Workspaces1/10

Semantic Search Playground

Name: mydevtools
Author: mydevtools

Real embedding-based search over your own text, computed entirely in your browser — no server, no API key

Model loading downloads assets from external hosts

Why: Semantic search downloads embedding model files from Hugging Face before computing embeddings in your browser. The inference runtime is bundled locally.

Data sent:

Model asset requests with ordinary HTTP metadata; document text and search queries are not uploaded

Data handling: Asset hosts receive download requests and may cache or log them. Documents, embeddings and search queries are processed locally in page memory.

This app does not log, trace or save request/response payloads or uploaded files on its server. External APIs and storage providers apply their own retention policies; this app cannot delete their copies.

Audit the source: src/app/tools/rag-search/page.tsx

About 91 MB of FP32 model files download from Hugging Face on first use. Cached files can be reused while browser storage retains them. The inference runtime is bundled locally, and your text is processed on this device.

Documents

mydevtools is a free, privacy-first collection of developer utilities that run entirely in the browser. It includes formatters for JSON, XML, SQL, and YAML that beautify and validate structured data. The validators catch syntax errors before code ships to production. Diff tools compare JSON, XML, CSV, and plain text side by side with highlighted changes. Data converters translate between JSON, XML, CSV, and YAML without a round trip to a server. Cryptography tools cover hashing, HMAC signing, AES encryption, and JWT decoding, all computed locally with the Web Crypto API. Certificate and key tools generate CSRs, inspect X.509 certificates, and parse PKCS12 keystores. Networking tools include a CORS tester, DNS lookup, and subnet calculator. The API request builder supports curl import and Postman collection export for testing REST endpoints. Image tools compress, resize, and optimize SVGs directly on the client. The artificial intelligence category includes protocol inspectors for MCP and A2A agents, a token counter with real BPE tokenization, and a fine-tuning dataset validator for JSONL files.

Chunk size: 300 chars

Chunk overlap: 60 chars

Current settings produce 6 chunks from 1,111 characters.

Try:

how do I scramble secrets so nobody can read them

tools for looking at pictures

checking that my markup is well formed

figuring out which machine a hostname points at

Top K:5

Min similarity: 0.00

Compare vs keyword search

Learn More About This Tool