LMWiki
An industrial-grade knowledge base application built on DeepSeek Harness. It includes document management, a dual-form knowledge graph, traceable question answering, and modules for online evaluation and optimization.
Timeline (last 90 days)
In a DeepSeek Harness practical tutorial, a Bilibili creator demonstrated building an industrial-grade knowledge base LMWiki from scratch using only 10 rounds of prompts, 8 Skills, 1 Agent preset, and 2 feature plugins.
The LMWiki knowledge base has functions including document management, compilation, upload, viewing Wiki compilation results, incremental document compilation, dual-form knowledge graph viewing, and traceable Q&A.
LMWiki has a built-in online evaluation and optimization module, which can reverse-optimize the compilation method and retrieval process based on metrics such as retrieval hit rate and evidence coverage, achieving continuous self-iteration.
The Skills layer used by the LMWiki project is a curated combination extracted from a library of more than 400 Skills in the 'Fufan Agent' provided to paying students.
The entire development of the LMWiki project was divided into 11 stages, advancing strictly according to the standard process of 'environment setup -> project foundation development -> core functional module development -> feature acceptance'.
The LMWiki knowledge base has officially launched on the team's self-developed DeepSeek Harness Studio desktop client, usable with a single click; the full set of learning materials (installation package, project source code, Agent preset, development knowledge base, Skills and architecture diagrams, round-by-round prompts, conversation run screenshots, and phased results screenshots) has been released to the relevant community.
All Facts (6)
The LMWiki knowledge base has functions including document management, compilation, upload, viewing Wiki compilation results, incremental document compilation, dual-form knowledge graph viewing, and traceable Q&A.
50%UnverifiedLMWiki has a built-in online evaluation and optimization module, which can reverse-optimize the compilation method and retrieval process based on metrics such as retrieval hit rate and evidence coverage, achieving continuous self-iteration.
50%UnverifiedThe Skills layer used by the LMWiki project is a curated combination extracted from a library of more than 400 Skills in the 'Fufan Agent' provided to paying students.
50%UnverifiedThe entire development of the LMWiki project was divided into 11 stages, advancing strictly according to the standard process of 'environment setup -> project foundation development -> core functional module development -> feature acceptance'.
50%UnverifiedIn a DeepSeek Harness practical tutorial, a Bilibili creator demonstrated building an industrial-grade knowledge base LMWiki from scratch using only 10 rounds of prompts, 8 Skills, 1 Agent preset, and 2 feature plugins.
50%UnverifiedThe LMWiki knowledge base has officially launched on the team's self-developed DeepSeek Harness Studio desktop client, usable with a single click; the full set of learning materials (installation package, project source code, Agent preset, development knowledge base, Skills and architecture diagrams, round-by-round prompts, conversation run screenshots, and phased results screenshots) has been released to the relevant community.
50%