items\": + {\n \"type\": \"object\",\n \"properties\": {\n \"name\": + {\n \"type\": \"string\",\n \"description\": \"Company + name\"\n },\n \"technology_summary\": {\n \"type\": + \"string\",\n \"description\": \"Brief summary of their chip technology + for LLM training\"\n },\n \"current_market_share\": {\n \"type\": + \"string\",\n \"description\": \"Estimated current market share percentage + or position\"\n },\n \"future_outlook\": {\n \"type\": + \"string\",\n \"description\": \"Brief paragraph on future prospects + and developments\"\n }\n },\n \"required\": [\n \"name\",\n \"technology_summary\",\n \"current_market_share\",\n \"future_outlook\"\n ]\n },\n \"description\": + \"List of top chip providers for LLM Training\"\n },\n \"overall_market_trends\": + {\n \"type\": \"string\",\n \"description\": \"Brief paragraph on + general trends in the LLM chip market\"\n }\n },\n \"required\": [\n \"providers\",\n \"overall_market_trends\"\n ]\n}\n\n\nYou + have access to the following tools:\n\n- `Search`: call a search tool and get + back some results\n- `ScrapeWebsite`: scrape a website and get relevant notes + about the given request. This will update the notes above.\n- `Info`: call this + when you are done and have gathered all the relevant info\n\nHere is the information + you have about the topic you are researching:\n\nTopic: Top 5 chip providers + for LLM training"}, {"role": "assistant", "content": [{"type": "tool_use", "name": + "search", "input": {"query": "Top 5 chip providers for LLM training market share + and technology"}, "id": "toolu_01WgfyEmetmVSsT91LvH5Dup"}]}, {"role": "user", + "content": [{"type": "tool_result", "content": "[{\"url\": \"https://www.cbinsights.com/company/langchain\", + \"content\": \"It was founded in 2023 and is based in San Francisco, California. + LangChain specializes in the development of large language model (LLM) applications, + providing a suite of products that support developers throughout the application + lifecycle. Use the CB Insights Platform to explore LangChain''s full profile.\"}, + {\"url\": \"https://en.wikipedia.org/wiki/LangChain\", \"content\": \"In October + 2023 LangChain introduced LangServe, a deployment tool designed to facilitate + the transition from LCEL (LangChain Expression Language) prototypes to production-ready + applications.[5]\\nIntegrations[edit]\\nAs of March 2023, LangChain included + integrations with systems including Amazon, Google, and Microsoft Azure cloud + storage; API wrappers for news, movie information, and weather; Bash for summarization, + syntax and semantics checking, and execution of shell scripts; multiple web + scraping subsystems and templates; few-shot learning prompt generation support; + finding and summarizing \\\"todo\\\" tasks in code; Google Drive documents, + spreadsheets, and presentations summarization, extraction, and creation; Google + Search and Microsoft Bing web search; OpenAI, Anthropic, and Hugging Face language + models; iFixit repair guides and wikis search and summarization; MapReduce for + question answering, combining documents, and question generation; N-gram overlap + scoring; PyPDF, pdfminer, fitz, and pymupdf for PDF file text extraction and + manipulation; Python and JavaScript code generation, analysis, and debugging; + Milvus vector database[6] to store and retrieve vector embeddings; Weaviate + vector database[7] to cache embedding and data objects; Redis cache database + storage; Python RequestsWrapper and other methods for API requests; SQL and + NoSQL databases including JSON support; Streamlit, including for logging; text + mapping for k-nearest neighbors search; time zone conversion and calendar operations; + tracing and recording stack symbols in threaded and asynchronous subprocess + runs; and the Wolfram Alpha website and SDK.[8] As a language model integration + framework, LangChain''s use-cases largely overlap with those of language models + in general, including document analysis and summarization, chatbots, and code + analysis.[2]\\nHistory[edit]\\nLangChain was launched in October 2022 as an + open source project by Harrison Chase, while working at machine learning startup + Robust Intelligence. In April 2023, LangChain had incorporated and the new startup + raised over $20 million in funding at a valuation of at least $200 million from + venture firm Sequoia Capital, a week after announcing a $10 million seed investment + from Benchmark.[3][4]\\n The project quickly garnered popularity, with improvements + from hundreds of contributors on GitHub, trending discussions on Twitter, lively + activity on the project''s Discord server, many YouTube tutorials, and meetups + in San Francisco and London. As of April 2023, it can read from more than 50 + document types and data sources.[9]\\nReferences[edit]\\nExternal links[edit]\"}, + {\"url\": \"https://www.forbes.com/companies/langchain/\", \"content\": \"LangChain. + Founded in late 2022, LangChain was an instant hit in Silicon Valley for its + software that enables anyone to build apps powered by large language models + like GPT-4 in as little as two ...\"}, {\"url\": \"https://tedai-sanfrancisco.ted.com/speakers-1/harrison-chase-/co-founder-and-ceo,-langchain/\", + \"content\": \"Harrison Chase, a Harvard graduate in statistics and computer + science, co-founded LangChain to streamline the development of Language Model + applications with open-source Python/Typescript packages. Chase''s experience + includes heading the Machine Learning team at Robust Intelligence, focusing + on the testing and validation of machine learning models, and leading the entity + linking team at Kensho ...\"}, {\"url\": \"https://www.crunchbase.com/organization/langchain\", + \"content\": \"Summary. LangChain may be growing as evidenced by securing $25 + million in funding, which is a significant financial milestone for the company. + Additionally, the launch of a new platform to facilitate the full lifecycle + of Large Language Model applications suggests expansion in their product offerings. + The integration of Gemini Pro API, enabling ...\"}, {\"url\": \"https://craft.co/langchain\", + \"content\": \"LangChain has 5 employees at their 1 location and $35 m in total + funding,. See insights on LangChain including office locations, competitors, + revenue, financials, executives, subsidiaries and more at Craft. ... Harrison + Chase, Co-Founder and Chief Executive Officer. Linkedin;\"}, {\"url\": \"https://www.langchain.com/about\", + \"content\": \"Ready to start shipping\\nreliable GenAI apps faster?\\nLangChain + and LangSmith are critical parts of the reference\\narchitecture to get you + from prototype to production. Announcing the General Availability of LangSmith\\nToday, + we\\u2019re thrilled to announce the general availability of LangSmith \\u2014 + our solution for LLM application development, monitoring, and testing.