Streamline sparse cosine dot-product lookups#195
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
|
Just as a heads up, I was blocked by some firewall rules while working on your feedback. Expand below for details. Warning Firewall rules blocked me from connecting to one or more addresses (expand for details)I tried to connect to the following addresses, but was blocked by firewall rules:
If you need me to access, download, or install something from one of these locations, you can either:
|
Codex
AI
changed the title
madara88645
merged commit 6ff01af
into
bolt-tfidf-sparse-dict-intersection-5659582442749844468
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
There was a problem hiding this comment.
Pull request overview
This PR optimizes the TF‑IDF cosine similarity hot loop used by semantic chunking by reducing sparse-vector dot-product lookups from two dictionary operations (in + []) to a single sentinel-backed dict.get, and documents the pattern for future use.
Changes:
- Update
cosine_similarityto use a sentinel-backedv2.get(k, sentinel)during sparse dot-product accumulation. - Update
.jules/bolt.mdto recommend the single-lookup sparse-dict intersection pattern for performance-critical paths.
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.
| File | Description |
|---|---|
| app/rag/simple_index.py | Collapses membership + fetch into one dict lookup per key in the sparse dot-product loop. |
| .jules/bolt.md | Documents the sentinel-backed dict.get pattern as the preferred sparse intersection optimization. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
You can also share your feedback on Copilot code review. Take the survey.
| missing = object() | ||
| dot = 0.0 | ||
| for k, v in v1.items(): |
…ictionary intersection (#193) * Optimize TF-IDF cosine similarity dictionary intersection Avoid allocating expensive sets and performing set intersection when calculating cosine similarity for sparse dictionaries (like TF-IDF scores). Instead, iterate over the smaller dictionary items directly checking for key existence in the larger dictionary. This provides roughly a 30-40% speed-up in this calculation hot loop during chunking. * Optimize TF-IDF cosine similarity dictionary intersection Avoid allocating expensive sets and performing set intersection when calculating cosine similarity for sparse dictionaries (like TF-IDF scores). Instead, iterate over the smaller dictionary items directly checking for key existence in the larger dictionary. This provides roughly a 30-40% speed-up in this calculation hot loop during chunking. * Streamline sparse cosine dot-product lookups (#195) * Initial plan * Optimize sparse cosine similarity lookups --------- Co-authored-by: openai-code-agent[bot] <242516109+Codex@users.noreply.github.com> --------- Co-authored-by: Codex <242516109+Codex@users.noreply.github.com>
Frequent TF-IDF cosine similarity calls were still doing double dictionary lookups per key in the hot loop despite prior intersection optimization, adding needless hashing overhead.
getto collapse membership + fetch into one lookup while preserving zero-value correctness.Example: