新增对话处理功能，优化意图识别逻辑，添加结果保存至Excel的功能，更新依赖项以支持新的数据库驱动和ORM，重构代码以提高可读性和维护性，删除冗余文件以简化项目结构。
This commit is contained in:
+105
@@ -0,0 +1,105 @@
|
||||
# Role Definition
|
||||
|
||||
- You are a **Python master**, a highly experienced **tutor**, a **world-renowned ML engineer**, and a **talented data scientist**.
|
||||
- You possess exceptional coding skills and a deep understanding of Python's best practices, design patterns, and idioms.
|
||||
- You are adept at identifying and preventing potential errors, and you prioritize writing efficient and maintainable code.
|
||||
- You are skilled in explaining complex concepts in a clear and concise manner, making you an effective mentor and educator.
|
||||
- You are recognized for your contributions to the field of machine learning and have a strong track record of developing and deploying successful ML models.
|
||||
- As a talented data scientist, you excel at data analysis, visualization, and deriving actionable insights from complex datasets.
|
||||
|
||||
# Technology Stack
|
||||
|
||||
- **Python Version:** Python 3.10+
|
||||
- **Dependency Management:** Poetry / Rye
|
||||
- **Code Formatting:** Ruff (replaces `black`, `isort`, `flake8`)
|
||||
- **Type Hinting:** Strictly use the `typing` module. All functions, methods, and class members must have type annotations.
|
||||
- **Testing Framework:** `pytest`
|
||||
- **Documentation:** Google style docstring
|
||||
- **Environment Management:** `conda` / `venv`
|
||||
- **Containerization:** `docker`, `docker-compose`
|
||||
- **Asynchronous Programming:** Prefer `async` and `await`
|
||||
- **Web Framework:** `fastapi`
|
||||
- **Demo Framework:** `gradio`, `streamlit`
|
||||
- **LLM Framework:** `langchain`, `transformers`
|
||||
- **Vector Database:** `faiss`, `chroma` (optional)
|
||||
- **Experiment Tracking:** `mlflow`, `tensorboard` (optional)
|
||||
- **Hyperparameter Optimization:** `optuna`, `hyperopt` (optional)
|
||||
- **Data Processing:** `pandas`, `numpy`, `dask` (optional), `pyspark` (optional)
|
||||
- **Version Control:** `git`
|
||||
- **Server:** `gunicorn`, `uvicorn` (with `nginx` or `caddy`)
|
||||
- **Process Management:** `systemd`, `supervisor`
|
||||
|
||||
# Coding Guidelines
|
||||
|
||||
## 1. Pythonic Practices
|
||||
|
||||
- **Elegance and Readability:** Strive for elegant and Pythonic code that is easy to understand and maintain.
|
||||
- **PEP 8 Compliance:** Adhere to PEP 8 guidelines for code style, with Ruff as the primary linter and formatter.
|
||||
- **Explicit over Implicit:** Favor explicit code that clearly communicates its intent over implicit, overly concise code.
|
||||
- **Zen of Python:** Keep the Zen of Python in mind when making design decisions.
|
||||
|
||||
## 2. Modular Design
|
||||
|
||||
- **Single Responsibility Principle:** Each module/file should have a well-defined, single responsibility.
|
||||
- **Reusable Components:** Develop reusable functions and classes, favoring composition over inheritance.
|
||||
- **Package Structure:** Organize code into logical packages and modules.
|
||||
|
||||
## 3. Code Quality
|
||||
|
||||
- **Comprehensive Type Annotations:** All functions, methods, and class members must have type annotations, using the most specific types possible.
|
||||
- **Detailed Docstrings:** All functions, methods, and classes must have Google-style docstrings, thoroughly explaining their purpose, parameters, return values, and any exceptions raised. Include usage examples where helpful.
|
||||
- **Thorough Unit Testing:** Aim for high test coverage (90% or higher) using `pytest`. Test both common cases and edge cases.
|
||||
- **Robust Exception Handling:** Use specific exception types, provide informative error messages, and handle exceptions gracefully. Implement custom exception classes when needed. Avoid bare `except` clauses.
|
||||
- **Logging:** Employ the `logging` module judiciously to log important events, warnings, and errors.
|
||||
|
||||
## 4. ML/AI Specific Guidelines
|
||||
|
||||
- **Experiment Configuration:** Use `hydra` or `yaml` for clear and reproducible experiment configurations.
|
||||
- **Data Pipeline Management:** Employ scripts or tools like `dvc` to manage data preprocessing and ensure reproducibility.
|
||||
- **Model Versioning:** Utilize `git-lfs` or cloud storage to track and manage model checkpoints effectively.
|
||||
- **Experiment Logging:** Maintain comprehensive logs of experiments, including parameters, results, and environmental details.
|
||||
- **LLM Prompt Engineering:** Dedicate a module or files for managing Prompt templates with version control.
|
||||
- **Context Handling:** Implement efficient context management for conversations, using suitable data structures like deques.
|
||||
|
||||
## 5. Performance Optimization
|
||||
|
||||
- **Asynchronous Programming:** Leverage `async` and `await` for I/O-bound operations to maximize concurrency.
|
||||
- **Caching:** Apply `functools.lru_cache`, `@cache` (Python 3.9+), or `fastapi.Depends` caching where appropriate.
|
||||
- **Resource Monitoring:** Use `psutil` or similar to monitor resource usage and identify bottlenecks.
|
||||
- **Memory Efficiency:** Ensure proper release of unused resources to prevent memory leaks.
|
||||
- **Concurrency:** Employ `concurrent.futures` or `asyncio` to manage concurrent tasks effectively.
|
||||
- **Database Best Practices:** Design database schemas efficiently, optimize queries, and use indexes wisely.
|
||||
|
||||
## 6. API Development with FastAPI
|
||||
|
||||
- **Data Validation:** Use Pydantic models for rigorous request and response data validation.
|
||||
- **Dependency Injection:** Effectively use FastAPI's dependency injection for managing dependencies.
|
||||
- **Routing:** Define clear and RESTful API routes using FastAPI's `APIRouter`.
|
||||
- **Background Tasks:** Utilize FastAPI's `BackgroundTasks` or integrate with Celery for background processing.
|
||||
- **Security:** Implement robust authentication and authorization (e.g., OAuth 2.0, JWT).
|
||||
- **Documentation:** Auto-generate API documentation using FastAPI's OpenAPI support.
|
||||
- **Versioning:** Plan for API versioning from the start (e.g., using URL prefixes or headers).
|
||||
- **CORS:** Configure Cross-Origin Resource Sharing (CORS) settings correctly.
|
||||
|
||||
# Code Example Requirements
|
||||
|
||||
- All functions must include type annotations.
|
||||
- Must provide clear, Google-style docstrings.
|
||||
- Key logic should be annotated with comments.
|
||||
- Provide usage examples (e.g., in the `tests/` directory or as a `__main__` section).
|
||||
- Include error handling.
|
||||
- Use `ruff` for code formatting.
|
||||
|
||||
# Others
|
||||
|
||||
- **Prioritize new features in Python 3.10+.**
|
||||
- **When explaining code, provide clear logical explanations and code comments.**
|
||||
- **When making suggestions, explain the rationale and potential trade-offs.**
|
||||
- **If code examples span multiple files, clearly indicate the file name.**
|
||||
- **Do not over-engineer solutions. Strive for simplicity and maintainability while still being efficient.**
|
||||
- **Favor modularity, but avoid over-modularization.**
|
||||
- **Use the most modern and efficient libraries when appropriate, but justify their use and ensure they don't add unnecessary complexity.**
|
||||
- **When providing solutions or examples, ensure they are self-contained and executable without requiring extensive modifications.**
|
||||
- **If a request is unclear or lacks sufficient information, ask clarifying questions before proceeding.**
|
||||
- **Always consider the security implications of your code, especially when dealing with user inputs and external data.**
|
||||
- **Actively use and promote best practices for the specific tasks at hand (LLM app development, data cleaning, demo creation, etc.).**
|
||||
Generated
+79
-59
@@ -2230,6 +2230,26 @@ type = "legacy"
|
||||
url = "http://mirrors.aliyun.com/pypi/simple"
|
||||
reference = "ali-mirrors"
|
||||
|
||||
[[package]]
|
||||
name = "pymysql"
|
||||
version = "1.1.1"
|
||||
description = "Pure Python MySQL Driver"
|
||||
optional = false
|
||||
python-versions = ">=3.7"
|
||||
files = [
|
||||
{file = "PyMySQL-1.1.1-py3-none-any.whl", hash = "sha256:4de15da4c61dc132f4fb9ab763063e693d521a80fd0e87943b9a453dd4c19d6c"},
|
||||
{file = "pymysql-1.1.1.tar.gz", hash = "sha256:e127611aaf2b417403c60bf4dc570124aeb4a57f5f37b8e95ae399a42f904cd0"},
|
||||
]
|
||||
|
||||
[package.extras]
|
||||
ed25519 = ["PyNaCl (>=1.4.0)"]
|
||||
rsa = ["cryptography"]
|
||||
|
||||
[package.source]
|
||||
type = "legacy"
|
||||
url = "http://mirrors.aliyun.com/pypi/simple"
|
||||
reference = "ali-mirrors"
|
||||
|
||||
[[package]]
|
||||
name = "python-dateutil"
|
||||
version = "2.9.0.post0"
|
||||
@@ -2579,68 +2599,68 @@ reference = "ali-mirrors"
|
||||
|
||||
[[package]]
|
||||
name = "sqlalchemy"
|
||||
version = "2.0.40"
|
||||
version = "2.0.41"
|
||||
description = "Database Abstraction Library"
|
||||
optional = false
|
||||
python-versions = ">=3.7"
|
||||
files = [
|
||||
{file = "SQLAlchemy-2.0.40-cp37-cp37m-macosx_10_9_x86_64.whl", hash = "sha256:ae9597cab738e7cc823f04a704fb754a9249f0b6695a6aeb63b74055cd417a96"},
|
||||
{file = "SQLAlchemy-2.0.40-cp37-cp37m-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:37a5c21ab099a83d669ebb251fddf8f5cee4d75ea40a5a1653d9c43d60e20867"},
|
||||
{file = "SQLAlchemy-2.0.40-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:bece9527f5a98466d67fb5d34dc560c4da964240d8b09024bb21c1246545e04e"},
|
||||
{file = "SQLAlchemy-2.0.40-cp37-cp37m-musllinux_1_2_aarch64.whl", hash = "sha256:8bb131ffd2165fae48162c7bbd0d97c84ab961deea9b8bab16366543deeab625"},
|
||||
{file = "SQLAlchemy-2.0.40-cp37-cp37m-musllinux_1_2_x86_64.whl", hash = "sha256:9408fd453d5f8990405cc9def9af46bfbe3183e6110401b407c2d073c3388f47"},
|
||||
{file = "SQLAlchemy-2.0.40-cp37-cp37m-win32.whl", hash = "sha256:00a494ea6f42a44c326477b5bee4e0fc75f6a80c01570a32b57e89cf0fbef85a"},
|
||||
{file = "SQLAlchemy-2.0.40-cp37-cp37m-win_amd64.whl", hash = "sha256:c7b927155112ac858357ccf9d255dd8c044fd9ad2dc6ce4c4149527c901fa4c3"},
|
||||
{file = "sqlalchemy-2.0.40-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:f1ea21bef99c703f44444ad29c2c1b6bd55d202750b6de8e06a955380f4725d7"},
|
||||
{file = "sqlalchemy-2.0.40-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:afe63b208153f3a7a2d1a5b9df452b0673082588933e54e7c8aac457cf35e758"},
|
||||
{file = "sqlalchemy-2.0.40-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:a8aae085ea549a1eddbc9298b113cffb75e514eadbb542133dd2b99b5fb3b6af"},
|
||||
{file = "sqlalchemy-2.0.40-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:5ea9181284754d37db15156eb7be09c86e16e50fbe77610e9e7bee09291771a1"},
|
||||
{file = "sqlalchemy-2.0.40-cp310-cp310-musllinux_1_2_aarch64.whl", hash = "sha256:5434223b795be5c5ef8244e5ac98056e290d3a99bdcc539b916e282b160dda00"},
|
||||
{file = "sqlalchemy-2.0.40-cp310-cp310-musllinux_1_2_x86_64.whl", hash = "sha256:15d08d5ef1b779af6a0909b97be6c1fd4298057504eb6461be88bd1696cb438e"},
|
||||
{file = "sqlalchemy-2.0.40-cp310-cp310-win32.whl", hash = "sha256:cd2f75598ae70bcfca9117d9e51a3b06fe29edd972fdd7fd57cc97b4dbf3b08a"},
|
||||
{file = "sqlalchemy-2.0.40-cp310-cp310-win_amd64.whl", hash = "sha256:2cbafc8d39ff1abdfdda96435f38fab141892dc759a2165947d1a8fffa7ef596"},
|
||||
{file = "sqlalchemy-2.0.40-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:f6bacab7514de6146a1976bc56e1545bee247242fab030b89e5f70336fc0003e"},
|
||||
{file = "sqlalchemy-2.0.40-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:5654d1ac34e922b6c5711631f2da497d3a7bffd6f9f87ac23b35feea56098011"},
|
||||
{file = "sqlalchemy-2.0.40-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:35904d63412db21088739510216e9349e335f142ce4a04b69e2528020ee19ed4"},
|
||||
{file = "sqlalchemy-2.0.40-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:9c7a80ed86d6aaacb8160a1caef6680d4ddd03c944d985aecee940d168c411d1"},
|
||||
{file = "sqlalchemy-2.0.40-cp311-cp311-musllinux_1_2_aarch64.whl", hash = "sha256:519624685a51525ddaa7d8ba8265a1540442a2ec71476f0e75241eb8263d6f51"},
|
||||
{file = "sqlalchemy-2.0.40-cp311-cp311-musllinux_1_2_x86_64.whl", hash = "sha256:2ee5f9999a5b0e9689bed96e60ee53c3384f1a05c2dd8068cc2e8361b0df5b7a"},
|
||||
{file = "sqlalchemy-2.0.40-cp311-cp311-win32.whl", hash = "sha256:c0cae71e20e3c02c52f6b9e9722bca70e4a90a466d59477822739dc31ac18b4b"},
|
||||
{file = "sqlalchemy-2.0.40-cp311-cp311-win_amd64.whl", hash = "sha256:574aea2c54d8f1dd1699449f332c7d9b71c339e04ae50163a3eb5ce4c4325ee4"},
|
||||
{file = "sqlalchemy-2.0.40-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:9d3b31d0a1c44b74d3ae27a3de422dfccd2b8f0b75e51ecb2faa2bf65ab1ba0d"},
|
||||
{file = "sqlalchemy-2.0.40-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:37f7a0f506cf78c80450ed1e816978643d3969f99c4ac6b01104a6fe95c5490a"},
|
||||
{file = "sqlalchemy-2.0.40-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:0bb933a650323e476a2e4fbef8997a10d0003d4da996aad3fd7873e962fdde4d"},
|
||||
{file = "sqlalchemy-2.0.40-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:6959738971b4745eea16f818a2cd086fb35081383b078272c35ece2b07012716"},
|
||||
{file = "sqlalchemy-2.0.40-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:110179728e442dae85dd39591beb74072ae4ad55a44eda2acc6ec98ead80d5f2"},
|
||||
{file = "sqlalchemy-2.0.40-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:e8040680eaacdce4d635f12c55c714f3d4c7f57da2bc47a01229d115bd319191"},
|
||||
{file = "sqlalchemy-2.0.40-cp312-cp312-win32.whl", hash = "sha256:650490653b110905c10adac69408380688cefc1f536a137d0d69aca1069dc1d1"},
|
||||
{file = "sqlalchemy-2.0.40-cp312-cp312-win_amd64.whl", hash = "sha256:2be94d75ee06548d2fc591a3513422b873490efb124048f50556369a834853b0"},
|
||||
{file = "sqlalchemy-2.0.40-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:915866fd50dd868fdcc18d61d8258db1bf9ed7fbd6dfec960ba43365952f3b01"},
|
||||
{file = "sqlalchemy-2.0.40-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:4a4c5a2905a9ccdc67a8963e24abd2f7afcd4348829412483695c59e0af9a705"},
|
||||
{file = "sqlalchemy-2.0.40-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:55028d7a3ebdf7ace492fab9895cbc5270153f75442a0472d8516e03159ab364"},
|
||||
{file = "sqlalchemy-2.0.40-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:6cfedff6878b0e0d1d0a50666a817ecd85051d12d56b43d9d425455e608b5ba0"},
|
||||
{file = "sqlalchemy-2.0.40-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:bb19e30fdae77d357ce92192a3504579abe48a66877f476880238a962e5b96db"},
|
||||
{file = "sqlalchemy-2.0.40-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:16d325ea898f74b26ffcd1cf8c593b0beed8714f0317df2bed0d8d1de05a8f26"},
|
||||
{file = "sqlalchemy-2.0.40-cp313-cp313-win32.whl", hash = "sha256:a669cbe5be3c63f75bcbee0b266779706f1a54bcb1000f302685b87d1b8c1500"},
|
||||
{file = "sqlalchemy-2.0.40-cp313-cp313-win_amd64.whl", hash = "sha256:641ee2e0834812d657862f3a7de95e0048bdcb6c55496f39c6fa3d435f6ac6ad"},
|
||||
{file = "sqlalchemy-2.0.40-cp38-cp38-macosx_10_9_x86_64.whl", hash = "sha256:50f5885bbed261fc97e2e66c5156244f9704083a674b8d17f24c72217d29baf5"},
|
||||
{file = "sqlalchemy-2.0.40-cp38-cp38-macosx_11_0_arm64.whl", hash = "sha256:cf0e99cdb600eabcd1d65cdba0d3c91418fee21c4aa1d28db47d095b1064a7d8"},
|
||||
{file = "sqlalchemy-2.0.40-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:fe147fcd85aaed53ce90645c91ed5fca0cc88a797314c70dfd9d35925bd5d106"},
|
||||
{file = "sqlalchemy-2.0.40-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:baf7cee56bd552385c1ee39af360772fbfc2f43be005c78d1140204ad6148438"},
|
||||
{file = "sqlalchemy-2.0.40-cp38-cp38-musllinux_1_2_aarch64.whl", hash = "sha256:4aeb939bcac234b88e2d25d5381655e8353fe06b4e50b1c55ecffe56951d18c2"},
|
||||
{file = "sqlalchemy-2.0.40-cp38-cp38-musllinux_1_2_x86_64.whl", hash = "sha256:c268b5100cfeaa222c40f55e169d484efa1384b44bf9ca415eae6d556f02cb08"},
|
||||
{file = "sqlalchemy-2.0.40-cp38-cp38-win32.whl", hash = "sha256:46628ebcec4f23a1584fb52f2abe12ddb00f3bb3b7b337618b80fc1b51177aff"},
|
||||
{file = "sqlalchemy-2.0.40-cp38-cp38-win_amd64.whl", hash = "sha256:7e0505719939e52a7b0c65d20e84a6044eb3712bb6f239c6b1db77ba8e173a37"},
|
||||
{file = "sqlalchemy-2.0.40-cp39-cp39-macosx_10_9_x86_64.whl", hash = "sha256:c884de19528e0fcd9dc34ee94c810581dd6e74aef75437ff17e696c2bfefae3e"},
|
||||
{file = "sqlalchemy-2.0.40-cp39-cp39-macosx_11_0_arm64.whl", hash = "sha256:1abb387710283fc5983d8a1209d9696a4eae9db8d7ac94b402981fe2fe2e39ad"},
|
||||
{file = "sqlalchemy-2.0.40-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:5cfa124eda500ba4b0d3afc3e91ea27ed4754e727c7f025f293a22f512bcd4c9"},
|
||||
{file = "sqlalchemy-2.0.40-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:8b6b28d303b9d57c17a5164eb1fd2d5119bb6ff4413d5894e74873280483eeb5"},
|
||||
{file = "sqlalchemy-2.0.40-cp39-cp39-musllinux_1_2_aarch64.whl", hash = "sha256:b5a5bbe29c10c5bfd63893747a1bf6f8049df607638c786252cb9243b86b6706"},
|
||||
{file = "sqlalchemy-2.0.40-cp39-cp39-musllinux_1_2_x86_64.whl", hash = "sha256:f0fda83e113bb0fb27dc003685f32a5dcb99c9c4f41f4fa0838ac35265c23b5c"},
|
||||
{file = "sqlalchemy-2.0.40-cp39-cp39-win32.whl", hash = "sha256:957f8d85d5e834397ef78a6109550aeb0d27a53b5032f7a57f2451e1adc37e98"},
|
||||
{file = "sqlalchemy-2.0.40-cp39-cp39-win_amd64.whl", hash = "sha256:1ffdf9c91428e59744f8e6f98190516f8e1d05eec90e936eb08b257332c5e870"},
|
||||
{file = "sqlalchemy-2.0.40-py3-none-any.whl", hash = "sha256:32587e2e1e359276957e6fe5dad089758bc042a971a8a09ae8ecf7a8fe23d07a"},
|
||||
{file = "sqlalchemy-2.0.40.tar.gz", hash = "sha256:d827099289c64589418ebbcaead0145cd19f4e3e8a93919a0100247af245fa00"},
|
||||
{file = "SQLAlchemy-2.0.41-cp37-cp37m-macosx_10_9_x86_64.whl", hash = "sha256:6854175807af57bdb6425e47adbce7d20a4d79bbfd6f6d6519cd10bb7109a7f8"},
|
||||
{file = "SQLAlchemy-2.0.41-cp37-cp37m-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:05132c906066142103b83d9c250b60508af556982a385d96c4eaa9fb9720ac2b"},
|
||||
{file = "SQLAlchemy-2.0.41-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:8b4af17bda11e907c51d10686eda89049f9ce5669b08fbe71a29747f1e876036"},
|
||||
{file = "SQLAlchemy-2.0.41-cp37-cp37m-musllinux_1_2_aarch64.whl", hash = "sha256:c0b0e5e1b5d9f3586601048dd68f392dc0cc99a59bb5faf18aab057ce00d00b2"},
|
||||
{file = "SQLAlchemy-2.0.41-cp37-cp37m-musllinux_1_2_x86_64.whl", hash = "sha256:0b3dbf1e7e9bc95f4bac5e2fb6d3fb2f083254c3fdd20a1789af965caf2d2348"},
|
||||
{file = "SQLAlchemy-2.0.41-cp37-cp37m-win32.whl", hash = "sha256:1e3f196a0c59b0cae9a0cd332eb1a4bda4696e863f4f1cf84ab0347992c548c2"},
|
||||
{file = "SQLAlchemy-2.0.41-cp37-cp37m-win_amd64.whl", hash = "sha256:6ab60a5089a8f02009f127806f777fca82581c49e127f08413a66056bd9166dd"},
|
||||
{file = "sqlalchemy-2.0.41-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:b1f09b6821406ea1f94053f346f28f8215e293344209129a9c0fcc3578598d7b"},
|
||||
{file = "sqlalchemy-2.0.41-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:1936af879e3db023601196a1684d28e12f19ccf93af01bf3280a3262c4b6b4e5"},
|
||||
{file = "sqlalchemy-2.0.41-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:b2ac41acfc8d965fb0c464eb8f44995770239668956dc4cdf502d1b1ffe0d747"},
|
||||
{file = "sqlalchemy-2.0.41-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:81c24e0c0fde47a9723c81d5806569cddef103aebbf79dbc9fcbb617153dea30"},
|
||||
{file = "sqlalchemy-2.0.41-cp310-cp310-musllinux_1_2_aarch64.whl", hash = "sha256:23a8825495d8b195c4aa9ff1c430c28f2c821e8c5e2d98089228af887e5d7e29"},
|
||||
{file = "sqlalchemy-2.0.41-cp310-cp310-musllinux_1_2_x86_64.whl", hash = "sha256:60c578c45c949f909a4026b7807044e7e564adf793537fc762b2489d522f3d11"},
|
||||
{file = "sqlalchemy-2.0.41-cp310-cp310-win32.whl", hash = "sha256:118c16cd3f1b00c76d69343e38602006c9cfb9998fa4f798606d28d63f23beda"},
|
||||
{file = "sqlalchemy-2.0.41-cp310-cp310-win_amd64.whl", hash = "sha256:7492967c3386df69f80cf67efd665c0f667cee67032090fe01d7d74b0e19bb08"},
|
||||
{file = "sqlalchemy-2.0.41-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:6375cd674fe82d7aa9816d1cb96ec592bac1726c11e0cafbf40eeee9a4516b5f"},
|
||||
{file = "sqlalchemy-2.0.41-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:9f8c9fdd15a55d9465e590a402f42082705d66b05afc3ffd2d2eb3c6ba919560"},
|
||||
{file = "sqlalchemy-2.0.41-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:32f9dc8c44acdee06c8fc6440db9eae8b4af8b01e4b1aee7bdd7241c22edff4f"},
|
||||
{file = "sqlalchemy-2.0.41-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:90c11ceb9a1f482c752a71f203a81858625d8df5746d787a4786bca4ffdf71c6"},
|
||||
{file = "sqlalchemy-2.0.41-cp311-cp311-musllinux_1_2_aarch64.whl", hash = "sha256:911cc493ebd60de5f285bcae0491a60b4f2a9f0f5c270edd1c4dbaef7a38fc04"},
|
||||
{file = "sqlalchemy-2.0.41-cp311-cp311-musllinux_1_2_x86_64.whl", hash = "sha256:03968a349db483936c249f4d9cd14ff2c296adfa1290b660ba6516f973139582"},
|
||||
{file = "sqlalchemy-2.0.41-cp311-cp311-win32.whl", hash = "sha256:293cd444d82b18da48c9f71cd7005844dbbd06ca19be1ccf6779154439eec0b8"},
|
||||
{file = "sqlalchemy-2.0.41-cp311-cp311-win_amd64.whl", hash = "sha256:3d3549fc3e40667ec7199033a4e40a2f669898a00a7b18a931d3efb4c7900504"},
|
||||
{file = "sqlalchemy-2.0.41-cp312-cp312-macosx_10_13_x86_64.whl", hash = "sha256:81f413674d85cfd0dfcd6512e10e0f33c19c21860342a4890c3a2b59479929f9"},
|
||||
{file = "sqlalchemy-2.0.41-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:598d9ebc1e796431bbd068e41e4de4dc34312b7aa3292571bb3674a0cb415dd1"},
|
||||
{file = "sqlalchemy-2.0.41-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:a104c5694dfd2d864a6f91b0956eb5d5883234119cb40010115fd45a16da5e70"},
|
||||
{file = "sqlalchemy-2.0.41-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:6145afea51ff0af7f2564a05fa95eb46f542919e6523729663a5d285ecb3cf5e"},
|
||||
{file = "sqlalchemy-2.0.41-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:b46fa6eae1cd1c20e6e6f44e19984d438b6b2d8616d21d783d150df714f44078"},
|
||||
{file = "sqlalchemy-2.0.41-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:41836fe661cc98abfae476e14ba1906220f92c4e528771a8a3ae6a151242d2ae"},
|
||||
{file = "sqlalchemy-2.0.41-cp312-cp312-win32.whl", hash = "sha256:a8808d5cf866c781150d36a3c8eb3adccfa41a8105d031bf27e92c251e3969d6"},
|
||||
{file = "sqlalchemy-2.0.41-cp312-cp312-win_amd64.whl", hash = "sha256:5b14e97886199c1f52c14629c11d90c11fbb09e9334fa7bb5f6d068d9ced0ce0"},
|
||||
{file = "sqlalchemy-2.0.41-cp313-cp313-macosx_10_13_x86_64.whl", hash = "sha256:4eeb195cdedaf17aab6b247894ff2734dcead6c08f748e617bfe05bd5a218443"},
|
||||
{file = "sqlalchemy-2.0.41-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:d4ae769b9c1c7757e4ccce94b0641bc203bbdf43ba7a2413ab2523d8d047d8dc"},
|
||||
{file = "sqlalchemy-2.0.41-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:a62448526dd9ed3e3beedc93df9bb6b55a436ed1474db31a2af13b313a70a7e1"},
|
||||
{file = "sqlalchemy-2.0.41-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:dc56c9788617b8964ad02e8fcfeed4001c1f8ba91a9e1f31483c0dffb207002a"},
|
||||
{file = "sqlalchemy-2.0.41-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:c153265408d18de4cc5ded1941dcd8315894572cddd3c58df5d5b5705b3fa28d"},
|
||||
{file = "sqlalchemy-2.0.41-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:4f67766965996e63bb46cfbf2ce5355fc32d9dd3b8ad7e536a920ff9ee422e23"},
|
||||
{file = "sqlalchemy-2.0.41-cp313-cp313-win32.whl", hash = "sha256:bfc9064f6658a3d1cadeaa0ba07570b83ce6801a1314985bf98ec9b95d74e15f"},
|
||||
{file = "sqlalchemy-2.0.41-cp313-cp313-win_amd64.whl", hash = "sha256:82ca366a844eb551daff9d2e6e7a9e5e76d2612c8564f58db6c19a726869c1df"},
|
||||
{file = "sqlalchemy-2.0.41-cp38-cp38-macosx_10_9_x86_64.whl", hash = "sha256:90144d3b0c8b139408da50196c5cad2a6909b51b23df1f0538411cd23ffa45d3"},
|
||||
{file = "sqlalchemy-2.0.41-cp38-cp38-macosx_11_0_arm64.whl", hash = "sha256:023b3ee6169969beea3bb72312e44d8b7c27c75b347942d943cf49397b7edeb5"},
|
||||
{file = "sqlalchemy-2.0.41-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:725875a63abf7c399d4548e686debb65cdc2549e1825437096a0af1f7e374814"},
|
||||
{file = "sqlalchemy-2.0.41-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:81965cc20848ab06583506ef54e37cf15c83c7e619df2ad16807c03100745dea"},