\\n We + build products that enable developers to go from an idea to working code in + an afternoon and in the hands of users in days or weeks. The biggest developer + community in \\u00a0GenAI\\nLearn alongside the 100K+ practitioners\\nwho are + pushing the industry forward.\\n And we built LangSmith to support all stages + of the AI engineering lifecycle, to get applications into production faster.\\n\"}, + {\"url\": \"https://www.bluelabellabs.com/blog/what-is-langchain/\", \"content\": + \"Similar setups to AutoTweeter would allow businesses to parse past and present + customer queries to understand drop-off points and build the appropriate flows + to solve customer problems on the spot, instead of spitting them into a queue + to wait for a human and plot how they\\u2019re going to steal from your business.\\n + Rather than creating a bot to tweet in your tone on your behalf about your love + for processed snacks, hockey, and political happenings, you could train such + a system on your brand voice, knowledge bases, and other resources to create + a highly valuable feature for your digital touchpoints.\\n Here, you essentially + create a curated ML system that\\u2019s tailored to your brand with specific + issues and requests from your customers \\u2013 with a little bit of dev magic, + you can quickly build a solution that integrates features like payment systems, + tech support, ordering, account management, and so much more.\\n Creating a + bot for social media that can post just like you\\nSocial media already has + enough bots engaging with people as if they\\u2019re real people, but for the + sake of making a functional example, our co-founder Bobby created a bot called + AutoTweeter for the social media platform we all still call Twitter.\\n How + does LangChain work?\\nLangchain essentially serves as a middleman between other + LLMs, data sources, and API-accessible functions \\u2013 more precisely, it\\u2019s + a comprehensive open-source library of components that allow users to build + better (i.e., more custom) solutions by interconnecting different resources.\\n\"}, + {\"url\": \"https://www.latent.space/p/langchain\", \"content\": \"And so you + can do a lot of things like reflection and kind of like making sure that you''re + on the right track and really put more safeguards and thinking about these agents + as opposed to relying on like chas and interface, like the bot we have that''s + on GitHub answering questions on the issues, I think probably gives better answers + than the bots that we have that are on chat on the website. [00:50:32]\\nHarrison: + And then between kind of like allowing for, I think one helpful thing that we + did in terms of like distinguishing between basically the base interfaces and + then more complex stuff is part of the separation around the docs is there''s + like the components piece, which has the model IO, the retrieval, the agents, + the callbacks, things like that. And then like, I know a lot of people are doing + this, but like a personal assistant for like managing like email calendar, basic + stuff, which I think is, I think that''s like a fantastic application for these + like agent like things, because if you think about personal assistants today, + you usually interact, I don''t have one, but I''m told you interact with them + over email. Show Notes\\nLangChain\\nLangChain\\u2019s Berkshire Hathaway Homepage\\nAbstractions + tweet\\nLangSmith\\nLangSmith Cookbooks repo\\nLangChain Retrieval blog\\nEvaluating + CSV Question/Answering blog and YouTube\\nMultiOn Partner blog\\nHarvard Sports + Analytics Collective\\nEvaluating RAG Webinar\\nawesome-langchain:\\nLLM Math + Chain\\nSelf-Ask\\nLangChain Hub UI\\n\\u201cLangChain is Pointless\\u201d\\nHarrison\\u2019s + links\\nsports - estimating player compatibility in the NBA\\nearly interest + in prompt injections\\nGitHub\\nTwitter\\nTimestamps\\n[00:00:00] Introduction\\n[00:00:48] + Harrison''s background and how sports led him into ML\\n[00:04:54] Memory (for + long running chats, whether a simple Buffer, Knowledge Graph, Summary, or Vector + Store)\\nUse-Cases: compositions of Components\\nChains: combining a PromptTemplate, + LLM Model and optional OutputParser\\nwith Router, Sequential, and Transform + Chains for advanced usecases\\nsavable, sharable schemas that can be loaded + from LangChainHub\\nAgents: a chain that has access to a suite of tools, of + nondeterministic8 length because the LLM is used as a reasoning engine to determine + which actions to take and in which order.\"}, {\"url\": \"https://www.langchain.com/\", + \"content\": \"Join Discord\\nContribute on GitHub\\n5.3M\\nMonthly Downloads\\n67k\\nGithub + Stars\\n1800+\\nContributors\\n34k+\\nApps Powered\\nReady to build?\\nGo To + Docs\\nGo To Docs\\nGo To Docs\\nContact Sales\\nContact Sales\\nContact Sales\\nProducts + and use-cases\\nLangChain\\nLangSmith\\nRetrieval\\nAgents\\nInspiration\\nCode\\nGitHub\\nLangChain + Hub\\nPython Docs\\nJS/TS Docs\\nSocial\\nTwitter\\nDiscord\\nBlog\\nLinkedIn\\nYouTube\\nTerms + of Service\\nSign up for our newsletter\\nProducts and use-cases\\nLangChain\\nLangSmith\\nRetrieval\\nAgents\\nInspiration\\nCode\\nGitHub\\nLangChain + Hub\\nPython Docs\\nJS/TS Docs\\nSocial\\nTwitter\\nDiscord\\nBlog\\nLinkedIn\\nYouTube\\nTerms + of Service\\nSign up for our newsletter\\nProducts and use-cases\\nLangChain\\nLangSmith\\nRetrieval\\nAgents\\nInspiration\\nCode\\nGitHub\\nLangChain + Hub\\nPython Docs\\nJS/TS Docs\\nSocial\\nTwitter\\nDiscord\\nBlog\\nLinkedIn\\nYouTube\\nTerms + of Service\\nSign up for our newsletter Chatbots\\nQ&A Over Docs\\nSummarization\\nCopilots\\nWorkflow + Automation\\nDocument Analysis\\nCustom Search\\nHarness the power\\u2013and + wrangle the complexity\\u2013 of LLMs\\nGet from proof-of-concept to production + on one platform\\nRely on a set of tools built especially for non-deterministic + models\\nLeverage new cognitive architectures and battle-tested orchestration\\nThanks + to our contributors and partners for making LangChain awesome\\nWhen you build + with LangChain, you build alongside a community of developers advancing the + frontier of what\\u2019s possible with LLMs.\\n \\ud83e\\udd9c\\ud83d\\udd17 + LangChain\\nLangSmith\\nLangServe\\nAgents\\nRetrieval\\nEvaluation\\nBlog\\nDocs\\n\\ud83e\\udd9c\\ud83d\\udd17 + LangChain\\n\\ud83e\\udd9c\\ud83d\\udd17 LangChain\\nLangChain Expression Language + makes it easy to create custom chains.