18 Services from dev-5090-box.tailfb4c1f.ts.net

Scans a code snippet for common security vulnerabilities including SQL injection, XSS, authentication bypass, secret leaks, and unsafe deserialization.

Generates unit tests for a given code snippet, automatically matching the file's existing test framework.

Repairs malformed JSON (trailing commas, single quotes, unquoted keys) and returns only valid, parseable JSON.

Converts a plain-English description of a pattern into a regex and a matching test string

Explains a SQL query in plain English, describing what it does step by step

Takes a raw regex pattern and returns a plain-English explanation of what it matches and how it works.

Performs structured code review returning numbered findings across correctness, security, edge cases, and reuse dimensions.

Fingerprints or verifies an OpenAI-compatible inference endpoint to detect silent model substitution by the seller

Runs multi-step chain-of-thought reasoning and planning tasks using the Qwen3.8-27B abliterated (Heretic) model on a local RTX 5090 via vLLM, without content restrictions.

Refactors a code snippet while preserving its behavior, returning a diff and a written rationale explaining the changes.

Converts a curl command into equivalent Python, JavaScript, or Go code

Generates uncensored code completions using the Qwen3-8B 27B abliterated (Heretic) model served on a local RTX 5090 via vLLM

Explains a pasted code snippet in plain English using an uncensored local LLM running on an RTX 5090.

Scores any public URL on agent-readiness out of 100 with per-check evidence, published weights, and a deterministic recompute recipe

Fetches live weather data, routes it through two local LLM stages, then an independent judge verifies the result against source data before returning a validated response

Generates 768-dimensional vector embeddings for one or more input text strings

Sends a chat message to an uncensored Qwen3-27B (abliterated) model hosted on a local RTX 5090 and returns a free-form AI response without content filtering.

Accepts image-understanding requests and currently returns text-fallback responses (no vision model loaded); billed at a reduced rate reflecting the preview/fallback status.