You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
@forloopcodes well you don't have a discussion page so this will do for now. Kinda wanted tool comparisons to be more common cus Smart Connections was a hit back then
@forloopcodes well you don't have a discussion page so this will do for now. Kinda wanted tool comparisons to be more common cus Smart Connections was a hit back then
Hey so vector search gets handled differently through peek, and we consider vectorization as long term storage/memory still for a vectorized store. Please see
So I “peeked” at foam and I see what you’re trying to give here. Could you elaborate further so we can discuss this more in depth if that’s alright. I do agree that what would happen as Foam would default to their link tree. I could be completely missing the mark since, again, I only peeked at that repo, so it would be good for the conversation to get the details from the expert who communicated their ideas effectively inside of their repo.
After pondering for a moment when it comes to the term “smart” connections…. Aka, linking the context coherently… I think what you’re throwing down, and I’ve written this in a preprint white paper, called “ topic intergration”
That is not my white paper but it Relates this formally for pull requests via semantics, but in order to get the full picture of that sort of graph you need clustering so that the model can remember that semantic pattern.
The above is my white paper that defines that type graph as clustering. So for example all of those .md files would get clustered into a superset topic simply known as “.md” or “ markdown” and then the next level of topics would be, essentially, “categories”. the major categories that emerge as a pattern of topic discussion, aka the the next level would be the topics/main-ideas that define the “Q”-category before the semantic search has to get fine granular.
@drQedwards thanks for the response, it is that Foam is a VSCode plugin mimicking Obsidian closely while being FOSS, BUT they do not have inherent "vault" support, or the crazy stuff from fuzzy search + graph traversal + vector search. So to find something that covers all of these points would be interesting (especially for agentic wiki that works along side codebase knowledge graphs)
@drQedwards thanks for the response, it is that Foam is a VSCode plugin mimicking Obsidian closely while being FOSS, BUT they do not have inherent "vault" support, or the crazy stuff from fuzzy search + graph traversal + vector search. So to find something that covers all of these points would be interesting (especially for agentic wiki that works along side codebase knowledge graphs)
Tell me more of the vault support. Is it like the atomic vaults one may find over at OpenSSL?
Okay so nothing like that… nothing crazy lol. I had to DYOR that with obisidian. I was recently trying to help assist with their woes of atomic operations over at OpenSSL.
They would have it: my .c file, nor anyone’s merges.
… so the answer is yes, there is support here. In fact maybe even atomic support for that obisidian vault.
I have made some "issues" on the future of this amazing work, speaking of how your node vectors may work in the memory graph architecture that has been collaborated upon.
Quick summary for anyone finding this later: Context+ builds its own local embedding + AST index (Tree-sitter symbol extraction, embeddings cached under .mcp_data, spectral clustering for semantic_navigate) and exposes it over MCP as graph-aware tools like get_blast_radius, semantic_navigate, and get_feature_hub. Obsidian Smart Connections is a note-vault semantic-link plugin; the overlap is vector search, but the focus here is cross-file code structure and usage tracing rather than note linking. Closing since the question's been answered, happy to keep discussing.
Activity
forloopcodes commentedon Mar 28, 2026
i'm sorry that's not an issue
though, i dont know about obsidian's smart connections, it generates and saves is own local config, and vector db of each file, thats how it works
TomLucidor commentedon Mar 29, 2026
@forloopcodes well you don't have a discussion page so this will do for now. Kinda wanted tool comparisons to be more common cus Smart Connections was a hit back then
forloopcodes commentedon Mar 29, 2026
i am going to look into this
drQedwards commentedon Apr 3, 2026
Hey so vector search gets handled differently through peek, and we consider vectorization as long term storage/memory still for a vectorized store. Please see
#28
For further details, please read PR #28
TomLucidor commentedon Apr 3, 2026
Compare and contrast with other designs bitbonsai/mcpvault#66
TomLucidor commentedon Apr 10, 2026
Foam might get basic MCP support but not vet vector/semantic search foambubble/foam#1521
drQedwards commentedon Apr 11, 2026
So I “peeked” at foam and I see what you’re trying to give here. Could you elaborate further so we can discuss this more in depth if that’s alright. I do agree that what would happen as Foam would default to their link tree. I could be completely missing the mark since, again, I only peeked at that repo, so it would be good for the conversation to get the details from the expert who communicated their ideas effectively inside of their repo.
drQedwards commentedon Apr 11, 2026
After pondering for a moment when it comes to the term “smart” connections…. Aka, linking the context coherently… I think what you’re throwing down, and I’ve written this in a preprint white paper, called “ topic intergration”
https://arxiv.org/abs/1710.10421
That is not my white paper but it Relates this formally for pull requests via semantics, but in order to get the full picture of that sort of graph you need clustering so that the model can remember that semantic pattern.
https://www.techrxiv.org/doi/full/10.36227/techrxiv.176523345.53664547/v1
The above is my white paper that defines that type graph as clustering. So for example all of those .md files would get clustered into a superset topic simply known as “.md” or “ markdown” and then the next level of topics would be, essentially, “categories”. the major categories that emerge as a pattern of topic discussion, aka the the next level would be the topics/main-ideas that define the “Q”-category before the semantic search has to get fine granular.
TomLucidor commentedon Apr 11, 2026
@drQedwards thanks for the response, it is that Foam is a VSCode plugin mimicking Obsidian closely while being FOSS, BUT they do not have inherent "vault" support, or the crazy stuff from fuzzy search + graph traversal + vector search. So to find something that covers all of these points would be interesting (especially for agentic wiki that works along side codebase knowledge graphs)
drQedwards commentedon Apr 12, 2026
Tell me more of the vault support. Is it like the atomic vaults one may find over at OpenSSL?
drQedwards commentedon Apr 12, 2026
Okay so nothing like that… nothing crazy lol. I had to DYOR that with obisidian. I was recently trying to help assist with their woes of atomic operations over at OpenSSL.
They would have it: my .c file, nor anyone’s merges.
… so the answer is yes, there is support here. In fact maybe even atomic support for that obisidian vault.
TomLucidor commentedon May 3, 2026
Bonus Q: how does ContextPlus store vectors? Cus the TODO shows that it needs GraphDB + VectorDB to be better https://github.com/forloopcodes/contextplus/blob/main/TODO.md?plain=1#L31
P.S. Some hope CoMfUcIoS/second-brain-mcp#10
drQedwards commentedon May 6, 2026
I have made some "issues" on the future of this amazing work, speaking of how your node vectors may work in the memory graph architecture that has been collaborated upon.
forloopcodes commentedon Jul 13, 2026
Quick summary for anyone finding this later: Context+ builds its own local embedding + AST index (Tree-sitter symbol extraction, embeddings cached under
.mcp_data, spectral clustering forsemantic_navigate) and exposes it over MCP as graph-aware tools likeget_blast_radius,semantic_navigate, andget_feature_hub. Obsidian Smart Connections is a note-vault semantic-link plugin; the overlap is vector search, but the focus here is cross-file code structure and usage tracing rather than note linking. Closing since the question's been answered, happy to keep discussing.