Merge branch 'main' into feature/pxc/env

modelscope · Oct 8, 2024 · 7e404f1 · 7e404f1
2 parents 7d90257 + 9d7af22
commit 7e404f1
Show file tree

Hide file tree

Showing 12 changed files with 153 additions and 93 deletions.
diff --git a/.github/workflows/unittest.yml b/.github/workflows/unittest.yml
@@ -24,9 +24,9 @@ jobs:
     - name: Install Minimal Dependencies
       run: |
         pip install -q -e .
-    - name: Run import tests
+    - name: Run minimal import tests
       run: |
-        python -c "import agentscope; print(agentscope.__version__)"
+        python tests/minimal.py
     - name: Install Full Dependencies
       run: |
         pip install -q -e .[full]

diff --git a/README.md b/README.md
@@ -35,7 +35,6 @@ Start building LLM-empowered multi-agent applications in an easier way.
 |----------------------------------------------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------|
 | <img src="https://gw.alicdn.com/imgextra/i1/O1CN01hhD1mu1Dd3BWVUvxN_!!6000000000238-2-tps-400-400.png" width="100" height="100"> | <img src="https://img.alicdn.com/imgextra/i2/O1CN01tuJ5971OmAqNg9cOw_!!6000000001747-0-tps-444-460.jpg" width="100" height="100"> |
 
-
 ----
 
 ## News
@@ -187,7 +186,6 @@ the following libraries.
   - <img src="https://img.alicdn.com/imgextra/i3/O1CN01SFL0Gu26nrQBFKXFR_!!6000000007707-2-tps-500-500.png" alt="new" width="30" height="30"/>[Conversation with CodeAct Agent](https://github.com/modelscope/agentscope/blob/main/examples/conversation_with_codeact_agent/)
   - <img src="https://img.alicdn.com/imgextra/i3/O1CN01SFL0Gu26nrQBFKXFR_!!6000000007707-2-tps-500-500.png" alt="new" width="30" height="30"/>[Conversation with Router Agent](https://github.com/modelscope/agentscope/blob/main/examples/conversation_with_router_agent/)
 
-
 - Game
   - [Gomoku](https://github.com/modelscope/agentscope/blob/main/examples/game_gomoku)
   - [Werewolf](https://github.com/modelscope/agentscope/blob/main/examples/game_werewolf)
@@ -236,7 +234,6 @@ optional dependencies. Full list of optional dependencies refers to
 Taking distribution mode as an example, you can install its dependencies
 as follows:
 
-
 #### On Windows
 
 ```bash
@@ -247,14 +244,14 @@ pip install agentscope[distribute]
 ```
 
 #### On Mac & Linux
+
 ```bash
 # From source
 pip install -e .\[distribute\]
 # From pypi
 pip install agentscope\[distribute\]
 ```
 
-
 ## Quick Start
 
 ### Configuration
@@ -391,35 +388,36 @@ pre-commit install
 
 Please refer to our [Contribution Guide](https://modelscope.github.io/agentscope/en/tutorial/302-contribute.html) for more details.
 
