| # ============================================================ | |
| # 环境变量配置 | |
| # .env 配置 | |
| OCR_ENGINE=api # 改为 api 模式 | |
| OCR_API_BASE=http://127.0.0.1:8015/v1 # vLLM 服务地址 | |
| OCR_API_MODEL=AXERA-TECH/PaddleOCR-VL-1.5 # 模型名 | |
| OCR_API_KEY=not-needed | |
| OCR_TASK=ocr # 任务类型 | |
| EMBEDDING_MODEL_NAME=AXERA-TECH/Qwen3-Embedding-0.6B | |
| EMBEDDING_API_BASE=http://127.0.0.1:8014/v1 | |
| EMBEDDING_API_KEY=sk-08ab126e77f04a0c99bb30154ab0876f | |
| EMBEDDING_BATCH_SIZE=4 | |
| LLM_API_BASE =http://127.0.0.1:8013/v1 | |
| LLM_API_KEY = not-needed | |
| LLM_MODEL_NAME =AXERA-TECH/Qwen3-1.7B-GPTQ-Int4 | |
| LLM_TEMPERATURE=0.1 | |
| LLM_MAX_TOKENS=2048 | |
| # ---- 向量数据库 ---- | |
| VECTOR_STORE_TYPE=chroma | |
| CHROMA_COLLECTION_NAME=pdf_ocr_knowledge | |
| # ---- PDF 渲染 ---- | |
| PDF_RENDER_DPI=300 | |
| MAX_FILE_SIZE_MB=50 | |
| # ---- 文本分割与检索 ---- | |
| CHUNK_SIZE=800 | |
| CHUNK_OVERLAP=150 | |
| RETRIEVAL_TOP_K=5 | |
| # ---- 日志 ---- | |
| LOG_LEVEL=INFO | |
| OCR_VL_BACKEND=native | |
| OCR_USE_LAYOUT=false | |
| OCR_LAYOUT_THRESHOLD=0.5 | |
| OCR_USE_CHART=false | |
| OCR_MAX_NEW_TOKENS=4096 | |
| OCR_TEMPERATURE=0.0 |