Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
keminghe
searching Neon…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
5 ms
·
1.
▲
by
keminghe
1y ago
You're absolutely right about the root cause being outdated AI knowledge bases/training data. I agree, my solution doesn't address that directly. Where this actually shines is with local LLMs (Ollama, etc) - smaller models, n
2.
▲
by
keminghe
1y ago
This is great feedback. I just improved the README and DockerHub overview to be more clear: "Stop getting out-of-date Python package manager commands from your AI. Cross-reference latest official pip, poetry, uv, and conda docs with au
3.
▲
by
keminghe
1y ago
Thank you so much for pointing that out, I just updated the docker instructions in the README and on DockerHub: ```shell # Pin to commit hash for production security # Get current hash from: https://hub.docker.com/r/kem
4.
▲
by
keminghe
1y ago
I see, yes, with this implementation, the networking component becomes necessary. I made different design decisions for `py-dep-man-companion` such that the tool is fully offline and can be used with local llms.
5.
▲
by
keminghe
1y ago
Genuinely curious: what are aspects of the demo you find less helpful? I want to improve it.
6.
▲
by
keminghe
1y ago
Appreciate you opening my eyes to this. Dash is indeed comprehensive, and a much bigger initiative. I wonder how it's handling the documentation staleness issue? New docs are published every minute.
7.
▲
by
keminghe
1y ago
Thank you. Yes, 50% of the core value prop is the self-updating automation and the traditional fuzzy + full text search capabilities that disrupt the embedding-centric RAG paradigm. Plus, Tantivy (Rust-based) is fast.
8.
▲
by
keminghe
1y ago
Appreciate the feedback. I will make it my todo to try out your suggestion of comparing Claude Code with and without MCP to measure the quantitative difference.
9.
▲
Show HN: Self-updating MCP server for official pip, uv, poetry and conda docs
(github.com)
30 points
by
keminghe
1y ago
|
18 comments
10.
▲
by
keminghe
1y ago
I will admit that the tool itself is still a bit janky (had to make a lot of workarounds with Streamlit). Will def continue to work with my professor for polish and additional features.
11.
▲
Show HN: Strain Seer – From demo-day hack to engineering blueprint
(github.com)
2 points
by
keminghe
1y ago
|
1 comments