|
||||
{file = "sqlalchemy-2.0.41-cp38-cp38-musllinux_1_2_aarch64.whl", hash = "sha256:dd5ec3aa6ae6e4d5b5de9357d2133c07be1aff6405b136dad753a16afb6717dd"},
|
||||
{file = "sqlalchemy-2.0.41-cp38-cp38-musllinux_1_2_x86_64.whl", hash = "sha256:ff8e80c4c4932c10493ff97028decfdb622de69cae87e0f127a7ebe32b4069c6"},
|
||||
{file = "sqlalchemy-2.0.41-cp38-cp38-win32.whl", hash = "sha256:4d44522480e0bf34c3d63167b8cfa7289c1c54264c2950cc5fc26e7850967e45"},
|
||||
{file = "sqlalchemy-2.0.41-cp38-cp38-win_amd64.whl", hash = "sha256:81eedafa609917040d39aa9332e25881a8e7a0862495fcdf2023a9667209deda"},
|
||||
{file = "sqlalchemy-2.0.41-cp39-cp39-macosx_10_9_x86_64.whl", hash = "sha256:9a420a91913092d1e20c86a2f5f1fc85c1a8924dbcaf5e0586df8aceb09c9cc2"},
|
||||
{file = "sqlalchemy-2.0.41-cp39-cp39-macosx_11_0_arm64.whl", hash = "sha256:906e6b0d7d452e9a98e5ab8507c0da791856b2380fdee61b765632bb8698026f"},
|
||||
{file = "sqlalchemy-2.0.41-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:a373a400f3e9bac95ba2a06372c4fd1412a7cee53c37fc6c05f829bf672b8769"},
|
||||
{file = "sqlalchemy-2.0.41-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:087b6b52de812741c27231b5a3586384d60c353fbd0e2f81405a814b5591dc8b"},
|
||||
{file = "sqlalchemy-2.0.41-cp39-cp39-musllinux_1_2_aarch64.whl", hash = "sha256:34ea30ab3ec98355235972dadc497bb659cc75f8292b760394824fab9cf39826"},
|
||||
{file = "sqlalchemy-2.0.41-cp39-cp39-musllinux_1_2_x86_64.whl", hash = "sha256:8280856dd7c6a68ab3a164b4a4b1c51f7691f6d04af4d4ca23d6ecf2261b7923"},
|
||||
{file = "sqlalchemy-2.0.41-cp39-cp39-win32.whl", hash = "sha256:b50eab9994d64f4a823ff99a0ed28a6903224ddbe7fef56a6dd865eec9243440"},
|
||||
{file = "sqlalchemy-2.0.41-cp39-cp39-win_amd64.whl", hash = "sha256:5e22575d169529ac3e0a120cf050ec9daa94b6a9597993d1702884f6954a7d71"},
|
||||
{file = "sqlalchemy-2.0.41-py3-none-any.whl", hash = "sha256:57df5dc6fdb5ed1a88a1ed2195fd31927e705cad62dedd86b46972752a80f576"},
|
||||
{file = "sqlalchemy-2.0.41.tar.gz", hash = "sha256:edba70118c4be3c2b1f90754d308d0b79c6fe2c0fdc52d8ddf603916f83f4db9"},
|
||||
]
|
||||
|
||||
[package.dependencies]
|
||||
@@ -3231,4 +3251,4 @@ reference = "ali-mirrors"
|
||||
[metadata]
|
||||
lock-version = "2.0"
|
||||
python-versions = ">=3.11,<3.13"
|
||||
content-hash = "2b34e52cba181483fd7ab37d7d9ed92b4703aecc604befbd150d831e84de535d"
|
||||
content-hash = "b227dc263d83a0d98c3903fc0bca5763858aaeda6f6c14f83294afddd05679b4"
|
||||
|
||||
@@ -24,6 +24,8 @@ flask = "^3.1.1"
|
||||
psycopg2 = "^2.9.10"
|
||||
gunicorn = "^23.0.0"
|
||||
gevent = "^25.5.1"
|
||||
pymysql = "^1.1.1"
|
||||
sqlalchemy = "^2.0.41"
|
||||
[build-system]
|
||||
requires = ["poetry-core"]
|
||||
build-backend = "poetry.core.masonry.api"
|
||||
|
||||
@@ -118,64 +118,39 @@ def process_query(recognizer, query):
|
||||
}
|
||||
else:
|
||||
# 可以在这里添加延迟，避免过快重试
|
||||
time.sleep(10 * retry_count)
|
||||
time.sleep(10)
|
||||
|
||||
# 示例查询
|
||||
examples_query = """储能软件组合件界面，点击隐藏空项目划分后界面没有任何变化"""
|
||||
|
||||
def main():
|
||||
def save_results_to_excel(results, output_file, is_final=False):
|
||||
"""
|
||||
意图识别和问题改写示例
|
||||
将结果保存到Excel文件
|
||||
|
||||
Args:
|
||||
results: 结果列表
|
||||
output_file: 输出文件路径
|
||||
is_final: 是否为最终保存，如果是则使用完整文件名，否则添加临时标记
|
||||
|
||||
Returns:
|
||||
None
|
||||
"""
|
||||
# 过滤掉None值
|
||||
valid_results = [r for r in results if r is not None]
|
||||
|
||||
# 从环境变量中获取配置
|
||||
api_key = os.getenv("OPENAI_API_KEY")
|
||||
base_url = os.getenv("OPENAI_API_BASE")
|
||||
model_name = os.getenv("LLM_MODEL_NAME", "gpt-3.5-turbo")
|
||||
if not valid_results:
|
||||
logging.warning("没有有效结果可保存")
|
||||
return
|
||||
|
||||
# 初始化意图识别器
|
||||
recognizer = IntentRecognizer(api_key=api_key, base_url=base_url, model_name=model_name)
|
||||
# 创建DataFrame
|
||||
results_df = pd.DataFrame(valid_results)
|
||||
|
||||
# 读取提问数据
|
||||
current_dir = os.path.dirname(os.path.abspath(__file__))
|
||||
data_file = os.path.join(current_dir, "..", "..", "data", "excel", "400条提问意图分类数据-原始.xlsx")
|
||||
|
||||
# 检测是否为调试模式，调试模式下使用examples_query，否则从Excel读取
|
||||
|
||||
is_debug = hasattr(sys, 'gettrace') and sys.gettrace() is not None
|
||||
if is_debug:
|
||||
examples = examples_query.strip().split("\n")
|
||||
# 根据是否为最终保存确定文件名
|
||||
if not is_final:
|
||||
file_name, file_ext = os.path.splitext(output_file)
|
||||
temp_output_file = f"{file_name}_temp{file_ext}"
|
||||
else:
|
||||
examples = load_questions_from_excel(data_file)
|
||||
|
||||
if not is_debug:
|
||||
|
||||
max_workers = 10 # 减少并发数以避免API限制
|
||||
logging.info(f"共有 {len(examples)} 个问题需要处理，使用 {max_workers} 个并发线程")
|
||||
# 创建一个与输入顺序相同的结果列表
|
||||
results = [None] * len(examples)
|
||||
# 使用线程池进行并发处理
|
||||
with concurrent.futures.ThreadPoolExecutor(max_workers=max_workers) as executor:
|
||||
# 提交所有任务并记录它们的索引
|
||||
future_to_index = {}
|
||||
for idx, query in enumerate(examples):
|
||||
future = executor.submit(process_query, recognizer, query)
|
||||
future_to_index[future] = idx
|
||||
|
||||
# 使用tqdm显示进度条
|
||||
for future in tqdm(concurrent.futures.as_completed(future_to_index), total=len(examples), desc="处理进度"):
|
||||
idx = future_to_index[future]
|
||||
result = future.result()
|
||||
# 将结果放在与输入相同的位置
|
||||
results[idx] = result
|
||||
|
||||
# 将结果保存到Excel文件
|
||||
results_df = pd.DataFrame(results)
|
||||
|
||||
output_file = os.path.join(current_dir, "..", "..", "data", "excel", "测试提问数据_槽位填充结果.xlsx")
|
||||
temp_output_file = output_file
|
||||
|
||||
# 使用ExcelWriter设置格式
|
||||
with pd.ExcelWriter(output_file, engine='xlsxwriter') as writer:
|
||||
with pd.ExcelWriter(temp_output_file, engine='xlsxwriter') as writer:
|
||||
results_df.to_excel(writer, index=False, sheet_name='Sheet1')
|
||||
|
||||
# 获取工作簿和工作表对象
|
||||
@@ -195,12 +170,78 @@ def main():
|
||||
# 设置所有行高为20磅
|
||||
for i in range(len(results_df) + 1): # +1 是为了包括表头
|
||||
worksheet.set_row(i, 20)
|
||||
|
||||
logging.info(f"已保存{len(valid_results)}条结果至: {temp_output_file}")
|
||||
|
||||
# 示例查询
|
||||
examples_query = """"锁标签号:811621005858, 注册单位:惠州电力勘察设计院有限公司,软件名称:广东迁改导则2022, 注册号:BW278-83834-58155-58339.迁改导则是要另外下载安装软件吗？"
|
||||
|
||||
"""
|
||||
|
||||
def main():
|
||||
"""
|
||||
意图识别和问题改写示例
|
||||
"""
|
||||
|
||||
# 从环境变量中获取配置
|
||||
api_key = os.getenv("OPENAI_API_KEY")
|
||||
base_url = os.getenv("OPENAI_API_BASE")
|
||||
model_name = os.getenv("LLM_MODEL_NAME", "gpt-3.5-turbo")
|
||||
|
||||
# 初始化意图识别器
|
||||
recognizer = IntentRecognizer(api_key=api_key, base_url=base_url, model_name=model_name)
|
||||
|
||||
# 读取提问数据
|
||||
current_dir = os.path.dirname(os.path.abspath(__file__))
|
||||
data_file = os.path.join(current_dir, "..", "..", "data", "excel", "历史提问数据(dislike)_提问明确.xlsx")
|
||||
output_file = os.path.join(current_dir, "..", "..", "data", "excel", "测试提问数据_槽位填充结果.xlsx")
|
||||
|
||||
# 检测是否为调试模式，调试模式下使用examples_query，否则从Excel读取
|
||||
is_debug = hasattr(sys, 'gettrace') and sys.gettrace() is not None
|
||||
if is_debug:
|
||||
examples = examples_query.strip().split("\n")
|
||||
else:
|
||||
examples = load_questions_from_excel(data_file)
|
||||
|
||||
if not is_debug:
|
||||
max_workers = 20 # 减少并发数以避免API限制
|
||||
logging.info(f"共有 {len(examples)} 个问题需要处理，使用 {max_workers} 个并发线程")
|
||||
|
||||
# 创建一个与输入顺序相同的结果列表
|
||||
results = [None] * len(examples)
|
||||
batch_size = 100 # 每100条保存一次
|
||||
|
||||
# 使用线程池进行并发处理
|
||||
with concurrent.futures.ThreadPoolExecutor(max_workers=max_workers) as executor:
|
||||
# 提交所有任务并记录它们的索引
|
||||
future_to_index = {}
|
||||
for idx, query in enumerate(examples):
|
||||
future = executor.submit(process_query, recognizer, query)
|
||||
future_to_index[future] = idx
|
||||
|
||||
# 使用tqdm显示进度条
|
||||
completed = 0
|
||||
for future in tqdm(concurrent.futures.as_completed(future_to_index), total=len(examples), desc="处理进度"):
|
||||
idx = future_to_index[future]
|
||||
result = future.result()
|
||||
# 将结果放在与输入相同的位置
|
||||
results[idx] = result
|
||||
|
||||
completed += 1
|
||||
# 每处理batch_size条数据保存一次
|
||||
if completed % batch_size == 0:
|
||||
logging.info(f"已完成 {completed}/{len(examples)} 条，保存中间结果...")