-## References
-
-If you find our work helpful for your research or application, please
-cite [our paper](https://arxiv.org/abs/2402.14034):
-
-```
-@article{agentscope,
-  author  = {Dawei Gao and
-             Zitao Li and
-             Xuchen Pan and
-             Weirui Kuang and
-             Zhijian Ma and
-             Bingchen Qian and
-             Fei Wei and
-             Wenhao Zhang and
-             Yuexiang Xie and
-             Daoyuan Chen and
-             Liuyi Yao and
-             Hongyi Peng and
-             Ze Yu Zhang and
-             Lin Zhu and
-             Chen Cheng and
-             Hongzhu Shi and
-             Yaliang Li and
-             Bolin Ding and
-             Jingren Zhou},
-  title   = {AgentScope: A Flexible yet Robust Multi-Agent Platform},
-  journal = {CoRR},
-  volume  = {abs/2402.14034},
-  year    = {2024},
-}
-```
+## Publications
+
+If you find our work helpful for your research or application, please cite our papers.
+
+1. [AgentScope: A Flexible yet Robust Multi-Agent Platform](https://arxiv.org/abs/2402.14034)
+
+    ```
+    @article{agentscope,
+        author  = {Dawei Gao and
+                   Zitao Li and
+                   Xuchen Pan and
+                   Weirui Kuang and
+                   Zhijian Ma and
+                   Bingchen Qian and
+                   Fei Wei and
+                   Wenhao Zhang and
+                   Yuexiang Xie and
+                   Daoyuan Chen and
+                   Liuyi Yao and
+                   Hongyi Peng and
+                   Ze Yu Zhang and
+                   Lin Zhu and
+                   Chen Cheng and
+                   Hongzhu Shi and
+                   Yaliang Li and
+                   Bolin Ding and
+                   Jingren Zhou}
+        title   = {AgentScope: A Flexible yet Robust Multi-Agent Platform},
+        journal = {CoRR},
+        volume  = {abs/2402.14034},
+        year    = {2024},
+    }
+    ```
diff --git a/README_ZH.md b/README_ZH.md
@@ -35,8 +35,6 @@
 |---------|----------|
 | <img src="https://gw.alicdn.com/imgextra/i1/O1CN01hhD1mu1Dd3BWVUvxN_!!6000000000238-2-tps-400-400.png" width="100" height="100"> | <img src="https://img.alicdn.com/imgextra/i2/O1CN01tuJ5971OmAqNg9cOw_!!6000000001747-0-tps-444-460.jpg" width="100" height="100"> |
 
-
-
 ----
 
 ## 新闻
@@ -56,7 +54,6 @@
 <img src="https://github.com/user-attachments/assets/dfffbd1e-1fe7-49ee-ac11-902415b2b0d6" width="45%" alt="agentscope-logo">
 </h5>
 
-
 - <img src="https://img.alicdn.com/imgextra/i3/O1CN01SFL0Gu26nrQBFKXFR_!!6000000007707-2-tps-500-500.png" alt="new" width="30" height="30"/>**[2024-07-15]** AgentScope 中添加了 Mixture of Agents 算法。使用样例请参考 [MoA 示例](https://github.com/modelscope/agentscope/blob/main/examples/conversation_mixture_of_agents)。
 
 - **[2024-06-14]** 新的提示调优（Prompt tuning）模块已经上线 AgentScope，用以帮助开发者生成和优化智能体的 system prompt。更多的细节和使用样例请参考 AgentScope [教程](https://modelscope.github.io/agentscope/en/tutorial/209-prompt_opt.html)！
@@ -232,6 +229,7 @@ pip install agentscope[distribute]
 ```
 
 #### On Mac & Linux
+
 ```bash
 # From source
 pip install -e .\[distribute\]
@@ -362,34 +360,36 @@ pre-commit install
 
 请参阅我们的[贡献指南](https://modelscope.github.io/agentscope/zh_CN/tutorial/302-contribute.html)了解更多细节。
 
