Rohit Ghumare comments

Results 224 comments of


                                            Rohit Ghumare

[Tracking] Cloud Native Sustainability Week 2024 London

Agenda is up: https://lu.ma/1yx3wzsk

[Tracking] Cloud Native Sustainability Week 2024 London

Any update on our blog? https://github.com/cncf/tag-env-sustainability/issues/567

providers: consolidate openai providers

Agreed — this would simplify both providers and reduce the surface area for embedding-vs-chat config divergence. Happy to take a PR. **Suggested shape** (open to alternatives): `src/providers/openai.ts` exports a single...

providers: consolidate openai providers

Picking this up after #240 lands. State of the world: - **#185** (your original) closed and superseded by **#240** (OpenAI-compatible LLM provider via raw fetch — covers OpenAI / DeepSeek...

Architecture: Migrate BM25 & Graph Search from In-Memory to SQLite

We might integrate this one: https://github.com/iii-hq/workers/tree/main/iii-database

Architecture: Migrate BM25 & Graph Search from In-Memory to SQLite

For integrating database worker, we need to migrate entire iii version to 0.11.7 or more. So, we should wait.

perf: improve graph searching by replacing BFS with Dijkstra's algorithm

Shipped in v0.9.19 via #463 — https://github.com/rohitg00/agentmemory/releases/tag/v0.9.19 Thanks @Tanmay-008 for the benchmark — the O(V·E) + `Array.shift()` profile was the real win on top of the weight-aware semantics.

fix(cli): accurately display bound viewer port on splash screen

Thanks @Tanmay-008 — exposing viewerPort + viewerSkipped on /livez is a clean way to make the splash truthful. Merged. Closes #521.

Architecture: Prevent Event Loop Freezes Caused by Heavy CPU-bound Tasks.

Thanks for the careful benchmark — the event-loop starvation profile is real and reproducible. A few directions before committing to worker_threads + SharedArrayBuffer (which is the right end-state for very...

Architecture: Prevent Event Loop Freezes Caused by Heavy CPU-bound Tasks.

Yes — go ahead. The 200ms→20-40ms result you got with sqlite-vec is exactly the kind of step-change that justifies skipping option (2) and (3) entirely for the current workload. Couple...