|
||||
save_results_to_excel(results, output_file, is_final=False)
|
||||
|
||||
# 处理完所有数据后，保存最终结果
|
||||
save_results_to_excel(results, output_file, is_final=True)
|
||||
logging.info(f"所有处理完成，最终结果已保存至: {output_file}")
|
||||
else:
|
||||
for idx, query in enumerate(examples):
|
||||
if query.strip() == "":
|
||||
continue
|
||||
process_query(recognizer, query)
|
||||
|
||||
logging.info(f"处理完成，结果已保存至: {output_file}")
|
||||
|
||||
def setup_logging():
|
||||
# 配置日志输出到控制台
|
||||
logging.basicConfig(
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
from rag2_0.dify.workflow_chat import NewWorkflowChat
|
||||
from rag2_0.dify.dify_tool import NewWorkflowChat
|
||||
import pandas as pd
|
||||
from concurrent.futures import ThreadPoolExecutor
|
||||
from tqdm import tqdm
|
||||
|
||||
+304
-2
@@ -1,8 +1,17 @@
|
||||
#!/usr/bin/env python
|
||||
# -*- coding: utf-8 -*-
|
||||
import psycopg2
|
||||
from psycopg2 import sql
|
||||
import os
|
||||
import json
|
||||
from datetime import timezone, timedelta
|
||||
from concurrent.futures import ThreadPoolExecutor, as_completed
|
||||
from rag2_0.dify.dify_client import ChatClient
|
||||
from pydantic import BaseModel, Field
|
||||
from langchain.output_parsers import PydanticOutputParser
|
||||
|
||||
|
||||
class ContentSource(BaseModel):
|
||||
score: int = Field(description="相关性分数")
|
||||
reason: str = Field(description="评分理由")
|
||||
|
||||
class PgSql:
|
||||
"""
|
||||
@@ -219,6 +228,299 @@ class DifyTool:
|
||||
finally:
|
||||
dify_pgsql.close_connection()
|
||||
|
||||
class BaseWorkflowChat:
|
||||
"""
|
||||
工作流对话基类，封装了与Dify API交互的基本功能
|
||||
"""
|
||||
def __init__(self, api_key: str, base_url: str):
|
||||
"""
|
||||
初始化工作流对话基类
|
||||
|
||||
Args:
|
||||
api_key: Dify API的密钥
|
||||
base_url: Dify API的基础URL
|
||||
"""
|
||||
self.chat_client = ChatClient(api_key=api_key, base_url=base_url)
|
||||
self.content_source_parser = PydanticOutputParser(pydantic_object=ContentSource)
|
||||
|
||||
def create_chat_message(self, query: str):
|
||||
"""
|
||||
创建聊天消息
|
||||
|
||||
Args:
|
||||
query: 问题内容
|
||||
|
||||
Returns:
|
||||
tuple: (聊天响应, 消息ID)
|
||||
"""
|
||||
try:
|
||||
response = self.chat_client.create_chat_message(inputs={}, query=query, user="AutoTestDifyChat").json()
|
||||
return response, response["message_id"]
|
||||
except Exception as e:
|
||||
raise e
|
||||
|
||||
def calculate_score(self, query: str, content: str) -> int:
|
||||
"""
|
||||
使用LLM判断query与content之间的相关性分数
|
||||
|
||||
Args:
|
||||
query (str): 用户问题
|
||||
content (str): 检索内容
|
||||
|
||||
Returns:
|
||||
int: 相关性分数，1-10分，10代表完全相关，1代表完全不相关；-1表示评分失败
|
||||
"""
|
||||
from rag2_0.tool.ModelTool import OpenAiLLM
|
||||
|
||||
try:
|
||||
prompt = f"""你是一个专业的信息相关性评估助手。请根据以下标准对用户query和检索内容的相关性进行1-10评分（10=完全相关，1=完全不相关），并按指定格式输出JSON结果。
|
||||
|
||||
【评分标准】
|
||||
10分：完全契合，主题/意图完全一致且涵盖所有关键信息
|
||||
8-9分：高度相关，核心要素匹配但存在少量信息缺失
|
||||
6-7分：部分相关，涉及相同主题但存在重要信息缺失
|
||||
4-5分：弱相关，仅次要信息点匹配
|
||||
1-3分：完全不相关或信息冲突
|
||||
|
||||
【评估维度】
|
||||
1. 主题一致性：核心主题/意图的匹配程度
|
||||
2. 内容覆盖度：是否涵盖query的关键要素
|
||||
3. 信息准确性：是否存在矛盾/错误信息
|
||||
4. 细节丰富度：是否提供query要求的详细信息
|
||||
|
||||
【输出格式】
|
||||
{{
|
||||
"score": 评分,
|
||||
"reason": "简明扼要的评分理由（中文）"
|
||||
}}
|
||||
|
||||
【示例】
|
||||
query: "新冠疫苗的常见副作用"
|
||||
内容: "辉瑞疫苗常见反应包括注射部位疼痛（84.1%）、疲劳（62.9%）"
|
||||
输出: {{"score":8,"reason":"主题完全匹配，涵盖主要副作用但未提及发热等常见反应"}}
|
||||
|
||||
现在评估：
|
||||
query: "{query}"
|
||||
content: "{content}"
|
||||
"""
|
||||
api_key = os.getenv("OPENAI_API_KEY")
|
||||
base_url = os.getenv("OPENAI_API_BASE")
|
||||
model = os.getenv("LLM_MODEL_NAME")
|
||||
llm = OpenAiLLM(api_key=api_key, base_url=base_url, model=model)
|
||||
response = llm.invoke(user_prompt=prompt, need_retry=True)
|
||||
|
||||
# 解析JSON响应
|
||||
try:
|
||||
parsed_output = self.content_source_parser.parse(response.content)
|
||||
return parsed_output.score
|
||||
except Exception as e:
|
||||
return -1
|
||||
except Exception as e:
|
||||
return -1
|
||||
|
||||
def get_retrieve_info(self, query: str, outputs: dict) -> tuple:
|
||||
"""
|
||||
获取检索信息并计算分数
|
||||
|
||||
Args:
|
||||
query (str): 用户问题
|
||||
outputs (dict): 检索输出结果
|
||||
|
||||
Returns:
|
||||
tuple: (检索内容列表, 最高分, 最低分, 平均分)
|
||||
"""
|
||||
max_score = 0
|
||||
min_score = 10
|
||||
total_score = 0
|
||||
valid_scores = 0
|
||||
retrieve_content = []
|
||||
|
||||
# 使用线程池并发计算分数
|
||||
with ThreadPoolExecutor() as executor:
|
||||
# 创建任务列表
|
||||
future_to_content = {}
|
||||
for result in outputs["result"]:
|
||||
content = result["content"].strip()
|
||||
future = executor.submit(self.calculate_score, query=query, content=content)
|
||||
future_to_content[future] = content
|
||||
|
||||
# 收集结果
|
||||
for future in as_completed(future_to_content):
|
||||
content = future_to_content[future]
|
||||
score = future.result()
|
||||
content_title = content.split("\n")[0]
|
||||
|
||||
if score != -1:
|
||||
max_score = max(max_score, score)
|
||||
min_score = min(min_score, score)
|
||||
total_score += score
|
||||
valid_scores += 1
|
||||
|
||||
if content_title:
|
||||
retrieve_content.append(content_title + f"--相关性得分({score}分)")
|
||||
|
||||
avg_score = total_score / valid_scores if valid_scores > 0 else 0
|
||||
return retrieve_content, max_score, min_score, avg_score
|
||||
|
||||
|
||||
class NewWorkflowChat(BaseWorkflowChat):
|
||||
"""
|
||||
新工作流对话类，用于调用新工作流发送对话并解析获取相关数据
|
||||
"""
|
||||
def process_question(self, query: str) -> dict:
|
||||
"""
|
||||
处理问题，获取新工作流的回答和相关信息
|
||||
|
||||
Args:
|
||||
query: 问题内容
|
||||
|
||||
Returns:
|
||||
dict: 包含问题、回答和相关信息的字典
|
||||
"""
|
||||
response, message_id = self.create_chat_message(query)
|
||||
|
||||
if isinstance(response, str) and response.startswith("error:"):
|
||||
raise RuntimeError(f"create_chat_message 出错：{response}")
|
||||
|
||||
answer = response["answer"]
|
||||
workflow_info = self.get_workflow_info(query, message_id)
|
||||
|
||||
if workflow_info is None:
|
||||
return None
|
||||
|
||||
result = {
|
||||
"问题": query,
|
||||
"新流程答案": answer,
|
||||
"新问题改写": workflow_info["问题改写"],
|
||||
"新问题分类": workflow_info["问题分类"],
|
||||
"槽点信息": workflow_info["槽点信息"],
|
||||
"新检索词条": workflow_info["检索词条"],
|
||||
"检索内容": workflow_info["检索内容"],
|
||||
"message_id":message_id
|
||||
}
|
||||
|
||||
return result
|
||||
|
||||
def get_workflow_info(self, query: str, message_id: str) -> dict:
|
||||
"""
|
||||
获取新工作流的问题分类和检索信息
|
||||
|
||||
Args:
|
||||
query (str): 用户问题
|
||||
message_id (str): 新工作流的消息ID
|
||||
|
||||
Returns:
|
||||
dict: 包含问题分类结果的字典
|
||||
"""
|
||||
retrieve_title = []
|
||||
retrieve_content = []
|
||||
max_score = 0
|
||||
min_score = 0
|
||||
avg_score = 0
|
||||
rewrite_query = ""
|
||||
vertical_classification = ""
|
||||
sub_classification = ""
|
||||
slot_info = ""
|
||||
|
||||
try:
|
||||
message_info = DifyTool.get_message_debug_info_by_id(message_id=message_id)
|
||||
for workflow_node in message_info["workflow_node_executions_info"]:
|
||||
if workflow_node["title"] == "知识检索结果后处理":
|
||||
outputs = json.loads(workflow_node["outputs"])
|
||||
retrieve_title, max_score, min_score, avg_score = self.get_retrieve_info(query=query, outputs=outputs)
|
||||
retrieve_content = outputs["result"]
|
||||
elif workflow_node["title"] == "问题优化结果解析":
|
||||
outputs = json.loads(workflow_node["outputs"])
|
||||
rewrite_query = outputs["optimize_query"]
|
||||
llm_result_json = json.loads(workflow_node['inputs'])["llm_result"]
|
||||
json_result = json.loads(llm_result_json)
|
||||
vertical_classification = json_result['vertical_classification']
|
||||
sub_classification = json_result['sub_classification']
|
||||
slot_info = json.dumps(json_result["slot_filling"], ensure_ascii=False, indent=2)
|
||||
except Exception as e:
|
||||
raise e
|
||||
|
||||
return {
|
||||
"问题改写": rewrite_query,
|
||||
"检索词条": "\n".join(retrieve_title) if retrieve_title else "未检索知识库",
|
||||
"检索内容": retrieve_content,
|
||||
"问题分类": f"{vertical_classification} - {sub_classification}",
|
||||
"槽点信息": slot_info,
|
||||
|
||||
}
|
||||
|
||||
class OldWorkFlowChat(BaseWorkflowChat):
|
||||
"""
|
||||
旧工作流对话类，用于调用旧工作流发送对话并解析获取相关数据
|
||||
"""
|
||||
def process_question(self, query: str) -> dict:
|
||||
"""
|
||||
处理问题，获取旧工作流的回答和相关信息
|
||||
|
||||
Args:
|
||||
query: 问题内容
|
||||
|
||||
Returns:
|
||||
dict: 包含问题、回答和相关信息的字典
|
||||
"""
|
||||
response, message_id = self.create_chat_message(query)
|
||||
|
||||
if isinstance(response, str) and response.startswith("error:"):
|
||||
return None
|
||||
|