\\n Go To Docs\\nGo To Docs\\nGo To Docs\\nContact + Sales\\nglobal corporations, STARTUPS, and TINKERERS build with LangChain\\nglobal + corporations, STARTUPS, and TINKERERS build with LangChain\\nglobal corporations, + STARTUPS, and TINKERERS build with LangChain\\nA complete set of powerful building + blocks\\nA complete set of powerful building blocks\\n Try it out here\\nGet + your LLM application from prototype to production\\nGet your LLM application + from prototype to production\\nBuild context-aware, reasoning applications with + LangChain\\u2019s flexible abstractions and AI-first toolkit.\\n\"}]", "tool_use_id": + "toolu_01WgfyEmetmVSsT91LvH5Dup", "is_error": false}]}], "model": "claude-3-5-sonnet-20240620", + "tool_choice": {"type": "any"}, "tools": [{"name": "scrape_website", "description": + "Scrape and summarize content from a given URL.", "input_schema": {"type": "object", + "properties": {"url": {"type": "string"}}, "required": ["url"]}}, {"name": "search", + "description": "Search for general results. This function performs a search + using the Tavily search engine, which is designed\nto provide comprehensive, + accurate, and trusted results. It''s particularly useful\nfor answering questions + about current events.", "input_schema": {"type": "object", "properties": {"query": + {"type": "string"}}, "required": ["query"]}}, {"name": "Info", "description": + "Call this when you have gathered all the relevant info", "input_schema": {"type": + "object", "properties": {"providers": {"type": "array", "items": {"type": "object", + "properties": {"name": {"type": "string", "description": "Company name"}, "technology_summary": + {"type": "string", "description": "Brief summary of their chip technology for + LLM training"}, "current_market_share": {"type": "string", "description": "Estimated + current market share percentage or position"}, "future_outlook": {"type": "string", + "description": "Brief paragraph on future prospects and developments"}}, "required": + ["name", "technology_summary", "current_market_share", "future_outlook"]}, "description": + "List of top chip providers for LLM Training"}, "overall_market_trends": {"type": + "string", "description": "Brief paragraph on general trends in the LLM chip + market"}}, "required": ["providers", "overall_market_trends"]}}]}' + headers: {} + method: POST + uri: https://api.anthropic.com/v1/messages + response: + body: + string: !!binary | + H4sIAAAAAAAAA4xWYW8bNwz9K8QBQzfACdyk61p/C5oh9doEHpq0A4bBoHX0HVGdeBV1dt0g/32g + 7uzY6ZL1m62TqMf3yEfdFlwWk6LRaj5+fvXqj9fviP969W39iVz17uvl9WnzZzEq0qYl20WqWFEx + KqJ4W0BV1oQhFaOikZJ8MSmcx66ko9OjX49UQqB0dDI+eTF+eTIuRoWTkCikYvL37TZoEvHzTi1q + hmL/u/n4+afrZjZr+CY5udSGPp69/LL8/bQYFQEbOzcNS7Ezoe1SMbkt2igrLilqDj5suvo4PZ+e + WQbk6iBeqs1cu6bBuCkmxVuu6qOW4lJig8ERXMxuFKRN3PA3KmEpEc6mgKGEBl3NgcATxsChGgEH + 57uSQwWpJjh7Ph7njW/th1JkUsu3i5FCmjcYP1Oaa43RYJ1LwwFDGgFp4gYTlSArivB6/BPIso84 + BVdzC/3RYlQsu9RFmkuXvMjnXXLACvS1JWdBkkCDHBJyAE5qcI0Si9OKcmIJI1h0CZboSC2HSKiW + hJOmpZR3HMN1TXkBwwZMMg4dqQXnEGSFiWDNqYZAa8Doak7kDJtmBlqMKfSX5jMqnktebjKgPpsd + mOPibrQT6+zy/DGlLqenO1oBnSNPEZNEHRnOBZsk8MbUMwRZxkG8tcTPXrB8Qo2LKOssqRGjib0H + BG3Qe2glGsyejNgTvKup7wQ5uzw3NSrs8aSILp/OZFn2b86vzg4Yy3D3kzvk/oG0g1y0z2ROAjjk + mnFWVxVsCI0ZE4Jd5zH6zXZHiQnBUUgU8931wxawq7tkUZSqhkLSA42mIZF/tJ9wgQHhPS70GVxg + V/JJ7p89vfKlXTsAnZkdwEdyrmbJsj2u0ocsiElEmnDhWWsqoY2kZLiH/P6/a3IGxiyHlXVfqKAm + XHHP0fZ8SSvy0hoBGTJyk4t5QU4agkYi3bfMinrZtHOOVK2DTaE9OgDdl477ms/xeliWMshySZFD + pbC20lsQuNg5Rp9LeFtM+3IfKHIhUnl6TJJrCioRZlEMmQW6CYbt5+vZjf4CJSlXgUrQlhwv2aH3 + m3zvQ8P7kUb6wFXIQUIyLp2XrjxaoFJpvCrFFTvSvs+CJFhzSX4DuEL2uPA7ESsKFNE/IWKf9KE3 + DZJlermxcdA3yvXsBu65OYZPNXuCNnKD0VTvDN/gl7IeYO+hTTVFeqbQio2urS6GPQKWNi0kgETw + 7Chkioc0etSHYkVsayfxUb1ydXquck1/r9r0ULUnBhRg23p2aOiekOyKXU3QetxQ7H2q6t0QIjmp + Ql+xP9xcu/z2fRBTJk5Cxmt5DGOEVwTT2c2BIx4aYI0PeTd4qbbBaRWby5UDDH/y1DZC9nLfn3ea + ohzOOlhGaQ4dJXNhrvfPqLChjN5vWUuRQqnWVjXB+/eX+3RszToyBZcJxJbLAW9flPb2UToctcMQ + H+TxGyjz2yCR3o+Ni4MKHvqH1kAhRQypN9X9HGwzVNx/y+PhP6zuIWd7Lp1FyEWPkHOGJGuMpfbG + Z+1ZbY5ouWTHW4dUh30Tq/huYB4V1mTjVDNT+oTf7I/qY3iTe3D3pMtJoNddkw/PLo59x3aapOmv + 6OvnUJLt/NzahS1iyG83NExNduQneslhiwv2nPj+0ZQfIlFUYYWRpbMPZacpT/HizqpHk7Rz2yvh + 8KGbPyh96azNi0novB8VXX5bT277J+08yWcKWkxenL54OSqkS/uLv41P7u7+BQAA//8DAB7P4A+8 + CwAA + headers: + CF-Cache-Status: + - DYNAMIC + CF-RAY: + - 8c4b20015f697ad0-SJC + Connection: + - keep-alive + Content-Encoding: + - gzip + Content-Type: + - application/json + Date: + - Tue, 17 Sep 2024 18:23:14 GMT + Server: + - cloudflare + Transfer-Encoding: + - chunked + X-Robots-Tag: + - none + anthropic-ratelimit-requests-limit: + - '4000' + anthropic-ratelimit-requests-remaining: + - '3999' + anthropic-ratelimit-requests-reset: + - '2024-09-17T18:23:54Z' + anthropic-ratelimit-tokens-limit: + - '400000' + anthropic-ratelimit-tokens-remaining: + - '400000' + anthropic-ratelimit-tokens-reset: + - '2024-09-17T18:23:14Z' + request-id: + - req_01Dv6svvdB3VpSpXUSrHhhgK + via: + - 1.1 google + x-cloud-trace-context: + - d7ef3c2e361965193d12822c4d4fbc0a + status: + code: 200 + message: OK +- request: + body: '{"max_tokens": 1024, "messages": [{"role": "user", "content": "You are + doing web research on behalf of a user. You are trying to figure out this information:\n\n\n{\n \"type\": + \"object\",\n \"properties\": {\n \"providers\": {\n \"type\": \"array\",\n \"items\": + {\n \"type\": \"object\",\n \"properties\": {\n \"name\": + {\n \"type\": \"string\",\n \"description\": \"Company + name\"\n },\n \"technology_summary\": {\n \"type\": + \"string\",\n \"description\": \"Brief summary of their chip technology + for LLM training\"\n },\n \"current_market_share\": {\n \"type\": + \"string\",\n \"description\": \"Estimated current market share percentage + or position\"\n },\n \"future_outlook\": {\n \"type\": + \"string\",\n \"description\": \"Brief paragraph on future prospects + and developments\"\n }\n },\n \"required\": [\n \"name\",\n \"technology_summary\",\n \"current_market_share\",\n \"future_outlook\"\n ]\n },\n \"description\": + \"List of top chip providers for LLM Training\"\n },\n \"overall_market_trends\": + {\n \"type\": \"string\",\n \"description\": \"Brief paragraph on + general trends in the LLM chip market\"\n }\n },\n \"required\": [\n \"providers\",\n \"overall_market_trends\"\n ]\n}\n\n\nYou + have access to the following tools:\n\n- `Search`: call a search tool and get + back some results\n- `ScrapeWebsite`: scrape a website and get relevant notes + about the given request. This will update the notes above.\n- `Info`: call this + when you are done and have gathered all the relevant info\n\nHere is the information + you have about the topic you are researching:\n\nTopic: Top 5 chip providers + for LLM training"}, {"role": "assistant", "content": [{"type": "tool_use", "name": + "search", "input": {"query": "Top 5 chip providers for LLM training market share + and technology"}, "id": "toolu_01WgfyEmetmVSsT91LvH5Dup"}]}, {"role": "user", + "content": [{"type": "tool_result", "content": "[{\"url\": \"https://www.cbinsights.com/company/langchain\", + \"content\": \"It was founded in 2023 and is based in San Francisco, California. + LangChain specializes in the development of large language model (LLM) applications, + providing a suite of products that support developers throughout the application + lifecycle. Use the CB Insights Platform to explore LangChain''s full profile.