Dong-Fang-Bu-Bai/btr-dualrag

选用指南

AI 根据 README 整理 · 2026/10/19 02:16:23

使用方式

通过命令行工具 btr-dualrag 使用，支持 ask 子命令进行查询，doctor 子命令检查环境。

查看 README 依据

BTR-MCL uses a separate environment based on `requirements/btr-mcl.source.txt`. See [Dependency environments](requirements/README.md) and [BTR-MCL service](services/btr_mcl_api/README.md). ## Use Copy `.env.local.example` to `.env.local`, replace the placeholders with local asset paths and service endpoints, then export the variables in the current shell. Check the configured environment: ```bash btr-dualrag doctor --profile functional_verified ``` Run a query: ```bash btr-dualrag ask --profile functional_verified \ "Which tools can calculate read depth for a BAM file?" ```

部署要求

需要 Python 环境，并安装 requirements/core.lock.txt 中的依赖。BTR-MCL 服务需要单独的环境，基于 requirements/btr-mcl.source.txt。大型运行时资产（如 BTVD/Chroma、BTKG/Neo4j、模型检查点等）不包含在仓库中，需要单独获取。

查看 README 依据

| `third_party/recbole/` | RecBole source required by BTR-MCL | | `docs/` | Architecture, asset, and paper-to-code notes | Large runtime assets are not included: BTVD/Chroma, BTKG/Neo4j, model checkpoints, BTR-MCL data files, local model caches, and evaluation datasets. See [Data and model assets](docs/data-and-models.md). ## Install Use a dedicated Python environment for the core package: ```bash python -m venv .venv-core python -m pip install -r requirements/core.lock.txt python -m pip install -e . --no-deps python -m unittest discover -s tests -q ```

BTR-MCL uses a separate environment based on `requirements/btr-mcl.source.txt`. See [Dependency environments](requirements/README.md) and [BTR-MCL service](services/btr_mcl_api/README.md). ## Use Copy `.env.local.example` to `.env.local`, replace the placeholders with local asset paths and service endpoints, then export the variables in the current shell. Check the configured environment: ```bash btr-dualrag doctor --profile functional_verified ``` Run a query: ```bash btr-dualrag ask --profile functional_verified \ "Which tools can calculate read depth for a BAM file?" ```

已知限制

大型运行时资产（如 BTVD/Chroma、BTKG/Neo4j、模型检查点、BTR-MCL 数据文件、本地模型缓存和评估数据集）不包含在仓库中。

查看 README 依据

| `third_party/recbole/` | RecBole source required by BTR-MCL | | `docs/` | Architecture, asset, and paper-to-code notes | Large runtime assets are not included: BTVD/Chroma, BTKG/Neo4j, model checkpoints, BTR-MCL data files, local model caches, and evaluation datasets. See [Data and model assets](docs/data-and-models.md). ## Install Use a dedicated Python environment for the core package: ```bash python -m venv .venv-core python -m pip install -r requirements/core.lock.txt python -m pip install -e . --no-deps python -m unittest discover -s tests -q ```