||||
answer = response["answer"]
|
||||
workflow_info = self.get_workflow_info(query, message_id)
|
||||
|
||||
if workflow_info is None:
|
||||
return None
|
||||
|
||||
result = {
|
||||
"问题": query,
|
||||
"旧流程答案": answer,
|
||||
"旧问题改写": workflow_info["问题改写"],
|
||||
"旧检索词条": workflow_info["检索词条"],
|
||||
"检索内容": workflow_info["检索内容"],
|
||||
"message_id":message_id
|
||||
}
|
||||
|
||||
return result
|
||||
|
||||
def get_workflow_info(self, query: str, message_id: str) -> dict:
|
||||
"""
|
||||
获取旧工作流的问题改写和检索信息
|
||||
|
||||
Args:
|
||||
query (str): 用户问题
|
||||
message_id (str): 旧工作流的消息ID
|
||||
|
||||
Returns:
|
||||
dict: 包含问题改写和检索信息的字典
|
||||
"""
|
||||
retrieve_title = []
|
||||
retrieve_content = []
|
||||
max_score = 0
|
||||
min_score = 0
|
||||
avg_score = 0
|
||||
rewrite_query = ""
|
||||
|
||||
try:
|
||||
message_info = DifyTool.get_message_debug_info_by_id(message_id=message_id)
|
||||
for workflow_node in message_info["workflow_node_executions_info"]:
|
||||
if workflow_node["title"] == "知识检索结果后处理":
|
||||
outputs = json.loads(workflow_node["outputs"])
|
||||
retrieve_title, max_score, min_score, avg_score = self.get_retrieve_info(query=query, outputs=outputs)
|
||||
retrieve_content = outputs["result"]
|
||||
elif workflow_node["title"] == "问题优化结果解析":
|
||||
outputs = json.loads(workflow_node["outputs"])
|
||||
rewrite_query = outputs["optimize_query"]
|
||||
except Exception as e:
|
||||
return None
|
||||
|
||||
return {
|
||||
"问题改写": rewrite_query,
|
||||
"检索词条": "\n".join(retrieve_title) if retrieve_title else "未检索知识库",
|
||||
"检索内容": retrieve_content,
|
||||
}
|
||||
|
||||
if __name__ == "__main__":
|
||||
try:
|
||||
|
||||
@@ -2,7 +2,8 @@
|
||||
# -*- coding: utf-8 -*-
|
||||
|
||||
import os
|
||||
from rag2_0.dify.dify_client import ChatClient, DifyClient
|
||||
from rag2_0.dify.dify_client import DifyClient
|
||||
from rag2_0.dify.dify_tool import NewWorkflowChat, OldWorkFlowChat
|
||||
import pandas as pd
|
||||
# 使用线程池并发执行
|
||||
from concurrent.futures import ThreadPoolExecutor, as_completed
|
||||
@@ -44,8 +45,9 @@ class DifyComparisonTester:
|
||||
max_workers: 最大工作线程数
|
||||
"""
|
||||
self.excel_path = excel_path
|
||||
self.old_chat = ChatClient(api_key=old_workflow_api_key, base_url=baseurl)
|
||||
self.new_chat = ChatClient(api_key=new_workflow_api_key, base_url=baseurl)
|
||||
# 使用NewWorkflowChat和OldWorkFlowChat代替ChatClient
|
||||
self.old_chat = OldWorkFlowChat(api_key=old_workflow_api_key, base_url=baseurl)
|
||||
self.new_chat = NewWorkflowChat(api_key=new_workflow_api_key, base_url=baseurl)
|
||||
|
||||
# 评判相关参数
|
||||
self.output_path = output_path or os.path.join(os.path.dirname(self.excel_path), "dify问答_综合评判结果.xlsx")
|
||||
@@ -78,13 +80,13 @@ class DifyComparisonTester:
|
||||
"""
|
||||
def get_old_answer():
|
||||
try:
|
||||
return self.old_chat.create_chat_message(inputs={}, query=q, user="AutoTestDifyChat").json()
|
||||
return self.old_chat.process_question(query=q)
|
||||
except Exception as e:
|
||||
return f"error: {str(e)}"
|
||||
|
||||
def get_new_answer():
|
||||
try:
|
||||
return self.new_chat.create_chat_message(inputs={}, query=q, user="AutoTestDifyChat").json()
|
||||
return self.new_chat.process_question(query=q)
|
||||
except Exception as e:
|
||||
return f"error: {str(e)}"
|
||||
|
||||
@@ -95,14 +97,15 @@ class DifyComparisonTester:
|
||||
try:
|
||||
old_result = future_old.result()
|
||||
new_result = future_new.result()
|
||||
old_message_id = old_result["message_id"]
|
||||
new_message_id = new_result["message_id"]
|
||||
|
||||
old_answer = old_result["answer"]
|
||||
new_answer = new_result["answer"]
|
||||
if isinstance(old_result, str) and old_result.startswith("error:"):
|
||||
return None, None
|
||||
if isinstance(new_result, str) and new_result.startswith("error:"):
|
||||
return None, None
|
||||
|
||||
except Exception as e:
|
||||
return None, None, None
|
||||
return {"问题": q, "旧流程答案": old_answer, "新流程答案": new_answer}, old_message_id, new_message_id
|
||||
return future_old, future_new
|
||||
|
||||
def find_wiki_link(self, query) -> str | None:
|
||||
"""
|
||||
@@ -407,22 +410,24 @@ content: "{content}"
|
||||
Returns:
|
||||
dict: 包含问题分类结果的字典
|
||||
"""
|
||||
retrieve_title=[]
|
||||
retrieve_content=[]
|
||||
max_score=0
|
||||
min_score=0
|
||||
avg_score=0
|
||||
rewrite_query=""
|
||||
vertical_classification=""
|
||||
sub_classification=""
|
||||
slot_info=""
|
||||
try:
|
||||
# 使用DifyTool直接获取消息信息
|
||||
new_message_info = DifyTool.get_message_debug_info_by_id(message_id=new_message_id)
|
||||
|
||||
# 初始化变量
|
||||
retrieve_title = []
|
||||
retrieve_content = []
|
||||
rewrite_query = ""
|
||||
vertical_classification = ""
|
||||
sub_classification = ""
|
||||
slot_info = ""
|
||||
|
||||
# 解析工作流节点信息
|
||||
for workflow_node in new_message_info["workflow_node_executions_info"]:
|
||||
if workflow_node["title"] == "知识检索结果后处理":
|
||||
outputs = json.loads(workflow_node["outputs"])
|
||||
retrieve_title, max_score, min_score, avg_score = self.get_retrieve_info(query=query, outputs=outputs)
|
||||
retrieve_content=outputs["result"]
|
||||
retrieve_content = outputs["result"]
|
||||
elif workflow_node["title"] == "问题优化结果解析":
|
||||
outputs = json.loads(workflow_node["outputs"])
|
||||
rewrite_query = outputs["optimize_query"]
|
||||
@@ -430,20 +435,21 @@ content: "{content}"
|
||||
json_result = json.loads(llm_result_json)
|
||||
vertical_classification = json_result['vertical_classification']
|
||||
sub_classification = json_result['sub_classification']
|
||||
slot_info=json.dumps(json_result["slot_filling"],ensure_ascii=False,indent=2)
|
||||
slot_info = json.dumps(json_result["slot_filling"], ensure_ascii=False, indent=2)
|
||||
except Exception as e:
|
||||
return None
|
||||
|
||||
return {
|
||||
"问题改写": rewrite_query,
|
||||
"检索词条": "\n".join(retrieve_title) if retrieve_title else "未检索知识库",
|
||||
"检索内容": retrieve_content,
|
||||
"问题分类": f"{vertical_classification} - {sub_classification}",
|
||||
"槽点信息":slot_info
|
||||
"槽点信息": slot_info
|
||||
}
|
||||
|
||||
def get_old_workflow_info(self, query:str, old_message_id:str) -> dict:
|
||||
"""
|
||||
获取新流程的问题分类
|
||||
获取旧流程的问题分类
|
||||
|
||||
Args:
|
||||
query (str): 用户问题
|
||||
@@ -452,24 +458,27 @@ content: "{content}"
|
||||
Returns:
|
||||
dict: 包含问题分类结果的字典
|
||||
"""
|
||||
retrieve_title=[]
|
||||
retrieve_content=[]
|
||||
max_score=0
|
||||
min_score=0
|
||||
avg_score=0
|
||||
rewrite_query=""
|
||||
try:
|
||||
# 使用DifyTool直接获取消息信息
|
||||
old_message_info = DifyTool.get_message_debug_info_by_id(message_id=old_message_id)
|
||||
|
||||
# 初始化变量
|
||||
retrieve_title = []
|
||||
retrieve_content = []
|
||||
rewrite_query = ""
|
||||
|
||||
# 解析工作流节点信息
|
||||
for workflow_node in old_message_info["workflow_node_executions_info"]:
|
||||
if workflow_node["title"] == "知识检索结果后处理":
|
||||
outputs = json.loads(workflow_node["outputs"])
|
||||
retrieve_title, max_score, min_score, avg_score = self.get_retrieve_info(query=query, outputs=outputs)
|
||||
retrieve_content=outputs["result"]
|
||||
retrieve_content = outputs["result"]
|
||||
elif workflow_node["title"] == "问题优化结果解析":
|
||||
outputs = json.loads(workflow_node["outputs"])
|
||||
rewrite_query = outputs["optimize_query"]
|
||||
except Exception as e:
|
||||
return None
|
||||
|
||||
return {
|
||||
"问题改写": rewrite_query,
|
||||
"检索词条": "\n".join(retrieve_title) if retrieve_title else "未检索知识库",
|
||||
@@ -512,13 +521,13 @@ content: "{content}"
|
||||
dict: 包含问题、回答和评判结果的字典
|
||||
"""
|
||||
# 获取基本的问题和回答
|
||||
basic_result, old_message_id, new_message_id = self.process_question(q)
|
||||
if basic_result is None:
|
||||
future_old, future_new = self.process_question(q)
|
||||
if future_old is None or future_new is None:
|
||||
return None
|
||||
|
||||
query = basic_result["问题"]
|
||||
old_answer = basic_result["旧流程答案"]
|
||||
new_answer = basic_result["新流程答案"]
|
||||
query = future_old["问题"]
|
||||
old_answer = future_old["旧流程答案"]
|
||||
new_answer = future_new["新流程答案"]
|
||||
|
||||
# 获取词条链接和标准答案
|
||||
wiki_url = self.find_wiki_link(query)
|
||||
@@ -541,32 +550,22 @@ content: "{content}"
|
||||
if judge_result is None:
|
||||
judge_result = ""
|
||||
|
||||
# 并行获取新旧流程信息
|
||||
with ThreadPoolExecutor(max_workers=2) as executor:
|
||||
future_new = executor.submit(self.get_new_workflow_info, query=query, new_message_id=new_message_id)
|
||||
future_old = executor.submit(self.get_old_workflow_info, query=query, old_message_id=old_message_id)