\"}, + {\"url\": \"https://en.wikipedia.org/wiki/LangChain\", \"content\": \"In October + 2023 LangChain introduced LangServe, a deployment tool designed to facilitate + the transition from LCEL (LangChain Expression Language) prototypes to production-ready + applications.[5]\\nIntegrations[edit]\\nAs of March 2023, LangChain included + integrations with systems including Amazon, Google, and Microsoft Azure cloud + storage; API wrappers for news, movie information, and weather; Bash for summarization, + syntax and semantics checking, and execution of shell scripts; multiple web + scraping subsystems and templates; few-shot learning prompt generation support; + finding and summarizing \\\"todo\\\" tasks in code; Google Drive documents, + spreadsheets, and presentations summarization, extraction, and creation; Google + Search and Microsoft Bing web search; OpenAI, Anthropic, and Hugging Face language + models; iFixit repair guides and wikis search and summarization; MapReduce for + question answering, combining documents, and question generation; N-gram overlap + scoring; PyPDF, pdfminer, fitz, and pymupdf for PDF file text extraction and + manipulation; Python and JavaScript code generation, analysis, and debugging; + Milvus vector database[6] to store and retrieve vector embeddings; Weaviate + vector database[7] to cache embedding and data objects; Redis cache database + storage; Python RequestsWrapper and other methods for API requests; SQL and + NoSQL databases including JSON support; Streamlit, including for logging; text + mapping for k-nearest neighbors search; time zone conversion and calendar operations; + tracing and recording stack symbols in threaded and asynchronous subprocess + runs; and the Wolfram Alpha website and SDK.[8] As a language model integration + framework, LangChain''s use-cases largely overlap with those of language models + in general, including document analysis and summarization, chatbots, and code + analysis.[2]\\nHistory[edit]\\nLangChain was launched in October 2022 as an + open source project by Harrison Chase, while working at machine learning startup + Robust Intelligence. In April 2023, LangChain had incorporated and the new startup + raised over $20 million in funding at a valuation of at least $200 million from + venture firm Sequoia Capital, a week after announcing a $10 million seed investment + from Benchmark.[3][4]\\n The project quickly garnered popularity, with improvements + from hundreds of contributors on GitHub, trending discussions on Twitter, lively + activity on the project''s Discord server, many YouTube tutorials, and meetups + in San Francisco and London. As of April 2023, it can read from more than 50 + document types and data sources.[9]\\nReferences[edit]\\nExternal links[edit]\"}, + {\"url\": \"https://www.forbes.com/companies/langchain/\", \"content\": \"LangChain. + Founded in late 2022, LangChain was an instant hit in Silicon Valley for its + software that enables anyone to build apps powered by large language models + like GPT-4 in as little as two ...\"}, {\"url\": \"https://tedai-sanfrancisco.ted.com/speakers-1/harrison-chase-/co-founder-and-ceo,-langchain/\", + \"content\": \"Harrison Chase, a Harvard graduate in statistics and computer + science, co-founded LangChain to streamline the development of Language Model + applications with open-source Python/Typescript packages. Chase''s experience + includes heading the Machine Learning team at Robust Intelligence, focusing + on the testing and validation of machine learning models, and leading the entity + linking team at Kensho ...\"}, {\"url\": \"https://www.crunchbase.com/organization/langchain\", + \"content\": \"Summary. LangChain may be growing as evidenced by securing $25 + million in funding, which is a significant financial milestone for the company. + Additionally, the launch of a new platform to facilitate the full lifecycle + of Large Language Model applications suggests expansion in their product offerings. + The integration of Gemini Pro API, enabling ...\"}, {\"url\": \"https://craft.co/langchain\", + \"content\": \"LangChain has 5 employees at their 1 location and $35 m in total + funding,. See insights on LangChain including office locations, competitors, + revenue, financials, executives, subsidiaries and more at Craft. ... Harrison + Chase, Co-Founder and Chief Executive Officer. Linkedin;\"}, {\"url\": \"https://www.langchain.com/about\", + \"content\": \"Ready to start shipping\\nreliable GenAI apps faster?\\nLangChain + and LangSmith are critical parts of the reference\\narchitecture to get you + from prototype to production. Announcing the General Availability of LangSmith\\nToday, + we\\u2019re thrilled to announce the general availability of LangSmith \\u2014 + our solution for LLM application development, monitoring, and testing.\\n We + build products that enable developers to go from an idea to working code in + an afternoon and in the hands of users in days or weeks. The biggest developer + community in \\u00a0GenAI\\nLearn alongside the 100K+ practitioners\\nwho are + pushing the industry forward.\\n And we built LangSmith to support all stages + of the AI engineering lifecycle, to get applications into production faster.\\n\"}, + {\"url\": \"https://www.bluelabellabs.com/blog/what-is-langchain/\", \"content\": + \"Similar setups to AutoTweeter would allow businesses to parse past and present + customer queries to understand drop-off points and build the appropriate flows + to solve customer problems on the spot, instead of spitting them into a queue + to wait for a human and plot how they\\u2019re going to steal from your business.\\n + Rather than creating a bot to tweet in your tone on your behalf about your love + for processed snacks, hockey, and political happenings, you could train such + a system on your brand voice, knowledge bases, and other resources to create + a highly valuable feature for your digital touchpoints.\\n Here, you essentially + create a curated ML system that\\u2019s tailored to your brand with specific + issues and requests from your customers \\u2013 with a little bit of dev magic, + you can quickly build a solution that integrates features like payment systems, + tech support, ordering, account management, and so much more.\\n Creating a + bot for social media that can post just like you\\nSocial media already has + enough bots engaging with people as if they\\u2019re real people, but for the + sake of making a functional example, our co-founder Bobby created a bot called + AutoTweeter for the social media platform we all still call Twitter.