7 posts tagged with "Python"

Debugging a silent failure where a document sync task marked everything failed in a RAG knowledge base project — full writeup below.

TL;DR​

A shared method was refactored with a new parameter signature, but one caller was missed. The caller passed arguments under the old contract and threw TypeError — except the call sat inside a try/except that quietly funneled the exception into a failed counter. No crash, no ERROR in the logs, just a number ticking up. These "silent failures" are the hardest bugs to track down. Two fixes: grep all callers after a signature refactor; and make except blocks log or re-raise, never swallow silently.

While building AI Agent systems for clients, we found that different tasks require vastly different model capabilities and costs: vision models for image analysis, lightweight models for text completion, and local models for internal data queries. MCP (Model Context Protocol) turns each capability into an independent tool that AI clients invoke on demand.

TL;DR​

Build a custom MCP Server in 30 minutes with Python FastMCP, connecting any OpenAI-compatible API based on scenario and cost. This article demonstrates the full workflow using the Doubao vision model, with extension templates for text generation, image generation, TTS, and more.

Encountered this issue while building an AI Agent platform for a client: needed to support multiple search providers (Tavily, Serper, Brave, Bing) while ensuring tool call failures don't interrupt the Agent's conversation flow.

TL;DR​

1. Define SearchProvider abstract base class + SearchResult data model for unified interface and output
2. Each provider inherits the base class, implements search() method with field mapping
3. Key design: Return SearchResult with error info on failure, never raise exceptions

Encountered this issue while building a multi-service SaaS system. Documenting the root cause and solution.

TL;DR​

Don't use async with client.post() pattern with httpx.AsyncClient. Create the client first, then call methods: response = await client.post().

TL;DR​

Pydantic v2 no longer supports class Config. Use model_config = ConfigDict(from_attributes=True) instead. If your model has a field named model_config, you must rename it to avoid conflict with the reserved attribute.

在 RAG 知识库项目中调试查询结果返回格式问题，以下是完整排查过程。

TL;DR​

RAG /query 接口返回的 sources 字段只包含 metadata，没有每条来源的 similarity 分数。解决方案：在组装响应时，将 metadatas 和 distances 合并，计算 similarity = 1 - distance。

在 RAG 知识库项目中调试混合检索评分问题，以下是完整排查过程。

TL;DR​

Milvus 混合搜索（Dense + Sparse）有四个常见坑：空稀疏向量报错、Collection 未加载、sparse 格式错误、阈值过高。本文给出每个问题的最小修复代码。