-## 引用
-
-如果您觉得我们的工作对您的研究或应用有帮助，请引用[我们的论文](https://arxiv.org/abs/2402.14034)。
-
-```
-@article{agentscope,
-  author  = {Dawei Gao and
-             Zitao Li and
-             Xuchen Pan and
-             Weirui Kuang and
-             Zhijian Ma and
-             Bingchen Qian and
-             Fei Wei and
-             Wenhao Zhang and
-             Yuexiang Xie and
-             Daoyuan Chen and
-             Liuyi Yao and
-             Hongyi Peng and
-             Zeyu Zhang and
-             Lin Zhu and
-             Chen Cheng and
-             Hongzhu Shi and
-             Yaliang Li and
-             Bolin Ding and
-             Jingren Zhou},
-  title   = {AgentScope: A Flexible yet Robust Multi-Agent Platform},
-  journal = {CoRR},
-  volume  = {abs/2402.14034},
-  year    = {2024},
-}
-```
+## 发表
+
+如果您觉得我们的工作对您的研究或应用有帮助，请引用如下论文
+
+1. [AgentScope: A Flexible yet Robust Multi-Agent Platform](https://arxiv.org/abs/2402.14034)
+
+    ```
+    @article{agentscope,
+        author  = {Dawei Gao and
+                   Zitao Li and
+                   Xuchen Pan and
+                   Weirui Kuang and
+                   Zhijian Ma and
+                   Bingchen Qian and
+                   Fei Wei and
+                   Wenhao Zhang and
+                   Yuexiang Xie and
+                   Daoyuan Chen and
+                   Liuyi Yao and
+                   Hongyi Peng and
+                   Ze Yu Zhang and
+                   Lin Zhu and
+                   Chen Cheng and
+                   Hongzhu Shi and
+                   Yaliang Li and
+                   Bolin Ding and
+                   Jingren Zhou}
+        title   = {AgentScope: A Flexible yet Robust Multi-Agent Platform},
+        journal = {CoRR},
+        volume  = {abs/2402.14034},
+        year    = {2024},
+    }
+    ```
diff --git a/docs/sphinx_doc/en/source/tutorial/104-usecase.md b/docs/sphinx_doc/en/source/tutorial/104-usecase.md
@@ -291,7 +291,7 @@ With the game logic and agents set up, you're ready to run the Werewolf game. By
 
 ```bash
 cd examples/game_werewolf
-python main.py  # Assuming the pipeline is implemented in main.py
+python werewolf.py  # Assuming the pipeline is implemented in werewolf.py
 ```
 
 It is recommended that you start the game in [AgentScope Studio](https://modelscope.github.io/agentscope/en/tutorial/209-gui.html), where you

diff --git a/docs/sphinx_doc/zh_CN/source/tutorial/104-usecase.md b/docs/sphinx_doc/zh_CN/source/tutorial/104-usecase.md
@@ -170,7 +170,7 @@ for i in range(1, MAX_GAME_ROUND + 1):
     # Night phase: werewolves discuss
     hint = HostMsg(content=Prompts.to_wolves.format(n2s(wolves)))
     with msghub(wolves, announcement=hint) as hub:
-      	set_parsers(wolves, Prompts.wolves_discuss_parser)
+        set_parsers(wolves, Prompts.wolves_discuss_parser)
         for _ in range(MAX_WEREWOLF_DISCUSSION_ROUND):
             x = sequentialpipeline(wolves)
             if x.metadata.get("finish_discussion", False):
@@ -295,7 +295,7 @@ for i in range(1, MAX_GAME_ROUND + 1):
 
 ```bash
 cd examples/game_werewolf
-python main.py  # Assuming the pipeline is implemented in main.py
+python werewolf.py  # Assuming the pipeline is implemented in werewolf.py
 ```
 
 建议您在在 [AgentScope Studio](https://modelscope.github.io/agentscope/zh_CN/tutorial/209-gui.html) 中启动游戏，在对应的链接中您将看到下面的内容输出。