\\n How + does LangChain work?\\nLangchain essentially serves as a middleman between other + LLMs, data sources, and API-accessible functions \\u2013 more precisely, it\\u2019s + a comprehensive open-source library of components that allow users to build + better (i.e., more custom) solutions by interconnecting different resources.\\n\"}, + {\"url\": \"https://www.latent.space/p/langchain\", \"content\": \"And so you + can do a lot of things like reflection and kind of like making sure that you''re + on the right track and really put more safeguards and thinking about these agents + as opposed to relying on like chas and interface, like the bot we have that''s + on GitHub answering questions on the issues, I think probably gives better answers + than the bots that we have that are on chat on the website. [00:50:32]\\nHarrison: + And then between kind of like allowing for, I think one helpful thing that we + did in terms of like distinguishing between basically the base interfaces and + then more complex stuff is part of the separation around the docs is there''s + like the components piece, which has the model IO, the retrieval, the agents, + the callbacks, things like that. And then like, I know a lot of people are doing + this, but like a personal assistant for like managing like email calendar, basic + stuff, which I think is, I think that''s like a fantastic application for these + like agent like things, because if you think about personal assistants today, + you usually interact, I don''t have one, but I''m told you interact with them + over email. Show Notes\\nLangChain\\nLangChain\\u2019s Berkshire Hathaway Homepage\\nAbstractions + tweet\\nLangSmith\\nLangSmith Cookbooks repo\\nLangChain Retrieval blog\\nEvaluating + CSV Question/Answering blog and YouTube\\nMultiOn Partner blog\\nHarvard Sports + Analytics Collective\\nEvaluating RAG Webinar\\nawesome-langchain:\\nLLM Math + Chain\\nSelf-Ask\\nLangChain Hub UI\\n\\u201cLangChain is Pointless\\u201d\\nHarrison\\u2019s + links\\nsports - estimating player compatibility in the NBA\\nearly interest + in prompt injections\\nGitHub\\nTwitter\\nTimestamps\\n[00:00:00] Introduction\\n[00:00:48] + Harrison''s background and how sports led him into ML\\n[00:04:54] Memory (for + long running chats, whether a simple Buffer, Knowledge Graph, Summary, or Vector + Store)\\nUse-Cases: compositions of Components\\nChains: combining a PromptTemplate, + LLM Model and optional OutputParser\\nwith Router, Sequential, and Transform + Chains for advanced usecases\\nsavable, sharable schemas that can be loaded + from LangChainHub\\nAgents: a chain that has access to a suite of tools, of + nondeterministic8 length because the LLM is used as a reasoning engine to determine + which actions to take and in which order.\"}, {\"url\": \"https://www.langchain.com/\", + \"content\": \"Join Discord\\nContribute on GitHub\\n5.3M\\nMonthly Downloads\\n67k\\nGithub + Stars\\n1800+\\nContributors\\n34k+\\nApps Powered\\nReady to build?\\nGo To + Docs\\nGo To Docs\\nGo To Docs\\nContact Sales\\nContact Sales\\nContact Sales\\nProducts + and use-cases\\nLangChain\\nLangSmith\\nRetrieval\\nAgents\\nInspiration\\nCode\\nGitHub\\nLangChain + Hub\\nPython Docs\\nJS/TS Docs\\nSocial\\nTwitter\\nDiscord\\nBlog\\nLinkedIn\\nYouTube\\nTerms + of Service\\nSign up for our newsletter\\nProducts and use-cases\\nLangChain\\nLangSmith\\nRetrieval\\nAgents\\nInspiration\\nCode\\nGitHub\\nLangChain + Hub\\nPython Docs\\nJS/TS Docs\\nSocial\\nTwitter\\nDiscord\\nBlog\\nLinkedIn\\nYouTube\\nTerms + of Service\\nSign up for our newsletter\\nProducts and use-cases\\nLangChain\\nLangSmith\\nRetrieval\\nAgents\\nInspiration\\nCode\\nGitHub\\nLangChain + Hub\\nPython Docs\\nJS/TS Docs\\nSocial\\nTwitter\\nDiscord\\nBlog\\nLinkedIn\\nYouTube\\nTerms + of Service\\nSign up for our newsletter Chatbots\\nQ&A Over Docs\\nSummarization\\nCopilots\\nWorkflow + Automation\\nDocument Analysis\\nCustom Search\\nHarness the power\\u2013and + wrangle the complexity\\u2013 of LLMs\\nGet from proof-of-concept to production + on one platform\\nRely on a set of tools built especially for non-deterministic + models\\nLeverage new cognitive architectures and battle-tested orchestration\\nThanks + to our contributors and partners for making LangChain awesome\\nWhen you build + with LangChain, you build alongside a community of developers advancing the + frontier of what\\u2019s possible with LLMs.\\n \\ud83e\\udd9c\\ud83d\\udd17 + LangChain\\nLangSmith\\nLangServe\\nAgents\\nRetrieval\\nEvaluation\\nBlog\\nDocs\\n\\ud83e\\udd9c\\ud83d\\udd17 + LangChain\\n\\ud83e\\udd9c\\ud83d\\udd17 LangChain\\nLangChain Expression Language + makes it easy to create custom chains.\\n Go To Docs\\nGo To Docs\\nGo To Docs\\nContact + Sales\\nglobal corporations, STARTUPS, and TINKERERS build with LangChain\\nglobal + corporations, STARTUPS, and TINKERERS build with LangChain\\nglobal corporations, + STARTUPS, and TINKERERS build with LangChain\\nA complete set of powerful building + blocks\\nA complete set of powerful building blocks\\n Try it out here\\nGet + your LLM application from prototype to production\\nGet your LLM application + from prototype to production\\nBuild context-aware, reasoning applications with + LangChain\\u2019s flexible abstractions and AI-first toolkit.\\n\"}]", "tool_use_id": + "toolu_01WgfyEmetmVSsT91LvH5Dup", "is_error": false}, {"type": "text", "text": + "I am thinking of calling the info tool with the info below. Is this good? Give + your reasoning as well. You can encourage the Assistant to look at specific + URLs if that seems relevant, or do more searches.\nIf you don''t think it is + good, you should be very specific about what could be improved.\n\n{\n \"providers\": + [\n {\n \"name\": \"NVIDIA\",\n \"technology_summary\": \"High-performance + GPUs optimized for AI and machine learning, including the A100 and H100 series\",\n \"current_market_share\": + \"Dominant, estimated over 90% of the AI chip market\",\n \"future_outlook\": + \"NVIDIA is expected to maintain its leadership position, but faces increasing + competition. The company continues to innovate with new architectures and partnerships + to solidify its market position.\"\n },\n {\n \"name\": \"AMD\",\n \"technology_summary\": + \"MI300 series accelerators, combining CPUs and GPUs for AI workloads\",\n \"current_market_share\": + \"Growing, but still a small portion compared to NVIDIA\",\n \"future_outlook\": + \"AMD is gaining traction with its CDNA architecture and MI300 series. The company + is expected to increase its market share in the coming years, particularly in + the data center and high-performance computing segments.