知识库 API
知识库、文件、搜索、分段与事件。
（待译）
Knowledge turns uploaded files into searchable knowledge bases with asynchronous ingestion and inspectable text segments. Concepts and behavior follow OpenAI vector stores (stores, file attachments, batches and search), so OpenAI vector stores experience applies. Differences: search has no next page (native results count chunks, while the compatible URL groups files); auto ranking does not rerank, so request default-2024-11-15 for reranking; chunk sizes use estimated tokens rather than a model tokenizer; ingestion accepts at most 100000000 bytes per file and also supports images and XLSX; native deletion retains a readable, restorable store for seven days. Start with praxis.knowledge.vectorStores.create, passing file_ids from File uploads to begin ingestion.
把文件加入知识库，按语义和关键词搜索相关分段。使用 API Key 作为 Bearer 凭据，建议从服务端调用。
https://api.deepraxis.ai/knowledge/v1