diff --git a/docs/sphinx_doc/zh_CN/source/tutorial/209-gui.md b/docs/sphinx_doc/zh_CN/source/tutorial/209-gui.md
@@ -70,7 +70,7 @@ agentscope.init(
     # ...
     project="xxx",
     name="xxx",
-    studio_url="http://127.0.0.1：5000"          # AgentScope Studio 的 URL
+    studio_url="http://127.0.0.1:5000"          # AgentScope Studio 的 URL
 )
 ```
 

diff --git a/examples/paper_llm_based_algorithm/README.md b/examples/paper_llm_based_algorithm/README.md
@@ -1,6 +1,5 @@
 # LLM-based algorithms
 
-
 This folder contains the source code for reproducing the experiment results in our arXiv preprint "On the Design and Analysis of LLM-Based Algorithms".
 
 Our work initiates a formal investigation into the design and analysis of LLM-based algorithms,
@@ -11,7 +10,6 @@ Within this folder, you can find our implementation for the key abstractions,
 the LLM-based algorithms in four concrete examples,
 and the experiments for validating our analysis in the manuscript.
 
-
 ## Tested Models
 
 The following models have been tested, which are also listed in `model_configs.json`:
@@ -20,26 +18,25 @@ GPT-3.5 Turbo,
 Llama3-8B (with ollama),
 Llama3-70B (with vLLM).
 
-
 ## Prerequisites
 
-
 1. Install AgentScope from source with `pip`, according to the [official instruction](../../README.md).
 2. Install matplotlib: `pip install matplotlib`.
-3. Change directory: `cd examples/llm_based_algorithm`.
+3. Change directory: `cd examples/paper_llm_based_algorithm`.
 4. Set up LLM model configs in `model_configs.json`.
 
-
-
 ## Usage
 
 ### Run experiments
 
 To run experiments for a certain task:
+
 ```bash
 bash ./scripts/exp_{task}.sh
 ```
+
 or copy a piece of scripts therein, modify the parameters, and run it in the terminal, for example:
+
 ```bash
 python3 run_exp_single_variable.py \
     --task counting \
@@ -52,6 +49,7 @@ python3 run_exp_single_variable.py \
 ```
 
 Parameters:
+
 - `task`: name of the task, {"counting", "sorting", "retrieval", "retrieval_no_needle", "rag"}.
 - `llm_model`: name of the LLM model, i.e. `config_name` in `model_configs.json`.
 - `variable_name`: "n" for problem size, or "m" for sub-task size.
@@ -60,30 +58,37 @@ Parameters:
 - `save_results`: if `True`, experiment results will be saved to `./out`; otherwise, results will be plotted and shown at the end of the experiment, and won't be saved.
 - `ntrials`: number of independent trials for each experiment config, i.e. each entry of `lst_variable`.
 
-
 ### Plot results
 
 To plot experiment results that have been saved:
+
 ```bash
 bash ./scripts/plot_{task}.sh
 ```
+
 or copy a piece of scripts therein and run it in the terminal, for example:
+
 ```bash
 python3 plot_exp_results.py \
     --folder ./out/counting/exp_counting_vary_n_model_ollama_llama3_8b-2024-06-19-11-11-13-kkwrhc
 ```
+
 The path to the experiment results need to be replaced with the actual one generated during your own experiment.
 The generated figures will be saved to the same folder.
 
-
 ## Reference
 
 For more details, please refer to our arXiv preprint:
+
 ```
-@article{chen2024llmbasedalgorithms,
-      title={On the Design and Analysis of LLM-Based Algorithms},
-      author={Yanxi Chen and Yaliang Li and Bolin Ding and Jingren Zhou},
-      year={2024},
+@article{llm_based_algorithms,
+  author       = {Yanxi Chen and
+                  Yaliang Li and
+                  Bolin Ding and
+                  Jingren Zhou},
+  title        = {On the Design and Analysis of LLM-Based Algorithms},
+  journal      = {CoRR},
+  volume       = {abs/2407.14788},
+  year         = {2024},
 }
 ```
-
diff --git a/src/agentscope/models/ollama_model.py b/src/agentscope/models/ollama_model.py
@@ -359,7 +359,7 @@ def format(
         system_content = "\n".join(system_content_template)
 
         system_message = {
-            "role": "system",
+            "role": "user",
             "content": system_content,
         }
 

diff --git a/src/agentscope/models/post_model.py b/src/agentscope/models/post_model.py
@@ -8,7 +8,6 @@
 import requests
 from loguru import logger
 
-from .gemini_model import GeminiChatWrapper
 from .openai_model import OpenAIChatWrapper
 from .model import ModelWrapperBase, ModelResponse
 from ..constants import _DEFAULT_MAX_RETRIES
@@ -221,6 +220,8 @@ def format(
 
         # Gemini
         elif model_name and model_name.startswith("gemini"):
+            from .gemini_model import GeminiChatWrapper
+
             return GeminiChatWrapper.format(*args)
 
         # Include DashScope, ZhipuAI, Ollama, the other models supported by

diff --git a/tests/format_test.py b/tests/format_test.py
@@ -283,7 +283,7 @@ def test_ollama_chat(self) -> None:
         # correct format
         ground_truth = [
             {
-                "role": "system",
+                "role": "user",
                 "content": (
                     "You are a helpful assistant\n"
                     "\n"