madewithlaya

Catalog / Documents & data

0079GitHub

Chat Seek: search past Claude Code, Codex and OpenCode chats in VS Code, reranked by Laya

Describe a past conversation in plain language and Chat Seek finds it across local agent histories, reranks with the receptron Laya runtime, and offers a Resume action.

Open source ↗ github.comcostfree, localtime-
fstandhartinger/chat-seek-vscodeREADME ↗
**Point your coding agent at this repository and ask it to install Chat Seek in your local VS Code—see [Agent installation instructions](#agent-installation-instructions) below.**

# Chat Seek

Find past **Claude Code, Codex, and OpenCode** conversations from a description, or ask a question about what those chats contain, inside VS Code.



[Demo video](media/demo.mp4) · [How indexing and search work (narrated)](media/search-explainer.mp4) · [Original explainer](media/explainer.mp4). All videos use illustrative data.

Chat Seek searches local user and assistant messages, then uses the [Laya local decision model](https://github.com/receptron/laya) to rerank likely matches. Optional one-sentence AI descriptions help you recognize a chat. Cards show its last activity (“1 week ago”), with the exact date on hover, a readable excerpt, and a **Resume in Claude Code / Codex / OpenCode** action.

For factual questions, Chat Seek asks Laya to identify the intent, then searches full original messages in likely chats for the query's exact terms. If `rg` (ripgrep) is available, it also checks matching transcript files across the archive. This early pass can find text beyond the 2,800-character lookup limit. PIN and passcode questions stay entirely local: when a code is clearly attached to the named account, Chat Seek copies it from the source with an exact citation, without a model or API deciding which nearby code to use. Other likely chunks pass through Laya and optional answer extraction. The full pass still reads every supported chat from the newest chat and newest chunk onward. Matching chunks appear as provisional cards; verified answers move to the top, and rejected candidates disappear. Expand **Read full original chunk**, open it in an editor, or resume the chat. The scan 

Also filed under Documents & data

  1. 0104

    zero-shot-ie-bench: GLiNER 2.5, GLiFormer, GLiClass, Laya, Von and Jev in one web UI

    Seven zero-shot extraction and classification families demoed, benchmarked and cross-compared, with sizes, licences and costs in one table.

    @umstek · Documents & data · free for the local rows

  2. 0103

    Zero-shot benchmark: Laya's advertised win over Jev does not reproduce on four datasets

    Frozen protocol, raw predictions, calibration and latency for Jev 1.13, PrismNLI-0.4B and Laya. Laya ties Jev on dair-ai/emotion and trails by 10–33 points elsewhere.

    @elcronos · Documents & data

  3. 0102

    jevbench: Jev vs GPT-5-mini, Claude Sonnet 5, fine-tuned DistilBERT, BART NLI and Laya

    Six classifiers over SST-2, AG News and more, reporting accuracy, macro-F1, ECE, p50/p95 latency, throughput and cost side by side. Laya runs locally on MPS or CPU.

    @dhruvmehra · Documents & data · $0 for the local rows

  4. 0098

    Agentic GraphRAG over S&P 10-Ks with Laya as a shadow decision engine beside Jev

    Anthony Alcaraz's talk demo routes typed decisions through Jev and shows Laya's answers beside them. On the TSMC run Laya rejected every true claim; that's the point.

    @AnthonyAlcaraz · Documents & data · free (Laya shadow); Bedrock for reasoning