An LLM-powered knowledge curation system that researches a topic and generates a full-length report with citations.
Python agentic-ragdeep-researchemnlp2024knowledge-curationlarge-language-modelsnaaclnlpreport-generationretrieval-augmented-generation
Updated ollama and mistral
zenith11084de6324504f
diff --git a/examples/run_storm_wiki_mistral.py b/examples/run_storm_wiki_mistral.pyindex 291d287..76f85dc 100644--- a/examples/run_storm_wiki_mistral.py+++ b/examples/run_storm_wiki_mistral.py@@ -15,33 +15,26 @@ args.output_dir/storm_gen_article.txt # Final article generatedstorm_gen_article_polished.txt # Polished final article (if args.do_polish_article is True)"""-import osfrom argparse import ArgumentParserfrom dspy import Example-from knowledge_storm import (- STORMWikiRunnerArguments,- STORMWikiRunner,- STORMWikiLMConfigs,-)+from knowledge_storm import STORMWikiRunnerArguments, STORMWikiRunner, STORMWikiLMConfigsfrom knowledge_storm.lm import VLLMClientfrom knowledge_storm.rm import YouRM, BingSearchfrom knowledge_storm.utils import load_api_keydef main(args):- load_api_key(toml_file_path="secrets.toml")+ load_api_key(toml_file_path='secrets.toml')lm_configs = STORMWikiLMConfigs()mistral_kwargs = {"model": "mistralai/Mistral-7B-Instruct-v0.2","port": args.port,"url": args.url,- "stop": (- "\n\n---",- ), # dspy uses "\n\n---" to separate examples. Open models sometimes generate this.+ "stop": ('\n\n---',) # dspy uses "\n\n---" to separate examples. Open models sometimes generate this.}conv_simulator_lm = VLLMClient(max_tokens=500, **mistral_kwargs)@@ -66,12 +59,10 @@ def main(args):# STORM is a knowledge curation system which consumes information from the retrieval module.# Currently, the information source is the Internet and we use search engine API as the retrieval module.- if args.retriever == "bing":- rm = BingSearch(- bing_search_api=os.getenv("BING_SEARCH_API_KEY"), k=engine_args.search_top_k- )- elif args.retriever == "you":- rm = YouRM(ydc_api_key=os.getenv("YDC_API_KEY"), k=engine_args.search_top_k)+ if args.retriever == 'bing':+ rm = BingSearch(bing_search_api=os.getenv('BING_SEARCH_API_KEY'), k=engine_args.search_top_k)+ elif args.retriever == 'you':+ rm = YouRM(ydc_api_key=os.getenv('YDC_API_KEY'), k=engine_args.search_top_k)runner = STORMWikiRunner(engine_args, lm_configs, rm)@@ -83,28 +74,26 @@ def main(args):find_related_topic_example = Example(topic="Knowledge Curation",related_topics="https://en.wikipedia.org/wiki/Knowledge_management\n"- "https://en.wikipedia.org/wiki/Information_science\n"- "https://en.wikipedia.org/wiki/Library_science\n",+ "https://en.wikipedia.org/wiki/Information_science\n"+ "https://en.wikipedia.org/wiki/Library_science\n")gen_persona_example = Example(topic="Knowledge Curation",examples="Title: Knowledge management\n"- "Table of Contents: History\nResearch\n Dimensions\n Strategies\n Motivations\nKM technologies"- "\nKnowledge barriers\nKnowledge retention\nKnowledge audit\nKnowledge protection\n"- " Knowledge protection methods\n Formal methods\n Informal methods\n"- " Balancing knowledge protection and knowledge sharing\n Knowledge protection risks",+ "Table of Contents: History\nResearch\n Dimensions\n Strategies\n Motivations\nKM technologies"+ "\nKnowledge barriers\nKnowledge retention\nKnowledge audit\nKnowledge protection\n"+ " Knowledge protection methods\n Formal methods\n Informal methods\n"+ " Balancing knowledge protection and knowledge sharing\n Knowledge protection risks",personas="1. Historian of Knowledge Systems: This editor will focus on the history and evolution of knowledge curation. They will provide context on how knowledge curation has changed over time and its impact on modern practices.\n"- "2. Information Science Professional: With insights from 'Information science', this editor will explore the foundational theories, definitions, and philosophy that underpin knowledge curation\n"- "3. Digital Librarian: This editor will delve into the specifics of how digital libraries operate, including software, metadata, digital preservation.\n"- "4. Technical expert: This editor will focus on the technical aspects of knowledge curation, such as common features of content management systems.\n"- "5. Museum Curator: The museum curator will contribute expertise on the curation of physical items and the transition of these practices into the digital realm.",+ "2. Information Science Professional: With insights from 'Information science', this editor will explore the foundational theories, definitions, and philosophy that underpin knowledge curation\n"+ "3. Digital Librarian: This editor will delve into the specifics of how digital libraries operate, including software, metadata, digital preservation.\n"+ "4. Technical expert: This editor will focus on the technical aspects of knowledge curation, such as common features of content management systems.