search

agiresearch / OpenAGI

OpenAGI: When LLM Meets Domain Experts
MIT License
1.82k stars 151 forks source link

feat: add more tool-calling enabled agents #67

Closed dongyuanjushi closed 1 month ago

dongyuanjushi commented 1 month ago

  1. add AcademicAgent with arxiv tool calling enabled

    • Task input example is: Summarize recent advancements in quantum computing from the past five years.
    • Tool calling example is: [AcademicAgent] It starts to call external tools [AcademicAgent] The solution to step 1: It will call the arxiv with the params as {'query': 'quantum computing advancements 2016-2021'}. The tool response is Published: 2007-08-02 Title: An Introduction to Quantum Computing Authors: Noson S. Yanofsky Summary: Quantum Computing is a new and exciting field at the intersection of mathematics, computer science and physics. It concerns a utilization of quantum mechanics to improve the efficiency of computation. Here we present a gentle introduction to some of the ideas in quantum computing. The paper begins by motivating the central ideas of quantum mechanics and quantum computation with simple toy models. From there we move on to a formal presentation of the small fraction of (finite dimensional) quantum mechanics that we will need for basic quantum computation. Central notions of quantum architecture (qubits and quantum gates) are described. The paper ends with a presentation of one of the simplest quantum algorithms: Deutsch's algorithm. Our presentation demands neither advanced mathematics nor advanced physics.

  2. add CreationAgent with text2image generation tool calling enabled

    • Task input example is: Create an image of a lush jungle with an ancient temple, evoking a sense of mystery and adventure.
    • Tool calling example is: [CreationAgent] It starts to call external tools Token indices sequence length is longer than the specified maximum sequence length for this model (91 > 77). Running this sequence through the model will result in indexing errors The following part of your input was truncated because CLIP can only handle sequences up to 77 tokens: ['somewhat eerie ambiance hints at hidden treasures and adventures that lie ahead.'] Token indices sequence length is longer than the specified maximum sequence length for this model (91 > 77). Running this sequence through the model will result in indexing errors The following part of your input was truncated because CLIP can only handle sequences up to 77 tokens: ['somewhat eerie ambiance hints at hidden treasures and adventures that lie ahead.'] 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00, 1.89it/s] demo

  3. revise RecAgent with IMDB rank calling enabled

    • Task input example is: Recommend two movies with groundbreaking visual effects released in the last fifteen years ranked between 1 and 20 with ratings above 8.0.
    • Tool calling example is: [RecAgent] It starts to call external tools [RecAgent] The solution to step 1: It will call the imdb_top_movies with the params as {'start': '1', 'end': '20'}. The tool response is Top 1-20 series ranked by IMDB are: The Godfather, ['Crime', 'Drama'], 9.2, published in 1972;The Dark Knight, ['Action', 'Crime', 'Drama'], 9.0, published in 2008;The Godfather Part II, ['Crime', 'Drama'], 9.0, published in 1974;12 Angry Men, ['Crime', 'Drama'], 9.0, published in 1957;Schindler's List, ['Biography', 'Drama', 'History'], 9.0, published in 1993;The Lord of the Rings: The Return of the King, ['Action', 'Adventure', 'Drama'], 9.0, published in 2003;Pulp Fiction, ['Crime', 'Drama'], 8.9, published in 1994;The Lord of the Rings: The Fellowship of the Ring, ['Action', 'Adventure', 'Drama'], 8.8, published in 2001;The Good, the Bad and the Ugly, ['Adventure', 'Western'], 8.8, published in 1966;Forrest Gump, ['Drama', 'Romance'], 8.8, published in 1994;Fight Club, ['Drama'], 8.8, published in 1999;The Lord of the Rings: The Two Towers, ['Action', 'Adventure', 'Drama'], 8.8, published in 2002;Inception, ['Action', 'Adventure', 'Sci-Fi'], 8.7, published in 2010;Star Wars: Episode V - The Empire Strikes Back, ['Action', 'Adventure', 'Fantasy'], 8.7, published in 1980;The Matrix, ['Action', 'Sci-Fi'], 8.7, published in 1999;Goodfellas, ['Biography', 'Crime', 'Drama'], 8.7, published in 1990;Spider-Man: Across the Spider-Verse, ['Animation', 'Action', 'Adventure'], 8.7, published in 2023;Interstellar, ['Adventure', 'Drama', 'Sci-Fi'], 8.7, published in 2014;One Flew Over the Cuckoo's Nest, ['Drama'], 8.6, published in 1975