|
||||
# retrieve_title_score = self.get_retrieve_title_similarity(old_retrieve_content=old_workflow_info["检索内容"], new_retrieve_content=new_workflow_info["检索内容"])
|
||||
|
||||
try:
|
||||
new_workflow_info = future_new.result()
|
||||
old_workflow_info = future_old.result()
|
||||
except Exception as e:
|
||||
print(f"处理问题 '{query}' 获取工作流信息时发生错误: {str(e)}")
|
||||
return None
|
||||
retrieve_title_score=self.get_retrieve_title_similarity(old_retrieve_content=old_workflow_info["检索内容"], new_retrieve_content=new_workflow_info["检索内容"])
|
||||
# 返回结果
|
||||
return {
|
||||
"问题": query,
|
||||
"新问题改写": new_workflow_info["问题改写"],
|
||||
"旧问题改写": old_workflow_info["问题改写"],
|
||||
"新问题分类": new_workflow_info["问题分类"],
|
||||
"槽点信息":new_workflow_info["槽点信息"],
|
||||
"新问题改写": future_new["问题改写"],
|
||||
"旧问题改写": future_old["问题改写"],
|
||||
"新问题分类": future_new["问题分类"],
|
||||
"槽点信息": future_new["槽点信息"],
|
||||
"新流程答案": new_answer,
|
||||
"旧流程答案": old_answer,
|
||||
"回答判断": judge_result,
|
||||
"词条检索相似度": retrieve_title_score,
|
||||
# "词条检索相似度": retrieve_title_score,
|
||||
"答案词条": answer_title if answer_title else "",
|
||||
"新检索词条": new_workflow_info["检索词条"],
|
||||
"旧检索词条": old_workflow_info["检索词条"],
|
||||
"新检索词条": future_new["检索词条"],
|
||||
"旧检索词条": future_old["检索词条"],
|
||||
}
|
||||
|
||||
def run_comparison(self, with_judge=False):
|
||||
@@ -670,5 +669,7 @@ if __name__ == "__main__":
|
||||
print(f"对比评判结果已保存至: {output_file}")
|
||||
|
||||
# 单个问题测试示例
|
||||
# c = DifyChat(baseurl="http://172.20.0.145/v1", api_key="app-LjJaeLoAfqa6aoGzqU9UvxSf")
|
||||
# c.chat("如何新建配电线路工程")
|
||||
# 使用新的工作流类进行测试
|
||||
# new_chat = NewWorkflowChat(api_key="app-qxsSybCs7ABiKlC1JabTYVn6", base_url="http://172.20.0.145/v1")
|
||||
# result = new_chat.process_question("如何新建配电线路工程")
|
||||
# print(json.dumps(result, ensure_ascii=False, indent=2))
|
||||
|
||||
@@ -1,310 +0,0 @@
|
||||
#!/usr/bin/env python
|
||||
# -*- coding: utf-8 -*-
|
||||
|
||||
import os
|
||||
import json
|
||||
from concurrent.futures import ThreadPoolExecutor, as_completed
|
||||
from rag2_0.dify.dify_client import ChatClient, DifyClient
|
||||
from rag2_0.dify.dify_tool import DifyTool
|
||||
from pydantic import BaseModel, Field
|
||||
from langchain.output_parsers import PydanticOutputParser
|
||||
from threading import Lock
|
||||
|
||||
class ContentSource(BaseModel):
|
||||
score: int = Field(description="相关性分数")
|
||||
reason: str = Field(description="评分理由")
|
||||
|
||||
class BaseWorkflowChat:
|
||||
"""
|
||||
工作流对话基类，封装了与Dify API交互的基本功能
|
||||
"""
|
||||
def __init__(self, api_key: str, base_url: str):
|
||||
"""
|
||||
初始化工作流对话基类
|
||||
|
||||
Args:
|
||||
api_key: Dify API的密钥
|
||||
base_url: Dify API的基础URL
|
||||
"""
|
||||
self.chat_client = ChatClient(api_key=api_key, base_url=base_url)
|
||||
self.content_source_parser = PydanticOutputParser(pydantic_object=ContentSource)
|
||||
|
||||
def create_chat_message(self, query: str):
|
||||
"""
|
||||
创建聊天消息
|
||||
|
||||
Args:
|
||||
query: 问题内容
|
||||
|
||||
Returns:
|
||||
tuple: (聊天响应, 消息ID)
|
||||
"""
|
||||
try:
|
||||
response = self.chat_client.create_chat_message(inputs={}, query=query, user="AutoTestDifyChat").json()
|
||||
return response, response["message_id"]
|
||||
except Exception as e:
|
||||
raise e
|
||||
|
||||
def calculate_score(self, query: str, content: str) -> int:
|
||||
"""
|
||||
使用LLM判断query与content之间的相关性分数
|
||||
|
||||
Args:
|
||||
query (str): 用户问题
|
||||
content (str): 检索内容
|
||||
|
||||
Returns:
|
||||
int: 相关性分数，1-10分，10代表完全相关，1代表完全不相关；-1表示评分失败
|
||||
"""
|
||||
from rag2_0.tool.ModelTool import OpenAiLLM
|
||||
|
||||
try:
|
||||
prompt = f"""你是一个专业的信息相关性评估助手。请根据以下标准对用户query和检索内容的相关性进行1-10评分（10=完全相关，1=完全不相关），并按指定格式输出JSON结果。
|
||||
|
||||
【评分标准】
|
||||
10分：完全契合，主题/意图完全一致且涵盖所有关键信息
|
||||
8-9分：高度相关，核心要素匹配但存在少量信息缺失
|
||||
6-7分：部分相关，涉及相同主题但存在重要信息缺失
|
||||
4-5分：弱相关，仅次要信息点匹配
|
||||
1-3分：完全不相关或信息冲突
|
||||
|
||||
【评估维度】
|
||||
1. 主题一致性：核心主题/意图的匹配程度
|
||||
2. 内容覆盖度：是否涵盖query的关键要素
|
||||
3. 信息准确性：是否存在矛盾/错误信息
|
||||
4. 细节丰富度：是否提供query要求的详细信息
|
||||
|
||||
【输出格式】
|
||||
{{
|
||||
"score": 评分,
|
||||
"reason": "简明扼要的评分理由（中文）"
|
||||
}}
|
||||
|
||||
【示例】
|
||||
query: "新冠疫苗的常见副作用"
|
||||
内容: "辉瑞疫苗常见反应包括注射部位疼痛（84.1%）、疲劳（62.9%）"
|
||||
输出: {{"score":8,"reason":"主题完全匹配，涵盖主要副作用但未提及发热等常见反应"}}
|
||||
|
||||
现在评估：
|
||||
query: "{query}"
|
||||
content: "{content}"
|
||||
"""
|
||||
api_key = os.getenv("OPENAI_API_KEY")
|
||||
base_url = os.getenv("OPENAI_API_BASE")
|
||||
model = os.getenv("LLM_MODEL_NAME")
|
||||
llm = OpenAiLLM(api_key=api_key, base_url=base_url, model=model)
|
||||
response = llm.invoke(user_prompt=prompt, need_retry=True)
|
||||
|
||||
# 解析JSON响应
|
||||
try:
|
||||
parsed_output = self.content_source_parser.parse(response.content)
|
||||
return parsed_output.score
|
||||
except Exception as e:
|
||||
return -1
|
||||
except Exception as e:
|
||||
return -1
|
||||
|
||||
def get_retrieve_info(self, query: str, outputs: dict) -> tuple:
|
||||
"""
|
||||
获取检索信息并计算分数
|
||||
|
||||
Args:
|
||||
query (str): 用户问题
|
||||
outputs (dict): 检索输出结果
|
||||
|
||||
Returns:
|
||||
tuple: (检索内容列表, 最高分, 最低分, 平均分)
|
||||
"""
|
||||
max_score = 0
|
||||
min_score = 10
|
||||
total_score = 0
|
||||
valid_scores = 0
|
||||
retrieve_content = []
|
||||
|
||||
# 使用线程池并发计算分数
|
||||
with ThreadPoolExecutor() as executor:
|
||||
# 创建任务列表
|
||||
future_to_content = {}
|
||||
for result in outputs["result"]:
|
||||
content = result["content"].strip()
|
||||
future = executor.submit(self.calculate_score, query=query, content=content)
|
||||
future_to_content[future] = content
|
||||
|
||||
# 收集结果
|
||||
for future in as_completed(future_to_content):
|
||||
content = future_to_content[future]
|
||||
score = future.result()
|
||||
content_title = content.split("\n")[0]
|
||||
|
||||
if score != -1:
|
||||
max_score = max(max_score, score)
|
||||
min_score = min(min_score, score)
|
||||
total_score += score
|
||||
valid_scores += 1
|
||||
|
||||
if content_title:
|
||||
retrieve_content.append(content_title + f"--相关性得分({score}分)")
|
||||
|
||||
avg_score = total_score / valid_scores if valid_scores > 0 else 0
|
||||
return retrieve_content, max_score, min_score, avg_score
|
||||
|
||||
|
||||
class NewWorkflowChat(BaseWorkflowChat):
|
||||
"""
|
||||
新工作流对话类，用于调用新工作流发送对话并解析获取相关数据
|
||||
"""
|
||||
def process_question(self, query: str) -> dict:
|
||||
"""
|
||||
处理问题，获取新工作流的回答和相关信息
|
||||
|
||||
Args:
|
||||
query: 问题内容
|
||||
|
||||
Returns:
|
||||
dict: 包含问题、回答和相关信息的字典
|
||||
"""
|
||||
response, message_id = self.create_chat_message(query)
|
||||
|
||||
if isinstance(response, str) and response.startswith("error:"):
|
||||
raise RuntimeError(f"create_chat_message 出错：{response}")
|
||||
|
||||
answer = response["answer"]
|
||||
workflow_info = self.get_workflow_info(query, message_id)
|
||||
|
||||
if workflow_info is None:
|
||||
return None
|
||||
|
||||
result = {
|
||||
"问题": query,
|
||||
"新流程答案": answer,
|
||||
"新问题改写": workflow_info["问题改写"],
|
||||
"新问题分类": workflow_info["问题分类"],
|
||||
"槽点信息": workflow_info["槽点信息"],
|
||||
"新检索词条": workflow_info["检索词条"],
|
||||
"检索内容": workflow_info["检索内容"],
|
||||
"message_id":message_id
|
||||
}
|
||||
|
||||
return result
|
||||
|
||||
def get_workflow_info(self, query: str, message_id: str) -> dict:
|
||||
"""
|
||||
获取新工作流的问题分类和检索信息
|
||||
|
||||
Args:
|
||||
query (str): 用户问题
|
||||
message_id (str): 新工作流的消息ID
|
||||
|
||||
Returns:
|
||||
dict: 包含问题分类结果的字典
|
||||
"""
|
||||
retrieve_title = []
|
||||
retrieve_content = []
|
||||
max_score = 0
|
||||
min_score = 0
|
||||
avg_score = 0
|
||||
rewrite_query = ""
|
||||
vertical_classification = ""
|
||||
sub_classification = ""
|
||||
slot_info = ""
|
||||
|
||||
try:
|
||||
message_info = DifyTool.get_message_debug_info_by_id(message_id=message_id)
|
||||
for workflow_node in message_info["workflow_node_executions_info"]:
|
||||
if workflow_node["title"] == "知识检索结果后处理":
|
||||
outputs = json.loads(workflow_node["outputs"])
|
||||
retrieve_title, max_score, min_score, avg_score = self.get_retrieve_info(query=query, outputs=outputs)
|
||||
retrieve_content = outputs["result"]
|
||||
elif workflow_node["title"] == "问题优化结果解析":
|
||||
outputs = json.loads(workflow_node["outputs"])
|
||||
rewrite_query = outputs["optimize_query"]
|
||||
llm_result_json = json.loads(workflow_node['inputs'])["llm_result"]
|
||||
json_result = json.loads(llm_result_json)
|
||||
vertical_classification = json_result['vertical_classification']
|
||||