\n"+ "5. Museum Curator: The museum curator will contribute expertise on the curation of physical items and the transition of these practices into the digital realm.")runner.storm_knowledge_curation_module.persona_generator.create_writer_with_persona.find_related_topic.demos = [- find_related_topic_example- ]+ find_related_topic_example]runner.storm_knowledge_curation_module.persona_generator.create_writer_with_persona.gen_persona.demos = [- gen_persona_example- ]+ gen_persona_example]# A trade-off of adding one-shot example is that it will increase the input length of the prompt. Also, some# examples may be very long (e.g., an example for writing a section based on the given information), which may@@ -116,28 +105,24 @@ def main(args):topic="Example Topic",conv="Wikipedia Writer: ...\nExpert: ...\nWikipedia Writer: ...\nExpert: ...",old_outline="# Section 1\n## Subsection 1\n## Subsection 2\n"- "# Section 2\n## Subsection 1\n## Subsection 2\n"- "# Section 3",+ "# Section 2\n## Subsection 1\n## Subsection 2\n"+ "# Section 3",outline="# New Section 1\n## New Subsection 1\n## New Subsection 2\n"- "# New Section 2\n"- "# New Section 3\n## New Subsection 1\n## New Subsection 2\n## New Subsection 3",+ "# New Section 2\n"+ "# New Section 3\n## New Subsection 1\n## New Subsection 2\n## New Subsection 3")- runner.storm_outline_generation_module.write_outline.write_page_outline.demos = [- write_page_outline_example- ]+ runner.storm_outline_generation_module.write_outline.write_page_outline.demos = [write_page_outline_example]write_section_example = Example(info="[1]\nInformation in document 1\n[2]\nInformation in document 2\n[3]\nInformation in document 3",topic="Example Topic",section="Example Section",output="# Example Topic\n## Subsection 1\n"- "This is an example sentence [1]. This is another example sentence [2][3].\n"- "## Subsection 2\nThis is one more example sentence [1].",+ "This is an example sentence [1]. This is another example sentence [2][3].\n"+ "## Subsection 2\nThis is one more example sentence [1].")- runner.storm_article_generation.section_gen.write_section.demos = [- write_section_example- ]+ runner.storm_article_generation.section_gen.write_section.demos = [write_section_example]- topic = input("Topic: ")+ topic = input('Topic: ')runner.run(topic=topic,do_research=args.do_research,@@ -149,87 +134,42 @@ def main(args):runner.summary()-if __name__ == "__main__":+if __name__ == '__main__':parser = ArgumentParser()# global arguments- parser.add_argument(- "--url", type=str, default="http://localhost", help="URL of the VLLM server."- )- parser.add_argument(- "--port", type=int, default=8000, help="Port of the VLLM server."- )- parser.add_argument(- "--output-dir",- type=str,- default="./results/mistral_7b",- help="Directory to store the outputs.",- )- parser.add_argument(- "--max-thread-num",- type=int,- default=3,- help="Maximum number of threads to use. The information seeking part and the article generation"- "part can speed up by using multiple threads. Consider reducing it if keep getting "- '"Exceed rate limit" error when calling LM API.',- )- parser.add_argument(- "--retriever",- type=str,- choices=["bing", "you"],- help="The search engine API to use for retrieving information.",- )+ parser.add_argument('--url', type=str, default='http://localhost',+ help='URL of the VLLM server.')+ parser.add_argument('--port', type=int, default=8000,+ help='Port of the VLLM server.')+ parser.add_argument('--output-dir', type=str, default='./results/mistral_7b',+ help='Directory to store the outputs.')+ parser.add_argument('--max-thread-num', type=int, default=3,+ help='Maximum number of threads to use. The information seeking part and the article generation'+ 'part can speed up by using multiple threads. Consider reducing it if keep getting '+ '"Exceed rate limit" error when calling LM API.')+ parser.add_argument('--retriever', type=str, choices=['bing', 'you'],+ help='The search engine API to use for retrieving information.')# stage of the pipeline- parser.add_argument(- "--do-research",- action="store_true",- help="If True, simulate conversation to research the topic; otherwise, load the results.",- )- parser.add_argument(- "--do-generate-outline",- action="store_true",- help="If True, generate an outline for the topic; otherwise, load the results.",- )- parser.add_argument(- "--do-generate-article",- action="store_true",- help="If True, generate an article for the topic; otherwise, load the results.",- )- parser.add_argument(- "--do-polish-article",- action="store_true",- help="If True, polish the article by adding a summarization section and (optionally) removing "- "duplicate content.",- )+ parser.add_argument('--do-research', action='store_true',+ help='If True, simulate conversation to research the topic; otherwise, load the results.')+ parser.add_argument('--do-generate-outline', action='store_true',+ help='If True, generate an outline for the topic; otherwise, load the results.')+ parser.add_argument('--do-generate-article', action='store_true',+ help='If True, generate an article for the topic; otherwise, load the results.')+ parser.add_argument('--do-polish-article', action='store_true',+ help='If True, polish the article by adding a summarization section and (optionally) removing '+ 'duplicate content.')