A-Series Disease Finder is an open-source Streamlit workbench for medical-data exploration, disease/ICD-10 selection, AI-assisted paper screening, local dataset inspection, and structured export.
A-Series 疾病查找器 是一个开源的 Streamlit 医学数据工作台，用于医学数据导入、疾病 / ICD-10 选择、AI 辅助文献筛选、本地数据探索与结构化导出。
⚠️ Research and educational use only. Not a medical device. Not a diagnostic system.
⚠️ 仅用于研究与学习，不是医疗器械，不是临床诊断系统。
A-Series Disease Finder helps users turn scattered medical research materials into a more organized workflow. It can import CSV/Excel tables and text documents, parse papers, select or normalize diseases with ICD-style information, screen papers with an AI backend, inspect loaded datasets, keep an audit trail, and export results as structured files.
A-Series 疾病查找器的目标，是把分散的医学数据、文献、疾病关键词和筛选结果整理进一个统一流程。它可以导入 CSV/Excel 数据表和文本资料，解析文献，选择或标准化疾病与 ICD 信息，用 AI 后端辅助筛选文献，查看本地数据表，保存审计日志，并把结果导出成结构化文件。
The project is useful when you want a local research assistant for questions like:
这个项目适合用来做这类事情：
- Which papers are relevant to a target disease?
- What tables and columns exist in my uploaded medical dataset?
- Which ICD-10-style disease entry should I use for a project?
- Can I package my screening results, included papers, metadata, and audit trail into one ZIP file?
- 哪些文献和目标疾病相关？
- 上传的医学数据表里有哪些字段？
- 某个研究项目应该使用哪个 ICD-10 疾病条目？
- 能不能把筛选结果、纳入文献、元数据和审计记录打包导出？
Upload and parse:
支持导入和解析：
- CSV / TSV
- Excel
.xlsx/.xls - TXT / Markdown
- JSON / JSONL
- DOCX
Tabular files are stored as local DataFrames; documents and papers are stored as local text entries.
表格文件会保存为本地 DataFrame；文档和论文会保存为本地文本条目。
A-Series includes a built-in ICD-style disease table and can optionally use AI to normalize disease names into structured disease configuration objects.
A-Series 内置 ICD 风格疾病表，也可以调用 AI 后端把疾病名称标准化成结构化疾病配置。
Upload papers or text files, set keywords and screening criteria, then use an OpenAI-compatible API, Ollama, or a local GGUF server to classify papers as included, excluded, or uncertain.
你可以上传论文或文本资料，设置关键词和筛选标准，然后使用 OpenAI-compatible API、Ollama 或本地 GGUF 服务，将文献分为纳入、排除或不确定。
View loaded tables, inspect row/column counts, filter columns, preview values, and export filtered data.
查看已导入的数据表，检查行列数量，筛选字段，预览样本值，并导出过滤结果。
The app records local audit events and maintains a simple cache for AI calls, so repeated operations are easier to inspect and reproduce.
应用会记录本地审计事件，并为 AI 调用维护简单缓存，方便检查和复现过程。
After paper screening or data preparation, export a ZIP package containing results, metadata, selected papers, and audit logs.
在文献筛选或数据整理完成后，可以导出包含结果、元数据、纳入文献和审计日志的 ZIP 包。
A-Series supports three backend styles:
A-Series 支持三类 AI 后端：
| Backend | Use case | 后端 | 用途 |
|---|---|---|---|
| OpenAI-compatible API | OpenAI, DeepSeek, Groq, local compatible gateways | OpenAI 兼容 API | OpenAI、DeepSeek、Groq 或本地兼容网关 |
| Ollama | Local model server at localhost:11434 |
Ollama | 本地 Ollama 模型服务 |
| Local GGUF / llama.cpp server | Local GGUF model inference | 本地 GGUF / llama.cpp server | 本地 GGUF 模型推理 |
The app can run in degraded mode if no AI backend is available. Some features will still work, but AI screening and normalization need a configured model.
如果没有可用 AI 后端，应用可以进入降级模式。部分基础功能仍可使用，但 AI 文献筛选和疾病标准化需要配置模型。
Recommended Python version:
推荐 Python 版本：
Python 3.10 or Python 3.11
install.bat
run.batchmod +x run.sh
./run.shgit clone https://github.com/<your-name>/A-Series-Disease-Finder.git
cd A-Series-Disease-Finder
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install -r requirements.txt
streamlit run app.pyThen open:
然后打开：
http://localhost:8501
python selfcheck.py
python smoke_test.py
streamlit run app.pyA release is considered minimally usable when:
当满足以下条件时，可以认为它达到最低可用状态：
- the Python files compile successfully;
- required project files exist;
- example files can be parsed;
- the Streamlit app can start;
- outputs and logs are readable rather than garbled;
- errors are understandable instead of silent crashes.
This tool may process medical or health-related data. You are responsible for what you upload and where your AI backend sends it.
这个工具可能处理医学或健康相关数据。你需要对自己上传的数据，以及 AI 后端会把数据发送到哪里负责。
Do not upload sensitive patient data, private health information, copyrighted medical documents, or confidential research files unless you have the legal right and technical safeguards to process them.
请不要随意上传敏感患者数据、个人健康信息、受版权保护的医学文档或保密研究文件，除非你明确拥有处理这些资料的权利，并且具备相应的技术保护措施。
This project does not provide medical advice. It must not be used to diagnose, treat, triage, or make clinical decisions.
本项目不提供医疗建议，不能用于诊断、治疗、分诊或临床决策。
A-Series-Disease-Finder/
├── app.py # Streamlit entry point
├── ai_engine/ # AI connector and prompt templates
├── core/ # datastore, cache, audit, task queue, config
├── icd_data/ # built-in ICD-style disease table
├── parsers/ # CSV, Excel, PDF, DOCX, TXT parsers
├── ui/ # Streamlit UI pages, theme, sidebar, i18n
├── examples/ # demo data and demo papers
├── docs/ # architecture, safety, workflow notes
├── tools/ # optional maintenance helpers
├── requirements.txt
├── selfcheck.py
└── smoke_test.py
This is an open-source preview release. The goal is to provide a working local research workflow, not a certified clinical product. Expect rough edges, especially around AI output quality, local model configuration, and data privacy workflows.
这是一个开源预览版本。它的目标是提供可运行的本地研究工作流，而不是认证临床产品。AI 输出质量、本地模型配置和数据隐私流程仍需要持续改进。
MIT License.