\"\n },\n {\n \"name\": + \"Intel\",\n \"technology_summary\": \"Habana Labs'' Gaudi2 AI accelerators + and upcoming Ponte Vecchio GPUs\",\n \"current_market_share\": \"Small + but established presence in the AI chip market\",\n \"future_outlook\": + \"Intel is investing heavily in AI chip development and aims to become more + competitive. The success of its Habana Labs acquisition and future GPU offerings + will be crucial for gaining market share.\"\n },\n {\n \"name\": + \"Google\",\n \"technology_summary\": \"Tensor Processing Units (TPUs) + designed specifically for machine learning workloads\",\n \"current_market_share\": + \"Significant in cloud-based AI services, but not widely available in the general + market\",\n \"future_outlook\": \"Google continues to develop and improve + its TPU technology. While primarily used in its own cloud services, there''s + potential for wider adoption or licensing in the future.\"\n },\n {\n \"name\": + \"Graphcore\",\n \"technology_summary\": \"Intelligence Processing Units + (IPUs) designed for AI and machine learning applications\",\n \"current_market_share\": + \"Niche player with growing recognition in the AI chip market\",\n \"future_outlook\": + \"Graphcore is gaining attention for its innovative IPU architecture. The company + has potential for growth, especially in specialized AI applications, but faces + strong competition from established players.\"\n }\n ],\n \"overall_market_trends\": + \"The LLM chip market is experiencing rapid growth and intense competition. + NVIDIA currently dominates with its GPU technology, but new entrants and established + tech giants are investing heavily in specialized AI accelerators. There''s a + trend towards more energy-efficient and scalable solutions, as well as chips + designed specifically for AI workloads. Cloud providers are also developing + their own custom chips. The market is expected to continue expanding as demand + for AI and machine learning capabilities increases across various industries.\"\n}"}]}], + "model": "claude-3-5-sonnet-20240620", "tool_choice": {"type": "tool", "name": + "InfoIsSatisfactory"}, "tools": [{"name": "InfoIsSatisfactory", "description": + "Validate whether the current extracted info is satisfactory and complete.", + "input_schema": {"properties": {"reason": {"description": "First, provide reasoning + for why this is either good or bad as a final result. Must include at least + 3 reasons.", "items": {"type": "string"}, "type": "array"}, "is_satisfactory": + {"description": "After providing your reasoning, provide a value indicating + whether the result is satisfactory. If not, you will continue researching.", + "type": "boolean"}}, "required": ["reason", "is_satisfactory"], "type": "object"}}]}' + headers: {} + method: POST + uri: https://api.anthropic.com/v1/messages + response: + body: + string: !!binary | + H4sIAAAAAAAAA0xT0U4bMRD8lZVf+uKgJAQo95aWigYRXmjaSghFlr13Z+LzXr17gRDl3yvfEZTH + s9czszNze+WdKlTD1Xo8mfP15Ft8fFull8f5XWvv3r4/PEyVVrJrMU8hs6lQaZUo5APD7FlMFKVV + Qw6DKpQNpnM4Oh9djJhiRBlNx9PZ+HI6VlpZioJRVPG0P4IKUVh3nFF7Kfm7W48ny6vGzt5WL9fv + t245+bta/bHVz1ZpFU2T3y1iSQt+NOK5NFYo7TJCbDtRxV4lNExRFU/qV43gY0mpMeIpQpto6x06 + sLTFxCA1glALF2Br3x6vE0NJCe7vlyDJ+OhjBYYh4b8OWdBp8NGGzuXzh9+Lm8Vcw3x5o2ERBYOG + W6IqoAYTHdwm09aWEp4prX4YW3+SfGHAKGn3AYaDmkziEzooPQbHBeSNNQjaOlKgagfcNY1JOw22 + SwmjQGPSBgW4NumDtOykSwjUSSDaaKj8tt8BLDVtwhoj+y1C9mDr8RWozNw+QUvse6N87NUM0Fl6 + djLPmxCOhJIwOgZGe+otg4GKyB11foB/qmUxgqcqnU8DAh9Hs/F9HgORhtpXdfBVLXmLDe6gJR+F + gTtb52SGEL4wOGp8NNFiH1HuQZ9106L0ew329Kln8SD0alLeoUXrTfDv6GC+AGMtBkxGKPGZetbK + 85pPy1aUJjAeDs9asVC7PjbutND9BefORIuqiF0IWnX9P1Tsh7KuhTYYWRWzr7NLraiT08PJ1fnh + 8B8AAP//AwBvxqRDpAMAAA== + headers: + CF-Cache-Status: + - DYNAMIC + CF-RAY: + - 8c4b20560e191722-SJC + Connection: + - keep-alive + Content-Encoding: + - gzip + Content-Type: + - application/json + Date: + - Tue, 17 Sep 2024 18:23:19 GMT + Server: + - cloudflare + Transfer-Encoding: + - chunked + X-Robots-Tag: + - none + anthropic-ratelimit-requests-limit: + - '4000' + anthropic-ratelimit-requests-remaining: + - '3999' + anthropic-ratelimit-requests-reset: + - '2024-09-17T18:23:54Z' + anthropic-ratelimit-tokens-limit: + - '400000' + anthropic-ratelimit-tokens-remaining: + - '400000' + anthropic-ratelimit-tokens-reset: + - '2024-09-17T18:23:19Z' + request-id: + - req_016Pta6sCCcxe74qtjr2Krgh + via: + - 1.1 google + x-cloud-trace-context: + - 2dbbddef4ed8fa1800e744fabf982670 + status: + code: 200 + message: OK version: 1 diff --git a/tests/integration_tests/test_graph.py b/tests/integration_tests/test_graph.py index 7a54b0a..7ce9e3f 100644 --- a/tests/integration_tests/test_graph.py +++ b/tests/integration_tests/test_graph.py @@ -7,7 +7,7 @@ @pytest.fixture(scope="function") -def template_schema() -> Dict[str, Any]: +def extraction_schema() -> Dict[str, Any]: return { "type": "object", "properties": { @@ -29,15 +29,90 @@ def template_schema() -> Dict[str, Any]: } + + @pytest.mark.asyncio @unit -async def test_researcher_simple_runthrough(template_schema: Dict[str, Any]) -> None: +async def test_researcher_simple_runthrough(extraction_schema: Dict[str, Any]) -> None: res = await graph.ainvoke( { "topic": "LangChain", - "template_schema": template_schema, + "extraction_schema": extraction_schema, } ) assert res["info"] is not None assert "harrison" in res["info"]["founder"].lower() + + +@pytest.fixture(scope="function") +def array_extraction_schema() -> Dict[str, Any]: + return { + "type": "object", + "properties": { + "providers": { + "type": "array", + "items": { + "type": "object", + "properties": { + "name": { + "type": "string", + "description": "Company name" + }, + "technology_summary": { + "type": "string", + "description": "Brief summary of their chip technology for LLM training" + }, + "current_market_share": { + "type": "string", + "description": "Estimated current market share percentage or position" + }, + "future_outlook": { + "type": "string", + "description": "Brief paragraph on future prospects and developments" + } + }, + "required": ["name", "technology_summary", "current_market_share", "future_outlook"] + }, + "description": "List of top chip providers for LLM Training" + }, + "overall_market_trends": { + "type": "string", + "description": "Brief paragraph on general trends in the LLM chip market" + } + }, + "required": ["providers", "overall_market_trends"] + } + + + +@pytest.mark.asyncio +@unit +async def test_researcher_list_type(array_extraction_schema: Dict[str, Any]) -> None: + res = await graph.ainvoke( + { + "topic": "Top 5 chip providers for LLM training", + "extraction_schema": array_extraction_schema, + } + ) + # Check that nvidia is amongst them lol + info = res["info"] + assert "providers" in info + assert isinstance(res["info"]["providers"], list) + assert len(info["providers"]) == 5 # Ensure we got exactly 5 providers + + # Check for NVIDIA's presence + nvidia_present = any(provider["name"].lower().strip() == "nvidia" for provider in info["providers"]) + assert nvidia_present, "NVIDIA should be among the top 5 chip providers for LLM training" + + # Validate structure of each provider + for provider in info["providers"]: + assert "name" in provider + assert "technology_summary" in provider + assert "current_market_share" in provider + assert "future_outlook" in provider + + # Check for overall market trends + assert "overall_market_trends" in info + assert isinstance(info["overall_market_trends"], str) + assert len(info["overall_market_trends"]) > 0 \ No newline at end of file From fd909ad8f0c778cedbeaea86f1ea33a395255916 Mon Sep 17 00:00:00 2001 From: William Fu-Hinthorn <13333726+hinthornw@users.noreply.github.com> Date: Tue, 17 Sep 2024 12:29:16 -0700 Subject: [PATCH 3/5] Remove dropdowns --- README.md | 90 ++++--------------------------------------------------- 1 file changed, 6 insertions(+), 84 deletions(-) diff --git a/README.md b/README.md index ea1d040..a9f6a19 100644 --- a/README.md +++ b/README.md @@ -40,9 +40,9 @@ The primary [search tool](./src/enrichment_agent/tools.py) [^1] used is [Tavily] Setup instruction auto-generated by `langgraph template lock`. DO NOT EDIT MANUALLY. --> -
-Setup for `model_name` -The `llm` configuration defaults are shown below: +### Setup Model + +The defaults values for `model_name` are shown below: ```yaml model_name: anthropic/claude-3-5-sonnet-20240620 @@ -50,7 +50,7 @@ model_name: anthropic/claude-3-5-sonnet-20240620 Follow the instructions below to get set up, or pick one of the additional options. -### Anthropic Chat Models +#### Anthropic To use Anthropic's chat models: @@ -60,17 +60,7 @@ To use Anthropic's chat models: ``` ANTHROPIC_API_KEY=your-api-key ``` -### Fireworks Chat Models - -To use Fireworks AI's chat models: - -1. Sign up for a [Fireworks AI account](https://app.fireworks.ai/signup) and obtain an API key. -2. Add your Fireworks AI API key to your `.env` file: - -``` -FIREWORKS_API_KEY=your-api-key -``` -#### OpenAI Chat Models +#### OpenAI To use OpenAI's chat models: @@ -80,7 +70,7 @@ To use OpenAI's chat models: OPENAI_API_KEY=your-api-key ``` -
+ @@ -158,74 +148,6 @@ Configuration auto-generated by `langgraph template lock`. DO NOT EDIT MANUALLY. "value": "anthropic/claude-instant-1.2", "variables": "ANTHROPIC_API_KEY" }, - { - "value": "fireworks/gemma2-9b-it", - "variables": "FIREWORKS_API_KEY" - }, - { - "value": "fireworks/llama-v3-70b-instruct", - "variables": "FIREWORKS_API_KEY" - }, - { - "value": "fireworks/llama-v3-70b-instruct-hf", - "variables": "FIREWORKS_API_KEY" - }, - { - "value": "fireworks/llama-v3-8b-instruct", - "variables": "FIREWORKS_API_KEY" - }, - { - "value": "fireworks/llama-v3-8b-instruct-hf", - "variables": "FIREWORKS_API_KEY" - }, - { - "value": "fireworks/llama-v3p1-405b-instruct", - "variables": "FIREWORKS_API_KEY" - }, - { - "value": "fireworks/llama-v3p1-405b-instruct-long", - "variables": "FIREWORKS_API_KEY" - }, - { - "value": "fireworks/llama-v3p1-70b-instruct", - "variables": "FIREWORKS_API_KEY" - }, - { - "value": "fireworks/llama-v3p1-8b-instruct", - "variables": "FIREWORKS_API_KEY" - }, - { - "value": "fireworks/mixtral-8x22b-instruct", - "variables": "FIREWORKS_API_KEY" - }, - { - "value": "fireworks/mixtral-8x7b-instruct", - "variables": "FIREWORKS_API_KEY" - }, - { - "value": "fireworks/mixtral-8x7b-instruct-hf", - "variables": "FIREWORKS_API_KEY" - }, - { - "value": "fireworks/mythomax-l2-13b", - "variables": "FIREWORKS_API_KEY" - }, - { - "value": "fireworks/phi-3-vision-128k-instruct", - "variables": "FIREWORKS_API_KEY" - }, - { - "value": "fireworks/phi-3p5-vision-instruct", - "variables": "FIREWORKS_API_KEY" - }, - { - "value": "fireworks/starcoder-16b", - "variables": "FIREWORKS_API_KEY" - }, - { - "value": "fireworks/yi-large", - "variables": "FIREWORKS_API_KEY" - }, { "value": "openai/gpt-3.5-turbo", "variables": "OPENAI_API_KEY" From 2ae63b64f11a0820b1e55827763a5ba9fb48e963 Mon Sep 17 00:00:00 2001 From: William Fu-Hinthorn <13333726+hinthornw@users.noreply.github.com> Date: Tue, 17 Sep 2024 13:12:13 -0700 Subject: [PATCH 4/5] Use haiku --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index a9f6a19..9da2e1c 100644 --- a/README.md +++ b/README.md @@ -84,7 +84,7 @@ End setup instructions ## How to customize 1. **Customize research targets**: Provide a custom `extraction_schema` when calling the graph to gather different types of information. -2. **Select a different model**: We default to anthropic (sonnet-35). You can select a compatible chat model using `provider/model-name` via configuration. Example: `openai/gpt-4o-mini`. +2. **Select a different model**: We default to anthropic (sonnet-35). You can select a compatible chat model using `provider/model-name` via configuration. Example: `anthropic/claude-3-haiku-20240307`. 3. **Customize the prompt**: We provide a default prompt in [prompts.py](./src/enrichment_agent/prompts.py). You can easily update this via configuration in the studio. You can also quickly extend this template by: From 8c62cf78214d838de8af5c7ddb2f98d91a766338 Mon Sep 17 00:00:00 2001 From: William Fu-Hinthorn <13333726+hinthornw@users.noreply.github.com> Date: Tue, 17 Sep 2024 15:09:53 -0700 Subject: [PATCH 5/5] Lint --- .gitignore | 1 + src/enrichment_agent/graph.py | 23 +++++++------ src/enrichment_agent/tools.py | 6 ++-- tests/integration_tests/test_graph.py | 47 ++++++++++++++------------- 4 files changed, 39 insertions(+), 38 deletions(-) diff --git a/.gitignore b/.gitignore index ab74e15..b8af3ff 100644 --- a/.gitignore +++ b/.gitignore @@ -161,3 +161,4 @@ cython_debug/ # option (not recommended) you can uncomment the following to ignore the entire idea folder. #.idea/ .DS_Store +uv.lock diff --git a/src/enrichment_agent/graph.py b/src/enrichment_agent/graph.py index c610794..099e81c 100644 --- a/src/enrichment_agent/graph.py +++ b/src/enrichment_agent/graph.py @@ -18,12 +18,12 @@ from enrichment_agent.tools import scrape_website, search from enrichment_agent.utils import init_model + # Define the nodes async def call_agent_model( state: State, *, config: Optional[RunnableConfig] = None ) -> Dict[str, Any]: - """ - Call the primary Language Model (LLM) to decide on the next research action. + """Call the primary Language Model (LLM) to decide on the next research action. This asynchronous function performs the following steps: 1. Initializes configuration and sets up the 'Info' tool, which is the user-defined extraction schema. @@ -47,7 +47,6 @@ async def call_agent_model( - If the LLM calls the 'Info' tool, it's considered as submitting the final answer. - If the LLM doesn't call any tool, a prompt to use a tool is appended to the messages. """ - # Load configuration from the provided RunnableConfig configuration = Configuration.from_runnable_config(config) @@ -98,6 +97,7 @@ async def call_agent_model( "loop_step": 1, } + class InfoIsSatisfactory(BaseModel): """Validate whether the current extracted info is satisfactory and complete.""" @@ -112,8 +112,7 @@ class InfoIsSatisfactory(BaseModel): async def reflect( state: State, *, config: Optional[RunnableConfig] = None ) -> Dict[str, Any]: - """ - Validate the quality of the data enrichment agent's output. + """Validate the quality of the data enrichment agent's output. This asynchronous function performs the following steps: 1. Prepares the initial prompt using the main prompt template. @@ -195,8 +194,7 @@ async def reflect( def route_after_agent( state: State, ) -> Literal["reflect", "tools", "call_agent_model", "__end__"]: - """ - Schedule the next node after the agent's action. + """Schedule the next node after the agent's action. This function determines the next step in the research process based on the last message in the state. It handles three main scenarios: @@ -210,7 +208,7 @@ def route_after_agent( the message history. Returns: - Literal["reflect", "tools", "call_agent_model", "__end__"]: + Literal["reflect", "tools", "call_agent_model", "__end__"]: - "reflect": If the agent has submitted info for review. - "tools": If the agent has called a tool other than "Info". - "call_agent_model": If an unexpected message type is encountered. @@ -225,7 +223,7 @@ def route_after_agent( """ last_message = state.messages[-1] - # "If for some reason the last message is not an AIMessage (due to a bug or unexpected behavior elsewhere in the code), + # "If for some reason the last message is not an AIMessage (due to a bug or unexpected behavior elsewhere in the code), # it ensures the system doesn't crash but instead tries to recover by calling the agent model again. if not isinstance(last_message, AIMessage): return "call_agent_model" @@ -236,11 +234,11 @@ def route_after_agent( else: return "tools" + def route_after_checker( state: State, config: RunnableConfig ) -> Literal["__end__", "call_agent_model"]: - """ - Schedule the next node after the checker's evaluation. + """Schedule the next node after the checker's evaluation. This function determines whether to continue the research process or end it based on the checker's evaluation and the current state of the research. @@ -252,7 +250,7 @@ def route_after_checker( the maximum number of allowed loops. Returns: - Literal["__end__", "call_agent_model"]: + Literal["__end__", "call_agent_model"]: - "__end__": If the research process should terminate. - "call_agent_model": If further research is needed. @@ -282,6 +280,7 @@ def route_after_checker( else: return "__end__" + # Create the graph workflow = StateGraph( State, input=InputState, output=OutputState, config_schema=Configuration diff --git a/src/enrichment_agent/tools.py b/src/enrichment_agent/tools.py index d21ef73..5e595b1 100644 --- a/src/enrichment_agent/tools.py +++ b/src/enrichment_agent/tools.py @@ -23,8 +23,7 @@ async def search( query: str, *, config: Annotated[RunnableConfig, InjectedToolArg] ) -> Optional[list[dict[str, Any]]]: - """ - Perform a general web search using the Tavily search engine. + """Perform a general web search using the Tavily search engine. This asynchronous function executes the following steps: 1. Extracts configuration from the provided RunnableConfig. @@ -72,8 +71,7 @@ async def scrape_website( state: Annotated[State, InjectedState], config: Annotated[RunnableConfig, InjectedToolArg], ) -> str: - """ - Scrape and summarize content from a given URL. + """Scrape and summarize content from a given URL. This asynchronous function performs the following steps: 1. Fetches the content of the specified URL. diff --git a/tests/integration_tests/test_graph.py b/tests/integration_tests/test_graph.py index 7ce9e3f..bb114a2 100644 --- a/tests/integration_tests/test_graph.py +++ b/tests/integration_tests/test_graph.py @@ -29,8 +29,6 @@ def extraction_schema() -> Dict[str, Any]: } - - @pytest.mark.asyncio @unit async def test_researcher_simple_runthrough(extraction_schema: Dict[str, Any]) -> None: @@ -55,37 +53,38 @@ def array_extraction_schema() -> Dict[str, Any]: "items": { "type": "object", "properties": { - "name": { - "type": "string", - "description": "Company name" - }, + "name": {"type": "string", "description": "Company name"}, "technology_summary": { "type": "string", - "description": "Brief summary of their chip technology for LLM training" + "description": "Brief summary of their chip technology for LLM training", }, "current_market_share": { "type": "string", - "description": "Estimated current market share percentage or position" + "description": "Estimated current market share percentage or position", }, "future_outlook": { "type": "string", - "description": "Brief paragraph on future prospects and developments" - } + "description": "Brief paragraph on future prospects and developments", + }, }, - "required": ["name", "technology_summary", "current_market_share", "future_outlook"] + "required": [ + "name", + "technology_summary", + "current_market_share", + "future_outlook", + ], }, - "description": "List of top chip providers for LLM Training" + "description": "List of top chip providers for LLM Training", }, "overall_market_trends": { "type": "string", - "description": "Brief paragraph on general trends in the LLM chip market" - } + "description": "Brief paragraph on general trends in the LLM chip market", + }, }, - "required": ["providers", "overall_market_trends"] + "required": ["providers", "overall_market_trends"], } - @pytest.mark.asyncio @unit async def test_researcher_list_type(array_extraction_schema: Dict[str, Any]) -> None: @@ -100,19 +99,23 @@ async def test_researcher_list_type(array_extraction_schema: Dict[str, Any]) -> assert "providers" in info assert isinstance(res["info"]["providers"], list) assert len(info["providers"]) == 5 # Ensure we got exactly 5 providers - + # Check for NVIDIA's presence - nvidia_present = any(provider["name"].lower().strip() == "nvidia" for provider in info["providers"]) - assert nvidia_present, "NVIDIA should be among the top 5 chip providers for LLM training" - + nvidia_present = any( + provider["name"].lower().strip() == "nvidia" for provider in info["providers"] + ) + assert ( + nvidia_present + ), "NVIDIA should be among the top 5 chip providers for LLM training" + # Validate structure of each provider for provider in info["providers"]: assert "name" in provider assert "technology_summary" in provider assert "current_market_share" in provider assert "future_outlook" in provider - + # Check for overall market trends assert "overall_market_trends" in info assert isinstance(info["overall_market_trends"], str) - assert len(info["overall_market_trends"]) > 0 \ No newline at end of file + assert len(info["overall_market_trends"]) > 0