# hyperparameters for the pre-writing stage- parser.add_argument(- "--max-conv-turn",- type=int,- default=3,- help="Maximum number of questions in conversational question asking.",- )- parser.add_argument(- "--max-perspective",- type=int,- default=3,- help="Maximum number of perspectives to consider in perspective-guided question asking.",- )- parser.add_argument(- "--search-top-k",- type=int,- default=3,- help="Top k search results to consider for each search query.",- )+ parser.add_argument('--max-conv-turn', type=int, default=3,+ help='Maximum number of questions in conversational question asking.')+ parser.add_argument('--max-perspective', type=int, default=3,+ help='Maximum number of perspectives to consider in perspective-guided question asking.')+ parser.add_argument('--search-top-k', type=int, default=3,+ help='Top k search results to consider for each search query.')# hyperparameters for the writing stage- parser.add_argument(- "--retrieve-top-k",- type=int,- default=3,- help="Top k collected references for each section title.",- )- parser.add_argument(- "--remove-duplicate",- action="store_true",- help="If True, remove duplicate content from the article.",- )+ parser.add_argument('--retrieve-top-k', type=int, default=3,+ help='Top k collected references for each section title.')+ parser.add_argument('--remove-duplicate', action='store_true',+ help='If True, remove duplicate content from the article.')- main(parser.parse_args())+ main(parser.parse_args())\ No newline at end of filediff --git a/examples/run_storm_wiki_ollama.py b/examples/run_storm_wiki_ollama.pyindex 76f85dc..35ba99e 100644--- a/examples/run_storm_wiki_ollama.py+++ b/examples/run_storm_wiki_ollama.py@@ -1,8 +1,8 @@"""-STORM Wiki pipeline powered by Mistral-7B-Instruct-v0.2 hosted by VLLM server and You.com search engine.+STORM Wiki pipeline powered by local model hosted by Ollama server and You.com or Bing search engine.You need to set up the following environment variables to run this script:- YDC_API_KEY: You.com API key; or, BING_SEARCH_API_KEY: Bing Search API key-You also need to have a VLLM server running with the Mistral-7B-Instruct-v0.2 model. Specify `--url` and `--port` accordingly.+You also need to have a Ollama server running with the llama3 model or other. Specify `--url`, `--port` and `--model` accordingly.Output will be structured as belowargs.output_dir/@@ -16,32 +16,34 @@ args.output_dir/storm_gen_article_polished.txt # Polished final article (if args.do_polish_article is True)"""import os+import sysfrom argparse import ArgumentParserfrom dspy import Example-from knowledge_storm import STORMWikiRunnerArguments, STORMWikiRunner, STORMWikiLMConfigs-from knowledge_storm.lm import VLLMClient-from knowledge_storm.rm import YouRM, BingSearch-from knowledge_storm.utils import load_api_key+sys.path.append('./src')+from lm import OllamaClient+from rm import YouRM, BingSearch+from storm_wiki.engine import STORMWikiRunnerArguments, STORMWikiRunner, STORMWikiLMConfigs+from utils import load_api_keydef main(args):load_api_key(toml_file_path='secrets.toml')lm_configs = STORMWikiLMConfigs()- mistral_kwargs = {- "model": "mistralai/Mistral-7B-Instruct-v0.2",+ ollama_kwargs = {+ "model": args.model,"port": args.port,"url": args.url,"stop": ('\n\n---',) # dspy uses "\n\n---" to separate examples. Open models sometimes generate this.}- conv_simulator_lm = VLLMClient(max_tokens=500, **mistral_kwargs)- question_asker_lm = VLLMClient(max_tokens=500, **mistral_kwargs)- outline_gen_lm = VLLMClient(max_tokens=400, **mistral_kwargs)- article_gen_lm = VLLMClient(max_tokens=700, **mistral_kwargs)- article_polish_lm = VLLMClient(max_tokens=4000, **mistral_kwargs)+ conv_simulator_lm = OllamaClient(max_tokens=500, **ollama_kwargs)+ question_asker_lm = OllamaClient(max_tokens=500, **ollama_kwargs)+ outline_gen_lm = OllamaClient(max_tokens=400, **ollama_kwargs)+ article_gen_lm = OllamaClient(max_tokens=700, **ollama_kwargs)+ article_polish_lm = OllamaClient(max_tokens=4000, **ollama_kwargs)lm_configs.set_conv_simulator_lm(conv_simulator_lm)lm_configs.set_question_asker_lm(question_asker_lm)@@ -138,10 +140,12 @@ if __name__ == '__main__':parser = ArgumentParser()# global argumentsparser.add_argument('--url', type=str, default='http://localhost',- help='URL of the VLLM server.')- parser.add_argument('--port', type=int, default=8000,- help='Port of the VLLM server.')- parser.add_argument('--output-dir', type=str, default='./results/mistral_7b',+ help='URL of the Ollama server.')+ parser.add_argument('--port', type=int, default=11434,+ help='Port of the Ollama server.')+ parser.add_argument('--model', type=str, default='llama3:latest',+ help='Model of the Ollama server.')+ parser.add_argument('--output-dir', type=str, default='./results/ollama',help='Directory to store the outputs.')parser.add_argument('--max-thread-num', type=int, default=3,help='Maximum number of threads to use. The information seeking part and the article generation'