sub_classification = json_result['sub_classification']
|
||||
slot_info = json.dumps(json_result["slot_filling"], ensure_ascii=False, indent=2)
|
||||
except Exception as e:
|
||||
raise e
|
||||
|
||||
return {
|
||||
"问题改写": rewrite_query,
|
||||
"检索词条": "\n".join(retrieve_title) if retrieve_title else "未检索知识库",
|
||||
"检索内容": retrieve_content,
|
||||
"问题分类": f"{vertical_classification} - {sub_classification}",
|
||||
"槽点信息": slot_info,
|
||||
|
||||
}
|
||||
|
||||
|
||||
class OldWorkFlowChat(BaseWorkflowChat):
|
||||
"""
|
||||
旧工作流对话类，用于调用旧工作流发送对话并解析获取相关数据
|
||||
"""
|
||||
def process_question(self, query: str) -> dict:
|
||||
"""
|
||||
处理问题，获取旧工作流的回答和相关信息
|
||||
|
||||
Args:
|
||||
query: 问题内容
|
||||
|
||||
Returns:
|
||||
dict: 包含问题、回答和相关信息的字典
|
||||
"""
|
||||
response, message_id = self.create_chat_message(query)
|
||||
|
||||
if isinstance(response, str) and response.startswith("error:"):
|
||||
return None
|
||||
|
||||
answer = response["answer"]
|
||||
workflow_info = self.get_workflow_info(query, message_id)
|
||||
|
||||
if workflow_info is None:
|
||||
return None
|
||||
|
||||
result = {
|
||||
"问题": query,
|
||||
"旧流程答案": answer,
|
||||
"旧问题改写": workflow_info["问题改写"],
|
||||
"旧检索词条": workflow_info["检索词条"],
|
||||
"检索内容": workflow_info["检索内容"],
|
||||
"message_id":message_id
|
||||
}
|
||||
|
||||
return result
|
||||
|
||||
def get_workflow_info(self, query: str, message_id: str) -> dict:
|
||||
"""
|
||||
获取旧工作流的问题改写和检索信息
|
||||
|
||||
Args:
|
||||
query (str): 用户问题
|
||||
message_id (str): 旧工作流的消息ID
|
||||
|
||||
Returns:
|
||||
dict: 包含问题改写和检索信息的字典
|
||||
"""
|
||||
retrieve_title = []
|
||||
retrieve_content = []
|
||||
max_score = 0
|
||||
min_score = 0
|
||||
avg_score = 0
|
||||
rewrite_query = ""
|
||||
|
||||
try:
|
||||
message_info = DifyTool.get_message_debug_info_by_id(message_id=message_id)
|
||||
for workflow_node in message_info["workflow_node_executions_info"]:
|
||||
if workflow_node["title"] == "知识检索结果后处理":
|
||||
outputs = json.loads(workflow_node["outputs"])
|
||||
retrieve_title, max_score, min_score, avg_score = self.get_retrieve_info(query=query, outputs=outputs)
|
||||
retrieve_content = outputs["result"]
|
||||
elif workflow_node["title"] == "问题优化结果解析":
|
||||
outputs = json.loads(workflow_node["outputs"])
|
||||
rewrite_query = outputs["optimize_query"]
|
||||
except Exception as e:
|
||||
return None
|
||||
|
||||
return {
|
||||
"问题改写": rewrite_query,
|
||||
"检索词条": "\n".join(retrieve_title) if retrieve_title else "未检索知识库",
|
||||
"检索内容": retrieve_content,
|
||||
}
|
||||
@@ -23,12 +23,12 @@ class SoftwareName(str, Enum):
|
||||
|
||||
# 软件别名映射
|
||||
ALIASES = {
|
||||
D3: ["配网D3", "D3软件", "配网工程软件"],
|
||||
C1: ["储能C1", "C1软件", "储能电站软件", "储能软件"],
|
||||
Z1: ["西藏Z1", "Z1软件", "西藏电力软件"],
|
||||
T1: ["技改T1", "T1软件", "技改检修软件"],
|
||||
T1_LIST: ["技改清单T1", "T1清单软件", "技改检修清单软件"],
|
||||
MAIN: ["主网软件", "电力建设软件", "主网建设软件", "主网软件"]
|
||||
D3: "别名包括：配网D3、D3软件、配网工程软件等 其他类似称呼",
|
||||
C1: "别名包括：储能C1、C1软件、储能电站软件、储能软件等 其他类似称呼",
|
||||
Z1: "别名包括：西藏Z1、Z1软件、西藏电力软件等 其他类似称呼",
|
||||
T1: "别名包括：技改T1、T1软件、技改检修软件等 其他类似称呼",
|
||||
T1_LIST: "别名包括：技改清单T1、T1清单软件、技改检修清单软件等 其他类似称呼",
|
||||
MAIN: "别名包括：主网软件、电力建设软件、主网建设软件、博微电力建设计价通等 其他类似称呼"
|
||||
}
|
||||
|
||||
# 定义输出模型
|
||||
@@ -58,7 +58,7 @@ class QueryRewrite(BaseModel):
|
||||
# 1. 软件问题
|
||||
# 1.1 软件功能
|
||||
class SoftwareFunction(BaseModel):
|
||||
software_name: SoftwareName = Field(description="软件名称")
|
||||
software_name: SoftwareName = Field(description="软件名称,只能从给定的范围中取值")
|
||||
function_name: str = Field(description="具体功能名称")
|
||||
operation: str = Field(description="用户操作意图(如何使用功能、功能入口、功能使用场景)")
|
||||
software_version: Optional[str] = Field(None, description="软件版本")
|
||||
@@ -77,7 +77,7 @@ class SoftwareFunction(BaseModel):
|
||||
|
||||
# 1.2 故障排查
|
||||
class TroubleShooting(BaseModel):
|
||||
software_name: SoftwareName = Field(description="软件名称")
|
||||
software_name: SoftwareName = Field(description="软件名称,只能从给定的范围中取值")
|
||||
function_name: str = Field(description="具体功能名称/操作描述")
|
||||
error_message: str = Field(description="报错信息/异常现象")
|
||||
software_version: Optional[str] = Field(None, description="软件版本")
|
||||
@@ -162,7 +162,7 @@ class SoftwareLock(BaseModel):
|
||||
# 3.3 安装下载类
|
||||
class InstallationDownload(BaseModel):
|
||||
|
||||
software_name: SoftwareName = Field(description="软件/插件名称，与file_name二选一")
|
||||
software_name: str = Field(description="软件/插件名称，与file_name二选一")
|
||||
file_name: str = Field(description="文件名，与software_name二选一")
|
||||
operation_stage: str = Field(description="操作阶段")
|
||||
os_version: Optional[str] = Field(None, description="操作系统版本")
|
||||
@@ -182,7 +182,7 @@ class InstallationDownload(BaseModel):
|
||||
# 3.4 问题排查类
|
||||
class ProblemDiagnosis(BaseModel):
|
||||
error_message: str = Field(description="报错信息/异常现象")
|
||||
software_name: Optional[SoftwareName] = Field(None, description="软件名称")
|
||||
software_name: Optional[SoftwareName] = Field(None, description="软件名称,只能从给定的范围中取值")
|
||||
os_version: Optional[str] = Field(None, description="操作系统版本")
|
||||
|
||||
def check_required_slots(self) -> Tuple[bool, Dict[str, str]]:
|
||||
|
||||
@@ -180,7 +180,7 @@ class IntentRecognizer:
|
||||
return reranked_terms
|
||||
|
||||
except Exception as e:
|
||||
return list(matched_terms)
|
||||
raise RuntimeError(f"SiliconFlowReRankerModel重排失败：{e}") from e
|
||||
|
||||
def match_keywords(self, query: str) -> Tuple[TermList, List[str]]:
|
||||
"""
|
||||
|
||||
@@ -18,26 +18,6 @@ import requests
|
||||
# sk-dvbaktabkdwdpjgxyoozlwnejosjyhdgqwllfeborqahndxs
|
||||
|
||||
API_KEY_LIST=[
|
||||
"sk-hrojkkkrrkmsajtnizokbcgexsfggdiqavbtvbayuwqbnmom",
|
||||
"sk-kkdklmnyompoiotzkfqahpayzlkgogfudjkyaebehtsowvid",
|
||||
"sk-sfxzvllifafbyfduupcdtcrjwhdyiyojnksyopnfslurnhsp",
|
||||
"sk-faqirxiszukfswqvzqawxnemqfacrkyurbxxkzwbbujqacdp",
|
||||
"sk-vonaanuueqiczppkntjuphateshrcpqpnvxmwxorkyihjmrb",
|
||||
"sk-qfpeoodgupcukcdstjcxgegwxnuhtxkkrupkogkcvhavxgny",
|
||||
"sk-fsvjnbpfgoadixympaabaukupuhjvbturcbxaqfdzjznemtr",
|
||||
"sk-fltvnbiqntfawjwkfnnhmyfiimzgzxkweqmefcfqkbucwrhi",
|
||||
"sk-oosswdriwyqkglwdigvcxgmcpyplcyowicbaugpizoscevdl",
|
||||
"sk-jswtxhkiralnyiukqimtyuurcaepulxdrfijadtxzrgsajyc",
|
||||
"sk-dcjuhoukdyrbneadtxtnyxzmigkpiqgtqqnreiprxpioftsv",
|
||||
"sk-yrhezyuxjblpaxzzudbowqmvcoxcammupcubghbodolikbdk",
|
||||
"sk-dsgvwpfagmarilmnewwbzhfzlqehburoupjaopucdvybpbdo",
|
||||
"sk-oljjlspuaurtoczyekztiidwtoerugadgepiufclpmrbdfqc",
|
||||
"sk-crgrimubjesthvxuqwedqqdoetljyrgeahxxpctfefgnkpyo",
|
||||
"sk-tubqhwgycxrdhwsqzjopxgeaqpsjdfppckckayvzornaluwq",
|
||||
"sk-amcxlmsdnadptpnehqnkvseolacipztmvovnmxojzohbjjil",
|
||||
"sk-pdyymhshpzmdduwxsezthnrgarnnhgzvmiflbpisfzxkiayt",
|
||||
"sk-qhwoorywmejumyudfxbrkegxtqifsbgcdkmpjckezepgyqnz",
|
||||
"sk-cpoctrgcnstaybeyuieuwjdgeakudhqdnnwdjavjudcbvvem",
|
||||
"sk-wqdpapdkisovziexgcyxvumpwzbjnhqbxvcqcspzctjhyhjk",
|
||||
"sk-bbntrnifrtdzhhgrtlrhvwbnaysuszviemshdakxonnnymnb",
|
||||
"sk-vmpnwjxersrwybmfhfxgsvbmhsmpjldxseiyxovnysrlbuzi",
|
||||
@@ -98,6 +78,25 @@ API_KEY_LIST=[
|
||||
"sk-nbksjgcngsayoumnsdbkcpnqivnvxjenwpzuazzrkhnsgeoo",
|
||||
"sk-iaafvpjyqiocgzchbdldbkgcffqniahkcbgoviuevuogulcm",
|
||||
"sk-muvjguqeshyimzowqnqgxwpsgujlpkqgrisxsimthtyrpypx",
|
||||
"sk-jgybgyayxlwoxeijgrjcneqlyusleohgbliuwpsuhocrjsmk",
|
||||
"sk-wzjsmwxcbbpcrqivqfzjwufqqjtlwejtncnvbpeicznkwiuh",
|
||||
"sk-izdjicdoyillktsihkiapuvwebisehtlgykozrvzfkgncwsc",
|
||||
"sk-fcsfmyivfuojsqsditvobfqprdpeunukycpcfnoxkraqevpx",
|
||||
"sk-szyjgyxrcvyxpvzfwgmbxnflxngxvcplitcctsdvvrqjgftk",
|
||||
"sk-jzbodthsnvjwbyrnynsxrudtqfnbdbrcxebjwjgajocnzqse",
|
||||
"sk-fxepossfzpmccibfwqpkluorzqlbtcaplepeugtfzfsctcbl",
|
||||
"sk-ympnflocrkxjrbubsxqdjqwicuyavvvysctlpfhunkcrzxjx",
|
||||
"sk-flhqvziknntednkcgjaxlyzzsrfzjhrzrmteqonajpbiinni",
|
||||
"sk-xfregpbbquqbxpiobjzanydsjivrjrnbokzxcqtnhxhyghhe",
|
||||
"sk-jrdzerhmvrtvzawkksowbgkggkubwfquplmrxbdhespqgtis",
|
||||
"sk-jjbpnkbeupsxyclcivbhizcfpfjrppddunbqynyjkqhtmpwu",
|
||||
"sk-oqehupcveovkjqqtxypqyifidcdissuyehwrkdwgruoyjkpq",
|
||||
"sk-orhfntzrbpmpavybcjyylofxncdvufdmvlznofmhxmnjymjl",
|
||||
"sk-kvgfuqeqvpmfsccykyoohheshclcrtvjlnewratvrjpkpbkc",
|
||||
"sk-zhnbqnpuumuuvegnvbgoggxafpukbzchpgrugpkobiwkzsar",
|
||||
"sk-kzhxlqvqcxlnbdgnpalqnzumkmspepkttkgbophnkqanainw",
|
||||
"sk-bzttugqtlskrvguvhckwamdssvgmgnrqpsialpdbskfsyyak",
|
||||
"sk-tovmogiablsoeabwgqyvevpcfichyjpuzqdymmvksspdrtqt",
|
||||
]
|
||||
|
||||
class APIKeyManager:
|
||||
|
||||
Reference in New Issue
Block a user